INFORMATION TO USERS
This material was produced from a microfilm copy of the original document. While the most advanced technological means to photograph and reproduce this document have been used, the quality is heavily dependent upon the quality of the original submitted.
The following explanation of techniques is provided to help you understand markings or patterns which may appear on this reproduction.
1. The sign or "target" for pages apparently lacking from the document photographed is "Missing Page(s)". If it was possible to obtain the mining page(s) or section, they are spliced into the film along with adjacent pages. This may have necessitated cutting thru an imaga and duplicating adjacent pages to insure you complete continuity.
2. When an imaga on the film is obliterated with a large round black mark, it is an indication that the photographer suspected that the copy may have moved during exposure and thus cause a blurred image. You will find a good image of the page in the adjacent frame.
3. When a map, drawing or chart, etc., was part of the material being photographed the photographer followed a definite method in "sectioning" the material. It is customary to begin photoing at the upper left hand corner of a large sheet and to continue photoing from left to right in equal sections with a small overlap. If necessary, sectioning is continued again — beginning below the first row and continuing on until complete.
4 . The majority of users indicate that the textual content is of greatest value, however, a somewhat higher quality reproduction could be made from "photographs" if essential to the understanding of the dissertation. Silver prints of "photographs" may be ordered at additional charge by writing the Order Department, giving the catalog number, title, author and specific pages you wish reproduced.
5 . PLEASE NOTE: Some pages may have indistinct print. Filmed as received.
Xerox University Microfilms 300 North Zoob Rood Ann Arbor. Michigan 40100 76-17,995
KELLY, Kevin Anthony, 1945- AN ON-AXIS FRESNEL ZONE PLATE IMAGING SYSTEM WITH A GERMANIUM GAMMA RAY CAMERA AND COMPUTER RECONSTRUCTION.
The Ohio State University, Ph.D., 1976 Engineering, nuclear
Xerox University Microfilms t Ann Arbor, Michigan 48100 AN ON-AXIS FRESNKL ZONE PLATE IMAGING SYSTEM
WITH A GERMANIUM GAMMA RAY CAMERA
AND COMPUTER RECONSTRUCTION
DISSERTATION
Presented in Partial Fulfillment of the Requirements
for the Degree Doctor of Philosophy in the
Graduate School of The Ohio State University
By Kevin A. Kelly, B.Sc., M.Sc.
The Ohio State University
1976
Reading Committee: Approved by:
Dr, Robert F. Redmond ^ Dr. Shoichiro Nakamura si ( / Dr. Philip A. Schlosser v / & & * * * < - Advisor Department of Nuclear Engineering Dedicated to
James and Anne Kelly,
my parents ACKNOWLEDGMENTS
Initially this research was funded by the National
Institutes of Health under contract NIH-NINDS-72-2323.
I wish to thank the Mechanical and Nuclear Engineering
Departments of The Ohio State University for their support during the latter part of this research. I am indebted to Dr. Arnold H. Deutchman of the Department of Nuclear
Medicine for several invaluable discussions of the role of gamma cameras in nuclear medicine, and for reviewing the rough draft of Chapter 7. I am grateful to Dr. Mark S.
Gerber, Dr. Philip A. Schlosser, and Mr. John W. Steidley
for discussions of the electronic and mechanical limi
tations on the spatial resolution of the orthogonal strip germanium camera. My sincerest thanks go to Ms. Carol
Edger for typing this dissertation under trying circum
stances. I wish to thank my reading committee, including
Dr. S. Nakamura, who taught me almost all I know about
nuclear reactor theory, and Dr. Philip Schlosser, who
directed the overall gamma camera research effort and
took an interest in me as a person as well as a student.
Above all, I would like to thank my advisor, Dr. Robert F.
Redmond, whose interest in me brought me to Ohio State.
Dr. Redmond suggested this research area and provided
direction and scope for the research. VITA
March 29, 194S Born - Croydon, United Kingdom
1967 B.Sc., Department of Physics, Uni versity of Notre Dame, Notre Dame, Indiana
1968 M.Sc. , Department of Physics, Yale University, New Haven, Connecticut
1968-1971 Peace Corps Volunteer, Ghana
1972-1975 Research Associate, Department of Nuclear Engineering, The Ohio State University, Columbus, Ohio
1974 M.Sc. , Department of Nuclear Engi neering, The Ohio State University, Columbus, Ohio TABLE OF CONTENTS
page
ACKNOWLEDGMENTS i• • n •
VITA iv
LIST OF TABLES .... ix
LIST OF FIGURES .... x
PARTIAL LIST OF SYMBOLS x* m > *
CHAPTER
1. INTRODUCTION . . .
1.1 Purpose of the Research 1 1.2 Rationale for the Research . . . 1 1.3 Goals of the Research ...... 3 1.4 Organization of the Dissertation 4
2 . RESEARCH MOTIVATION AND RELATED RESEARCH BY OTHER INVESTIGATORS . . . 5
2.1 Motivation for the Res earch 5
2 .1.1 Imaging Methods ...... 6 2 .1.2 Coded Apertures ...... 9 2 .1.3 Fresnel Zone Pl|ate Apertures 16 2.1.4 Optical Reconstruction . . 18 2.1.5 Digital Reconstruction . . 22
2.2 Related Research by Other Investigators 23
2.2.1 Shadowgrams with Other Detectors 23 2.2.2 Digital Reconstruction of Holo grams and Shadowgrams ...... 28
v TABLE OF CONTENTS (continued)
page ’
3. RELEVANT THEORY ...... 30
3.1 Optical Theory of Shadowgrams ...... 30 3.2 Gamma Ray Imaging by Shadowgrams...... 33 3.3 Digital Signal Processing Theory ...... 34 3.4 Image Enhancement Theory ...... 40
4. OUTLINE OF SYSTEM DESIGN ...... 44
4.1 The Germanium Detector ...... 44 4.2 The Fresnel Zone Plate S y s t e m ...... 45
4.2.1 Zone Plate Paramet ers...... 45 4.2.2 Relations Among Parameters .... 50 4.2.3 Method of Choosing Parameters . . . 56 4.2.4 Parameters for the Germanium Camera 64 4.2.5 Mechanical Components ...... 68
4.3 Image Processing and Display ...... 71 4.4 The Complete S y s t e m ...... 74
5. DIGITAL RECONSTRUCTION OF SHADOWGRAMS ...... 76
5.1 The Reconstruction Model ...... 76 5.2 Reconstruction of Optical Shadowgrams . . 79
5.2.1 Experimental Apparatus and P r o c e d u r e ...... 80 5.2.2 R e s u l t s ...... 85
5.3 Reconstruction of Computer-Generated Shadowgrams ...... 98
5.3.1 Generation of the Shadowgrams . . . 98 5.3.2 Point S o u r c e s ...... 99 5.3.3 Small Extended Sources ...... 102 5.3.4 Large Extended Sources ...... 107 S. 3.5 Extended Sources with Cold Spots . 108 5.3.6 Sources of Nonuniform Intensity . . 113 5.3.7 Small Strong Sources in a Weak Source Field ...... 116 5.3.8 Sources in Different Tomographic P l a n e s ...... 117
vi TABIiE OF CONTENTS (continued) page
5.4 Summary of Results ...... 125
6. IMAGE ENHANCEMENT STUDIES ...... 127
6.1 Introduction...... 127 6.2 A Fourier Transform Technique ...... 129
7. CLTNICAL ADVANTAGES AND LIMITATIONS OF THE SYSTEM ...... 143
7.1 Limitation on Source S i z e ...... 143 7.2 Imaging Whole Organs ...... 146 7.3 Imaging Small Lesions ...... 148 7.4 Conclusions ...... 151
8. SUMMARY OF RESULTS AND CONCLUSIONS ...... 153
8.1 Summary of the R e s e a r c h ...... 153 8.2 Summary of the Results ...... 154 8.3 Extensions of the R e sea rch...... 156 8.4 Conclusion...... 157
APPENDICES
A. OPTICAL THEORY OF FRESNEL ZONE PLATES AND HOLOGRAMS - AN OVERVIEW ...... 158
A.l The Frcsnel Zone P l a t e ...... 158
A. 1.1 I n t r o d u c t i o n ...... 158 A. 1.2 Basic Zone Plate T h e o r y * . 159 A.1.3 Applications of Zone Plates .... 173 A.1.4 Manufacturing of Zone Plates . . . 173
A.2 Holography ...... 174
A. 2.1 Basic H o l o g r a p h y ...... 174 A.2.2 History of Holography ...... 181
A.3 Relationship of Fresnel Zone Plates to H o l o g r a m s ...... 187
B. THEORY OF ZONE PLATE IMAGING OF GAMMA RAYS . . 190
vii TABLE OF CONTENTS (continued)
page
C. THE GERMANIUM GAMMA CAMERA ...... 196
C.l The Orthogonal Strip Detector ...... 196 C.2 Detector A r r a y ...... 199 C.3 E l e c t r o n i c s ...... 201 C.4 Mechanical Components ...... 204
D. THEORY OF DIGITAL RECONSTRUCTION OF SHADOWGRAMS 208
D.l The Diffraction Integ ral...... 208 D.2 The Fresnel Approximation ...... 213 D.3 The Fraunhofer Approximation ...... 216 D.4 The Reconstructed Image as a Fourier Transform...... 217 D.5 Application of Sampling Theory ...... 2X9 D.6 Digitizing the Reconstruction Integral . . 223 D.7 Reconstructing the True I m a g e ...... 226
BIBLIOGRAPHY . 229 LIST OF TABLES
Table page
4-1 Parameters of the Imaging System 50
4-2 System Parameters and Equations for Optimum System Resolution 63
4-3 System Parameters and Equations for a Specified Field of View 65
4-4 Zone Plate - Germanium Camera System Parameters 67
5-1 System Parameters for 1/8-inch Diameter Circular Source Shadowgram 86
5-2 Line Printer Code 89
5-3 System Parameters for Shadowgram of the Letter "E" 91
5-4 Relative Intensity of Weaker Source and Weaker Image 114
ix LIST OF FIGURES
Figure page 2.1 Conventional apertures 7 2.2 The Fresnel zone plate aperture 17
2.3 The optical reconstruction system 19
3.1 Recovery of a sampled waveform 38
4.1 A square detector of size Di divided into resolution elements of size Axi 47
4.2 Zone plate system field of view 48
4.3 The objcct-to-zone plate distance, si, and the zone plate-to-dctcctor distance, s2 49
4.4 The relationship between object plane resolution At and detector plane resolution AXj 53
4.5 Radiation from a point source in the object plane passing through the outermost zone 55
4.6 Relationships among the zone plate, detector and object plane parameters 57
4.7 Attenuation of gamma rays passing through the outermost zone 59
4.8 Mechanical components of the zone plate system 69
4.9 Block diagram of the complete imaging system 75
5.1 Experimental apparatus for casting optical shadowgrams 81
5.2 Image of a point source from an optical shadowgram 88
5.3 Imago of the letter "E" from an optical shadowgram 92
x LIST OF FIGURES (continued) Figure page
5.4 Alignment of "E" arms with rows of resolu tion elements 95
5.5 The source pattern for the "F." 104
5.6 The shadowgram for the "E" 105
5.7 The image of the "E** 106
5.8 Image of triangle and square focussed on plane 8.0 121
5.9 Image of triangle and square focussed on plane 8.5 122
5.10 Image of triangle and square focussed on plane 9.0 123
6.1 Results of transfer function calculation for a point source at the center of the field of view 136
6.2 Results of transfer function calculation for a point source in the corner of the field of view 137
6.3 Enhanced image of the letter "E" 140
A.l Spherical wavefront expanding from point 0 160
A. 2 Wavefront geometry 162
A.3 A Fresnel zone plate 166
A.4 Imaging an off-axis point source 167
A.5 Reference beam and signal arriving at the film . . 175
A.6 The diffracted beams 177
A.7 Hologram of a point scatterer 179
A.8 Reconstructing a point image 180
A.9 Hologram of a transparent object 185
xi LIST OF FIGURES (continued) Figure page
A. 10 Hologram of an opaque object 186
C.l Schematic of the orthogonal strip, charge splitting detector 198
C.2 Illustration of the 4 x 4 detector array 200
C.3 Schematic diagram of the electronics required to process signals emerging from one quadrant of the detector array 203
C.4 The camera assembly consisting of camera stand, camera head, and adjustable counter weights 205
C.5 Cutaway view of the camera head 206
D.l Coherent radiation incident on a shadowgram 210
D.2 The reconstruction geometry 212 PARTIAL LIST OF SYMBOLS
The symbols listed here are those which recurrently appear in the text.
Symbol Definition Equation
R„ Radius of the n-th Fresnel zone 3-1 n
Ar Width of the n-th Fresnel zone 3-2 n
SNR Signal-to-noise ratio 3-5
s^ Object-to-zone plate distance 4-1
&2 Zone plate-to-detector distance 4-1
a Magnification parameter 4-1
N, Number of resolution elements per L side 4-2
Length of square detector side 4-2
Ax-. Length of square resolution element 1 side 4-2
T Field of view 4-3
At Object plane resolution 4-3
N Number of zones in zone plate 4-4
D^j Diameter of the zone plate 4-4
tz Thickness of the zone plate 4-11
* « i li CHAPTER 1
INTRODUCTION
1.1 The Purpose of the Research
This study examines the feasibility of obtaining gamma ray camera images by means of an on-axis Fresnel zone plate aperture to produce shadowgrams and by digital reconstruction of images from the shadowgrams produced. A zone plate aperture system for a germanium gamma camera is designed and evaluated by means of computer studies.
The design is based on the characteristics of a high purity germanium camera currently being developed by the Nuclear
Engineering Department at The Ohio State University. The evaluation examines the ability of the system to provide images useful for the accurate diagnosis of disease in the nuclear medicine clinic.
1.2 Rationale for the Research
The current development of a high resolution germanium camera has prompted consideration of large area apertures for efficient collection of gamma radiation. The most useful such aperture, the Fresnel zone plate, results in 2 a pattern on the detector which resembles a hologram of the source intensity distribution. Normally the image of the source is recovered by beaming coherent laser light through a transparency of the holographic pattern. The resulting diffraction pattern is the image. A difficulty with this technique for the on-axis zone plate has been the obscuring of the diffraction pattern by the undiffractcd beam of light. This difficulty has prompted those researchers using X-ray film as the detection medium to use an off- axis zone plate in combination with a half-tone screen; however, the spatial resolution of this system is then degraded by a factor of three, and the efficiency by a factor of two. This solution is not acceptable for a ger manium gamma camera.
An alternative solution to this difficulty is con sidered here, the digital calculation of the diffraction pattern. Little previous work has been done with computer reconstruction of holograms because the number of resolu tion elements in an optical hologram is usually very large, and the calculation requires a great deal of computer time. The number of resolution elements in the germanium gamma camera considered here is relatively small, and the output of the camera is necessarily in digital form.
Hence the feasibility of digitally reconstructing camera holograms merits consideration. In addition, the 3 availability of image data in digital form permits the application of image enhancement techniques, developed for space and military purposes, to nuclear medicine. These considerations are the subject of this research.
1.3 The Goals of the Research
This dissertation describes the development and evaluation of a design for a practical coded aperture system with digital reconstruction. The system contains a high purity germanium, gamma ray imaging camera with an on-axis Fresnel zone plate aperture. The design of the coded aperture system is based on constraints dic tated by the characteristics of the camera detector and its associated electronics. These characteristics were established by research into gamma camera design currently being carried out by the Nuclear Engineering Department at The Ohio State University.
There are four primary goals of the research described in this dissertation. First, the physical system is to be designed. Second, the feasibility of computer recon struction is to be demonstrated. Third, the optimal digital image enhancement scheme is to be determined.
Fourth, the clinical capabilities and limitations of the zone plate, germanium camera system and associated data processing system are to be determined. 1.4 The Organization of the Dissertation
A chapter-by-chapter overview of the dissertation is presented here. Chapter 2 is a qualitative discussion of the advantages and disadvantages of coded aperture imaging with a germanium camera; this discussion motivates the consideration of digital reconstruction to overcome the difficulties. In this chapter also, closely related research by other investigators is summarized. Theory relevant to this work, but not developed by the author, is presented in Chapter 3. The terminology, concepts and equations of Chapter 3 will be used extensively in the following chapters. Chapter 4 presents the overall system design. The germanium camera is described, and a method ology of Fresnel zone plate system design is presented.
Chapter 5 deals with the digital reconstruction of shadow grams, beginning with an outline of the theory of digital reconstruction. The simulation of shadowgrams which would be formed by the design system is discussed, and the quality of the images obtained from these shadowgrams is evaluated.
Chapter 6 discusses the improvement in image quality obtained in the image enhancement studies. The clinical practicability of the overall system is evaluated in
Chpater 7. Chapter 8 contains a concise summary of the results and the conclusions drawn from these results. CHAPTER 2
RESEARCH MOTIVATION AND RELATED RESEARCH
BY OTHER INVESTIGATORS
2.1 The Motivation for the Research
In nuclear medicine, gamma ray cameras are used to form images of the distribution of radioisotope uptake in many bodily organs. Conventional cameras use a sodium iodide scintillation crystal as the detection medium; the scintillation camera has a spatial resolution of one to two centimeters. An aperture must be placed between
the organ and the detector to form an image of the organ on the detector face. The parameters of the aperture are determined by the spatial resolution of the detector.
Germanium detectors are expected to become available in
gamma cameras during the next few years (1.-4.) » and will
be capable of a spatial resolution of about two millimeters.
Present gamma camera apertures are unsuitable for a high
resolution camera. This research investigates the feasi
bility of using a particular aperture suitable for a high
resolution camera, collecting image data in coded form,
decoding the data and enhancing the image by digital
processing.
5 6
2.1.1 Imaging Methods
Traditional gamma camera apertures are variations
of two basic types, the pinhole aperture and the multi
hole collimator (see Figure 2.1). Consider a distribution
of radioisotopes to be made up of many small point
sources of radiation. With both types of apertures,
although a point source at A emits radiation in all
directions, only radiation emitted into a small cone
passing through a hole in the aperture is allowed to
reach the detector. The point source irradiates an
approximately circular area on the detector, and the
aperture is designed so that the diameter of this circle
equals the spatial resolution of the detector.
An improvement in detector resolution requires smaller
aperture holes, and hence more radiation. As the holes
become smaller, the cone into which gamma rays are emitted
bocomes smaller. The solid angle subtended by this cone
is proportional to the square of the detector resolution.
.For example, consider a case in which the spatial resolu
tion of a camera is improved from 2 cm to 2 mm. To
realize this factor of ten improvement requireis reducing
the aperture hole diameter by a factor of ton, but then
the number of gamma rays reaching the detector face is
reduced to 1% of the former value. This reduction to 1%
applies only to the pinhole aperture; for the multihole
collimator more holes may be added to the design, but 7
A
B
source
detector aperture
(a)
source aperture detector (b)
Figure 2.1 Conventional apertures. (a) the pinhole aperture; (b) the multihole collimating aperture. 8 limitations on the hole spacing imposed by geometric, material and gamma ray penetrability considerations still result in a reduction in the number of gamma rays reaching the detector.
In addition, for either type of aperture, a 2 cm by
2 cm square on the detector constitutes one resolution element before the improvement, but constitutes 100 resolution elements after the resolution is improved to
2 mm. Suppose that an average of n gamma rays must be detected in each resolution element to obtain a statis tically well-defined image. In the former case, n gamma events must be counted in the 2 cm square. In the latter case, n counts per 2 mm square are required; equivalently,
100 n counts are required in the 2 cm square in order to obtain an image having a statistical quality equal to the quality obtainable by the former detector. As the need for radiation increases by a factor of 100 per unit area, the amount of radiation actually received per unit area decreases (for a pinhole) by a factor of 100. It is desirable to obtain images of small organs with a germanium camera such that these images have the same statistical quality obtained previously with a scintillation camera for images of organs ten times larger. This requires that the organ emit 10,000 times as many gamma rays as before. 9
To meet this requirement for more radiation, one may' either increase the dose given to the patient by a factor of 10^ or increase the time for collecting radiation by
10^. Neither alternative is acceptable. One may not
increase patient dose from several millicuries to tens of
curies. The collection time is limited by the need for many patients to use the camera, by demands on the clini
cian's time, and by the need to minimize discomfort to the
patient. In addition, a high resolution camera actually
requires a shorter collection time than conventional
cameras because patient motion is expected to be a problem
in obtaining high resolution images. Furthermore, a high
resolution gamma camera is expected to significantly
expand the role of dynamic function studies in nuclear
medicine: if the image of a changing radioisotope
distribution can be formed every few seconds, one has in
effect a motion picture of the body's internal organ.
These dynamic function studies demand a minimal collection
time. Small collection time and reasonable patient dose
impose a limitation on the spatial resolution achievable
by a gamma camera using a traditional camera aperture.
2.1.2 Coded Apertures
The collection time can be reduced for dynamic
studies and the dose to the patient may be reduced for
static images by abandoning these traditional camera apertures which reproduce directly on the detector face the radioisotope distribution in the body. There are a variety of apertures which provide a large open area for the transmission of radiation but which do not result directly in the image (5). Instead, the aperture forms a pattern on the detector face which contains the image in coded form, and the pattern must be decoded to obtain the image. In this case, the aperture is termed a coded aperture.
The principle advantage of the coded aperture is that a larger fraction of the radiation emitted by the source reaches the detector, and this fraction is independent of detector spatial resolution. Replacing a pinhole or multi- hole aperture with a coded aperture is comparable to replacing an optical pinhole camera with a lens camera.
The lens in a camera has two functions: it provides a large opening for collecting light efficiently and it focuses the collected light into an image. The coded aperture has the advantage that it provides a large opening for collecting gamma rays efficiently. Its disadvantage is that obtaining the image is a two-step process. The
first step, shadowcasting, is the collection of radiation to form a coded pattern. The second step, reconstruction,
is the decoding of the pattern to retrieve the image. 11
A coded aperture system is especially suitable for a ' germanium gamma camera. The coded aperture provides the high gamma ray collection efficiency, and hence short collection time, needed for optimal utilization of the germanium camera. At the same time, this camera has the qualitites of high detection efficiency, good scatter discrimination and excellent spatial resolution needed for successful implementation of the coded aperture con cept.
A number of coded apertures having arrays of multiple pinholes has been proposed. A coded aperture consisting of an array of N pinholes collects N times as many gamma rays as cither a pinhole aperture or a multihole collimator.
A multihole collimator is made thick enough so that ideally gamma rays from a point source pass through only one hole in the collimating aperture. On the other hand, the multi hole coded aperture is made thin enough so that no colli- mation occurs, that is, gamma rays from a point source are able to pass through any hole in the coded aperture.
Four multihole coded apertures have been proposed: the random array, the nonredundant array, the linearly modu lated array, and the time modulated array.
The random array (6) divides the aperture into a mesh of small squares, half of which are opaque and half of which are transparent holes. The positions of these holes are randomly selected. Each point source in the object to be imaged casts a shadow of this aperture pattern onto the detector face. These overlapping aperture patterns create a coded image. Several optical reconstruction schemes have been suggested, each requiring a positive photographic transparency of the coded image. One such scheme involves using the original aperture pattern as a filter for light beams directed through the coded transparency. For each of the overlapping aperture patterns in the transparency a beam direction exists for which that pattern is aligned with the original aperture pattern.
The filtered light beams arc focussed onto separate points in the image plane with a lens. For beams not encountering aligned patterns, the light intensity is reduced by a factor of two in passing through the coded trans parency and again by a factor of two in passing through the aperture pattern, for an overall reduction by four.
However, if the transparency pattern and aperture pattern are properly aligned, the light intensity is reduced only by a factor of two. Hence there will be a brighter spot on the image plane corresponding to each point source in the original object.
The biggest disadvantage of the random array aperture is that the intensity of the image is never more than twice the intensity of the background. This low contrast may not be a problem in such applications as X-ray or gamma ray astronomy where separate point souces are to be imaged. But for nuclear medicine applications concerned with imaging continuous-tone objects, the low contrast attainable with the random array aperture has limited further consideration of its use.
The nonredundant array (7) alleviates the problem of low contrast by sacrificing considerable collection efficiency. Only about 24 of the aperture area allows radiation to pass through to the detector. Typical non redundant apertures contain 15 to 27 pinholes. These pinholes are spaced nonredundantly, that is, the vector displacement between any pair of holes is unique. Coded transparency formation and image reconstruction with this aperture is similar to the procedures outlined for the random array. As before, when the transparency and aper ture patterns are properly aligned during reconstruction, light will pass through all the holes in both arrays.
However, if the nonredundant arrays are not properly aligned, the light will pass through, on the average, only a single pair of holes. As a result, both the contrast between the image and background and the gamma ray collection efficiency increase with the number of holes in the array. 14
A possible limitation on the nonredundant array idea' is the excessive computer time required to generate non redundant array patterns when the number of holes is more than about 25. Nevertheless, the nonredundant array is a promising method of coded aperture imaging which has only recently begun to receive consideration.
The linearly modulated array (8) is a single line of pinholes for which the interhole spacing decreases from one end of the line to the other. More precisely, the spatial frequency of the array at any point is proportional to its horizontal position along the array. The motiva tion for this aperture is the simplicity of the reconstruc tion process which requires no precision optical system.
The coded image is viewed with a television camera which converts each horizontal line of the coded image to a temporal signal. When this video signal is passed through an appropriate dispersive delay line, each overlapping aperture pattern in the signal is compressed in the time domain into a single pulse which results in a bright spot on a television monitor. The resulting group of bright spots is the decoded image. Although the feasibility of this scheme has been demonstrated, the quality of the images obtained has been unsatisfactory, and no current research using this scheme is being carried out. IS
The time modulated array (9) is an attractive scheme' recently introduced which promises to provide resolution
> * which is uniform over the camera field of view. Most coded aperture schemes suffer from the fact that object points near the edge of the field of view cast shadows of only part of the aperture pattern onto the detector face. The reconstruction process, which is essentially a spatial integration over the recorded data, results in
images of these points having reduced resolution because of the restricted spatial limits of integration. With
the time modulated scheme, reconstruction is performed by integrating over the time domain in such a way that
the equal time limits of integration for all points produce
equal resolution for all points.
The aperture is a long lead strip containing a pseudo
random sequence of holes. During data collection the
strip moves along a line at right angles to the line from
source to detector, so that the portion of the strip which
is the effective aperture is a function of time. Data
is recorded digitally as number of counts per detector
resolution element per unit time. The decoding scheme
requires digital data processing. Good results have been
obtained with the time modulated array. The chief
disadvantage of using this array at present lies in the 4 16 excessive computer storage and processing time required to implement it.
2.1.3 Fresnel Zone Plate Aperture
The coded aperture which has received the most attention is the Fresnel zone plate. It is attractive because, at least in theory, reconstruction is accomplished most easily with this aperture. The Fresnel zone plate consists of concentric annular regions of equal area, alternately opaque and transparent to radiation. This pattern has the property that it acts as a diffraction lens to focus monochromatic light.
The use of a Fresnel zone plate aperture with a gamma camera is illustrated in Figure 2.2. Gamma rays passing through the aperture undergo no diffraction; each point source in the organ simply casts its own shadow of the pattern onto the detector face. The distribution and intensity of the overlapping patterns represent coded information about the distribution and concentration of the source material taken up by the organ to be imaged.
The coded pattern even contains three-dimensional infor mation about the source distribution because sources close to the aperture cast larger shadow patterns than those farther away. In addition, the whole detector face contains information about each point source, pro vided the aperture material is thin enough so no source aperture detector
Figure 2.2 The Fresnel zone plate aperture. 18 collimation occurs. If half the detector pattern were somehow lost, the remaining half would still contain information about the whole source distribution. The reader familiar with holography will recognize the simi larity between the pattern encoded on the detector and a hologram. In fact, this pattern is a hologram, formed by shadowcasting rather than by interference. The Fresnel zone plate pattern is similar to the hologram of a point source interfering with a plane wave, and the pattern on the detector face is similar to a hologram of the source distribution.
2.1.4 Optical Reconstruction
As in the case of an interference hologram, the
image encoded in the shadowcast hologram can be recon
structed using a source of monochromatic light. Figure 2.3
shows a photographic transparency of the detector pattern
illuminated with monochromatic laser light. Each zone plate pattern on the transparency acts as a lens focussing
the light to a separate point on the screen. The focal
length of each "lens" depends on the size of the zone
plate pattern. Because sources at various distances cast
shadows of various sizes, the images of these sources will
be formed at varying distances from the coded pattern
transparency. Hence by moving the screen in Figure 2.3, laser lenses coded pattern transparency
screen
Figure 2.5 The optical reconstruction system. i-* to 20 one can bring into focus various tomographic planes in the radioisotope tagged organ.
Optical reconstruction of shadowcast holograms, called shadowgrams, involves a difficulty not encountered with interference holograms. With interference holograms there is an angular separation of the dim, diffracted image beam from the intense, undiffracted beam passing straight through the hologram transparency. With shadowgrams there is no angular separation; the diffracted image is overlapped and obscured by the undiffracted beam. The obscuring effect can be virtually eliminated when forming an image of a small object, but an image of a large object can be formed only with unacceptably low contrast (10J .
Angular separation can be achieved by using only a portion of a Fresnel zone plate for shadowcasting. How ever, this portion, called an off-axis zone plate, results in an image having less detail than the image formed by a complete, on-axis zone plate. Use of an off- axis plate requires sacrificing by a factor of three (11) the high spatial resolution capability of the germanium camera - the capability which motivated the use of a coded aperture. In addition, the off-axis plate cannot adequately encode information about large objects unless used in series with a second aperture acting as a half tone screen (12). This dual aperture system reduces the 21 radiation collected by a factor of two, but more seriously it sometimes results in the appearance in the image of small nonexistent artifacts (13).
There are other serious difficulties with the optical reconstruction of shadowgrams. High contrast in the image requires a large dynamic range in the film transparency, that is, a large ratio of radiant energy required for maximum film exposure to that for minimum exposure. The limited dynamic range of film limits the contrast attain able in holographic images (L4) . Another difficulty is the painstaking film processing procedures required for the production of high quality holograms (1J5); these pro cedures would be difficult to carry out routinely in a clinical situation. In addition to processing for the purpose of maintaining the high dynamic range of the film, shadowgrams should be bleached and masked (16). Bleaching enhances the brightness and contrast of the image obtained.
Since more gamma rays are collected from sources near the center of the camera field of view than from those near the edge, the central sources appear brighter in the image; masking techniques during film processing can eliminate this bias. These masking techniques in particular require much more time and attention to detail than could reason ably be expected from a technician in a busy nuclear medicine clinic. 22
The difficulties with the optical reconstruction of shadowgrams led to consideration of the feasibility of digital reconstruction.
2.1.5 Digital Reconstruction
Difficulties encountered with optical reconstruction would not be encountered with digital computer reconstruc tion. The coded shadowgrams can be formed with an on-axis zone plate. The diffraction pattern resulting from illuminating the shadowgram with a beam of coherent radia
tion can be calculated digitally, and this diffraction pattern is the desired image. In digital reconstruction the image is not obscured by an undiffracted beam. Image uniformity can be obtained by a digital masking technique.
Computer processing can simulate any optical processing possible in a research laboratory, eliminating the need
for a clinician to operate a precision laser and optical
system and to carry out the special film processing pro
cedures required for holographic work.
In addition to avoiding the difficulties of optical
reconstruction, digital reconstruction of germanium
camera data in the nuclear medicine clinic has a number of
advantages. Data from a germanium camera is digital and
hence no loss of information is involved in digitizing the
data; in fact, digital processing can extract the maximum 23
information from the data. With computer reconstruction, no information is lost by using a CRT display to transfer
the data onto photographic film having film grain noise
and limited dynamic range, and no optical system aberra
tions are introduced during reconstruction. The "dynamic
range" of computer data memory can be chosen as large as
needed. Also, digital image enhancement can be readily
included in the data processing procedure to obtain an
image of maximum diagnostic value. Finally, quantitative
results are readily available. An integrated system of
germanium camera, minicomputer and visual display would
provide in a few seconds a high resolution image of an
internal organ for immediate diagnosis.
2.2 Related Research by Other Investigators
No research has been reported by other investigators
considering the use of the Fresnel zone plate aperture with
a high-purity germanium camera. All published research
to date has looked at the suitability of this aperture
for use with other detection media and, with one exception,
only optical reconstruction was employed.
2.2.1 Shadowgrams With Other Detectors
The first application of the Fresnel zone plate
aperture to radioisotope imaging was reported by Barrett 24 in 1972 (17). An on-axis zone plate created the shadow gram pattern by transmitting the gamma radiation received from a thyroid phantom. The shadowgram was detected by an Anger scintillation camera, and photographed from a
CRT display with a polaroid camera using transparency film.
This transparency was then illuminated by a beam of coherent light from a laser to reconstruct the image.
Although the fcasilbility of the concept was demonstrated, the poor resolution of the scintillation camera (18) severely limited the attainable image quality.
A group headed by W. L. Rogers (1£) then carried out a series of experiments testing the zone plate concept by simulating radioactive sources with optical sources and detectors with photographic film. They showed that better optical reconstructions could be obtained using an off-axis
zone plate. In addition, a significant increase in the
signal-to-noise ratio of the image occurred when a half
tone screen was used together with the off-axis zone plate.
These optical experiments indicated that a zone plate
aperture could be used successfully with a gamma camera
provided (1) the camera detects gamma rays efficiently,
and (2) the camera has good spatial resolution. Rogers'
group concluded that the use of a zone plate aperture
requires a detector resolution which is a factor of three
better than the resolution of the scintillation camera. 25
The use of a zone plate with the Michigan high reso lution image intensifier camera was investigated next by this group. This camera has a theoretical line spread function of 1.5 mm full width at half maximum for 122 keV gamma rays, but has an intrinsic efficiency of only
12% for 140 keV gamma rays. In zone plate experiments a standard thyroid phantom had to be loaded with the rela tively high activity of 5 mCi of ^ mTc to reduce the ampli fier noise to acceptable levels. In addition, the actual resolution of this camera was significantly poorer than that predicted theoretically because distortion in the intensifier tube had a defocussing effect on the shadow gram. A conclusion drawn from these experiments was that the system appears useful for small sources but not for larger sources containing voids. Also, the successful use of a zone plate aperture requires a camera with high intrinsic efficiency and a good actual spatial resolution.
Rogers concluded that "the zone plate imaging system opens a new dimension for imaging techniques in nuclear medi cine. This field still is in its infancy and considerable research is still needed, both theoretically and experi mentally" [20) .
Another detector capable of high resolution which has been used with a zone plate aperture is the multiwire proportional chamber. An on-axis zone plate was used with this detector by Macdonald and his colleagues (2JL) to image arrays of point sources. The purpose of the work was to verify a theoretical formula for the signal-to-noise ratio attainable as a function of the number of point sources in the array. Although there were no imaging studies of extended sources, the multiwire proportional chamber appears to be less suitable than a germanium detector for coded aperture applications in nuclear medicine because of its low sensitivity and poor energy resolution.
Only 301 efficiency is achieved for low energy 22 keV gamma rays, and the energy resolution is only about 15%.
The most successful use of a Fresnel zone plate aper ture to date has been with X-ray film as the detection medium. Barrett (22J has developed an experimental camera employing a half-tone screen, an off-axis zone plate, and an X-ray film cassette incorportating two fast intensivying screens. The X-ray transparency, after being photoreduced, is a shadowgram which produces a reconstructed image when inserted into an optical reconstruction system.
Experiments were carried out laying radio-opaque strip patterns of various spatial frequencies over an extended source. High resolution images were obtained for strip widths and spacings ranging from more than 1 cm down to about 5 mm. For a higher frequency spacing, the image contrast went abruptly to zero. Hence the system 27 appeared to have a step-like spatial frequency response.
Further experiments with a high resolution gold 2one plate and a thyroid phantom resulted in a system resolu tion between 2 and 3 mm.
The ability of the system to separate various tomo graphic planes was also tested. When one plane is in focus, other planes are out of focus to the extent that they are distant from the in-focus plane. Experimentally,
Barrett's group was able to clearly distinguish planes
2-1/2 cm apart. An object clearly in focus would begin to be distinctly out of focus if the system were focussed on a plane 5 mm behind that object. A difficulty with the three-dimensional capability of the system is that the out-of-focus image may not be simply blurred but can be made up of scrambled, but sharply defined components of the true image. This effect, along with the introduction into the image of nonexistent artifacts, is possibly due to the high spatial frequency content created by the half tone screen. In a clinical situation, this could lead to a false interpretation of the existence or position of a source. Clinical evaluation of this system is now being carried out (23).
The major disadvantage of this system is the method of radiation detection. Barrett reports that "although it has the virtues of high resolution and extreme 28 simplicity, X-ray film is obviously not the ultimate detector. It does not have the sensitivity for dynamic
function studies and does not have an energy discrimination capability which is important for rejecting scattered
radiation" (24).
2.2.2 Digital Reconstruction of Holograms and Shadowgrams
Although some research has been carried out in the
area of computer generated holograms for optical reconstruc
tion, very little work has been done on the digital recon
struction of holograms generated either by interference or
by shadowcasting. The number of resolution elements in
most holograms is so large that the computer time required
for digital reconstruction is excessive, and therefore
optical reconstruction is more practical and inexpensive.
In addition, the various tomographic planes of the optically
reconstructed image can be brought into view by simply
moving the screen on which the image is formed. On the
other hand, a separate calculation beginning with the basic
hologram data is required with digital reconstruction to
determine the image of each tomographic plane.
Goodman and Lawrence (25J demonstrated that a digital
image formed from an electronically scanned hologram is
comparable in quality with the optical image of the same
hologram. Aoki has compared optical and digital 29 reconstructions of both microwave (26) and sound-wave holograms (27) and shown them to be of similar quality.
The computer reconstruction of Fresnel zone plate shadow grams has received published consideration only by
Macdonald's group (28) in their work with the multiwire proportional chamber: in determining the signal-to-noise ratio attainable with a zone plate system, digital recon struction was chosen to eliminate the various noise sources introduced by optical processing. CHAPTER 3
RELEVANT THEORY
This chapter presents theoretical, material relevant to this dissertation but not developed by the author.
The basic optical theory which provides an understanding of the imaging ability of shadowgrams is covered first.
Then the theory of gamma ray imaging by shadowgrams is presented. Selected topics from the theory of digital signal processing are next discussed; these topics contribute to an understanding of the computer codes developed in the course of this investigation. Finally, relevant aspects of image enhancement theory are outlined.
3.1 Optical Theory of Shadowgrams
An overview of basic Fresnel zone plate theory and
its relationship to holography is presented here. Only
those features of the theory important in later chapters are mentioned. A more complete discussion, with refer ences, occurs in Appendix A. This appendix not only explains the theory but also gives a complete historical survey of the literature which contributed to the develop ment of the theory.
30 The Fresnel zone plate consists of concentric annular zones of equal area, alternately opaque and trans parent to an advancing wavefront, as shown in Figure A.3.
Denoting the first zone radius by R ^ , the radius of the n-th zone is given by
Rn * R1 /5P ' (3"X)
The width of the zones decreases with radius and is given by
Arn “ Rj ( /TT - / n-1 ) . (3-2)
A radiant wavefront encountering a zone plate behaves
in many ways as if it had encountered a lens, but the
effects are caused by diffraction rather than refraction.
A single zone plate can both converge and diverge incident
radiation, and multiple real and virtual focal points
exist for a single zone plate. The zone plate "lens" has
extreme chromatic aberration. The principle focal lengths,
real and virtual, of a zone plate for radiation of wave
length X are given by 32
Image formation by holography is also a diffraction process. The hologram of a point source is the pattern formed by the interference of a plane wave with a point scatterer of radiation. This pattern consists of concen tric annular light and dark zones having radii also deter mined by Equation 3-1. However, instead of zones of well-defined width, the pattern has a sinusoidal intensity variation along any radius. This pattern has the ability both to converge and diverge radiation in a lenslike manner. But because the zones do not have sharply defined edges, the pattern does not focus incident radi ation to focal points other than the two principal foci.
These focal lengths are also given by Equation 3-3.
Any extended object may be considered a distribution of point scatterers of radiation, and a hologram of such an object can be considered a superposition of overlapping
and interfering zone plate patterns. Radiation incident
on a hologram is diffracted by overlapping zone plate
patterns to form both a real image and a virtual image of
the original object.
Shadowgrams, on the other hand, are formed by a super
position of overlapping, but not interfering, zone plate
patterns. A shadowgram of a small object, consisting of
few equivalent point sources of radiation, is a pattern 33 very similar to a hologram of that object. However, for a large object, the shadowgram consists of many over lapping zone plate patterns, and the noninterference of these patterns can result in a pattern quite different from a hologram of the object.
3.2 Gamma Ray Imaging by Shadowgrams
Important features of the theory of gamma ray imaging by shadowgrams are presented here. A more complete discussion, with references, appears in Appendix B.
Most of the theory of gamma ray imaging of shadow grams has been developed by H. H. Barrett and his colleagues,
His approach to imaging theory is to use the mathematical
framework of communications theory. By finding the trans
fer function of a zone plate for the spatial frequencies contained in the source distribution, Barrett shows that the source spatial frequency imaged best by the system
is given by
so Rw f - -?■ -Jl . (3-4) S1 Rj
Barrett also relates the signal-to-noise ratio for a
zone plate of diameter DN to that for a pinhole aperture 34 of diameter Dp. For imaging a point source the relation- • ship is
CSNR>zp - WIT CSNRJp . (3-5)
Because D^ is much greater than Dp, for equal exposure times clearly the zone plate will have a much better signal-to-noise ratio. However, this is true only for imaging a point source. For imaging a small object having an area of M resolution elements, the signal-to-noise
- 1/2 ratio is reduced by a factor of M ' from its value for a single point. This is because gamma rays from all resolution elements contribute to the noise but only those
from one resolution element contribute to the signal for that resolution element.
3.3 Digital Signal Processing Theory
A computer code was developed in the course of this
study for digital reconstruction of shadowgrams. Optical
reconstruction involves converting a continuous two-
dimensional shadowgram intensity distribution into a
continuous two-dimensional image intensity distribution.
It is shown in Appendix D that the image is a two-dimensional 35
Fourier transform of the shadowgram and hence that the shadowgram may be considered a map of the spatial fre quency content of the image. Digital reconstruction requires that the shadowgram be reduced to a finite number of input data values and that the image be reduced to a finite number of output data values. The theory of digital signal processing includes a theory for carrying out this reduction without loss of information. An outline of the theory is presented here in the one-dimensional time and frequency domains for clarity. The results are easily extended to the two-dimensional spatial and spatial frequency domains.
Consider a continuous function of time, h(t) , that is sampled for purposes of digital processing. If the period between samples is T, then after sampling one retains only the information contained in the time function, hR (t), given by
00 6 (t - nT) ne-oo (3-6)
00 = h(nT) 6(t - nt) , 36 where n refers to the sample number and 6 is the delta function. Let the sampling function be given by
CO d(t) » £ 6(t - nT) . (3-7)
ne .00
Then
hn (t) » h(t)d(t) . (3-8)
Let a capital letter denote a Fourier transform to the frequency domain, f. The well-known convolution theorem then leads to the result
Hn (f) = H(f)*D(f) , (3-9)
where an asterisk denotes convolution. D(f) is given by
(29) to be
00
D(f) - ^ Y j {(f ■ f) (3‘10) n«-»
Hence Hn (f) is a function which repeats the function H(f) at intervals of period ^ in the frequency domain. H(f) contains complete information about the original function h(t). Therefore Hn (f) also contains this information 37 provided the repetitions of H(f) are spaced far enough apart so they do not overlap. Overlapping introduces into Hn (f) false information about the frequency content of H(f). This is called aliasing.
Figure 3-1 illustrates these concepts. Figure 3-1(a) shows h(t) and its Fourier transform which is zero for any frequency value of magnitude greater than f , the cutoff frequency. Figure 3-1 (b) shows d(t) and its
Fourier transform. In Figure 3-l(c), Hn (f) is shown to consist of periodic repetitions of H(f) provided the bandwidth of H(f) is less than the period of Hn (f) , that is, provided fc < ^r.
Now H(f) can be recovered by multiplying Hn (f) by a unit rectangular function, Q(f), given by
1 -L < f < 1 2T — — 2T
Q(f) - (3-11)
as shown in Figure 3-1(d). Then
H(f) » Hn (f)Q(f) (3-12) as shown in Figure 3-1(e). Application of the convolution theorem then yields h(t) H(f) 38
t
(a)
1 * i >
-3T -2T 21 3T I7T 2/T y t )
-3T -2T -T 31 Q(fJ
-l------u IT h (t)*q(t)
Figure 3.1 The recovery of a sampled waveform. Figures (a) through (e) illustrate relevant functions and their Fourier transforms. 39
jh(nT) Y, 6(t ‘ »T) IqCt) L n=-°°
00
n»-t»
00 sinn ( £ - n ) E Mm) - u - >-T , (3 -1 3 ) B *00 * ( r - ")
where q(t) is given by (30J. Hence the original continuous function can be fully recovered from samples spaced at interval T, provided ^ 2fc.
For a two-dimensional spatial function h(x,y) of extent T and T , the corresponding result is x y
« 00 simr(~- - n)sinit^^- - m) h(x,y) m Y Y h(nT ,mT ) ---- j— 5----- r - j — *— r— , y " (rr - ») ( f • m ) a y
(3-14) provided 2fc and ^~ il 2fc » where fc is the maximum x y spatial frequency content of h(x,y). This result is 40 applied in the theory of digital reconstruction presented in Appendix D to finding the best spatial interval between sample points in the image plane.
3.4 Image Enchancement Theory
An imaging system maps a set of input functions into a set of output functions. These functions can be real valued, dealing with intensities, or complex valued, deal
ing with field amplitudes. These are usually functions
of the two-dimensional independent variable, space.
An imaging system can be described by a mathematical
operator, M, which operates on input functions to produce
output functions. Let the system output be denoted by
v(Xi,yi) and the input by u(x2 ,y2). Then the relationship
between them is
vCxj.yi) = M {u(x2 ,y2)} . (3-15)
4
Almost all image enhancement studies have dealt with
systems which are linear. A system is linear if the
superposition principle holds for all input functions,
that is,
M {aUj + bu2> = aM{uj) + bM{u2) , (3-16) 41 where a and b are complex constants. The output of a
linear system can be expressed as a superposition of
outputs from weighted impulse responses. An impulse
response is the response of a system to a point source
(delta function) input. For a system described by opera
tor M, the impulse response is
(3-17)
It is easily shown (31^) that input and output are then
related by
00 uU.i-OhCx. ,y ;5,n)dedn
Most imaging systems are linear if both light amplitude
and phase are considered.
Some linear systems are also spatially invariant.
For such systems the functional form of the output does
not change as an input point source moves about in the
input field; only the position of the output changes. Then
the impulse response depends only on the displacements
(Xj - £) and (y^ - n). Hence the output may be expressed
as a convolution integral, as follows, 42
v(xl»yi) m J J u(C,n)h(x1 - Z ty1 - n)d^dn . (3-19) -00
Denoting the convolution by an asterisk, we obtain
v ** u*h . (3-20)
In terms of Fourier transforms, this relationship becomes a simple multiplication; that is,
V ■» UH •» HU (3-21)
The function, H, is called the transfer function of the system. The transfer function can be extremely useful for image enhancement. But its use is limited to linear, invariant systems.
We consider here only that use of the transfer func tion which proved of some use in this study. Suppose an imaging system produces a known output, v, from an unknown input, u, which must be determined. This can be accomplished by first determining the transfer function of the system.
A known point source, up , is imaged by the system and the corresponding output, v , is determined. The Fourier r transform of Vp, denoted by Vp, can then bo determined either optically (32) or digitally (33) . The Fourier 43 transform of un is a constant because u is a delta P P function. Then to within a multiplicative constant,
Equation 3*21 reduces to
Vp - H , (3-22)
and the transfer function of the system is determined.
The output, v, from the unknown input can be Fourier trans formed, giveing V. Then we obtain the Fourier transform of input, U, given by
U - f . (3-23) P
Finally, the unknown input, u, is found by taking the inverse Fourier transform of U.
Studies of space variant systems have recently begun to appear in the literature (3£). The approach has been to try to add another parameter to the transfer function to account for the variance in cases where the variance is known to have a symmetry, such as radial symmetry.
Although some theories have been advanced, it appears that no practical results have yet been obtained. Nonlinear systems remain intractable. CHAPTER 4
OUTLINE OF SYSTEM DESIGN
Most important in this chapter is the methodology for determining the physical parameters of a Fresnel zone plate imaging system. Because the choice of these parameters depends on the detector characteristics, the detection system of the high purity germanium gamma camera is discussed at the outset. A zone plate system suitable for this camera is then designed. Alternative data processing systems are considered next, and finally the design of the overall imaging system is outlined.
4.1 The Germanium Detector
The imaging system described in this chapter is a
Fresnel zone plate system designed for use with a high purity germanium gamma camera. The features of the camera design most relevant to the zone plate design are presented here. A detailed description of the camera is presented in Appendix C.
The detection system of the camera consists of an array of orthogonal strip detectors, each of which is
44 45 cut from a single crystal of high purity germanium (see
Figure C.l). The spatial resolution of the camera depends on the spacing of the strips and the electronics.
A strip spacing of 1 mm could be achieved at present; but the electronics, as currently designed, would not allow the resolution of individual strips. Improvements
in electronic resolution are possible if the energy resolution of the system is allowed to degrade or if the cost of the electronics is allowed to increase signifi cantly. The design considered here has an electronics resolution of 2.4 mm.
The camera is designed to have 16 individual detectors
in a four-by-four square array. The side of the square will be 15.3 cm in length (see Figure C.2).
4.2 The Fresnel Zone Plate System
Several important parameters must be determined in
setting up a zone plate imaging system. This section
specifies these parameters, determines relations among
them, outlines a method for choosing these parameters
and applies this method to choosing the parameters for
the germanium camera described in the previous section.
4,2.1 Zone Plate Parameters
The choice of all zone plate system parameters
depends on the characteristics of the detector in the 46 system. It is assumed here that the detector is a square ' of side made up of resolution elements of size Ax^ as shown in Figure 4.1. The number of resolution elements per side is N^.
After the shadowgram is received on the detector and digitally processed, an image is obtained which shows a
square area of the object plane. The side of this square
is defined here to be the field of view, T, of the system,
and the dimension of the resolution element in the object plane is At, as illustrated in Figure 4.2.
A Fresnel zone plate is placed between the object plane and the detector as shown in Figure 4.3. It is
assumed for the present that the detector and zone plate
thicknesses and the object depth of field are small
compared to s^, the distance from the object plane to the
zone plate, and S2 » the zone plate to detector distance.
A very useful parameter is the ratio of these distances,
given by
The zone plate is circular having N zones. Denoting the
radius of the n-th zone by Rn , then Rjj is the radius of
the entire zone plate, that is, of the outermost trans
parent zone. The diameter of the zone plate is denoted Figure 4.1 A square detector of size Dj divided into resolu tion elements of size Axi. There are Ni resolution elements per side. Here Nj * 5. 48
| At ---1
Figure 4.2 Zone plate system field of view. The area is a square of side T, containing Nj by Nj square resolution elements of size, At. Here » 5. 49
>
object zone plate detector plane plane plane
Figure 4. The object-to-zone plate distance, si, and the zone plate-to-detector distance, s7.
4 so by D^. The width of the n-th zone is Arn and hence the width of the outermost zone is Ar^. The thickness of the zone plate is t2»
The parameters of importance are those summarized in Table 4-1.
Table 4-1. Parameters of the Imaging System
Symbol Definition number of resolution elements per side Ni length of square detector side Di Ax^ length of square resolution element side
T field of view
At object plane resolution
N number of zones in zone plate
diameter of zone plate d n width of outermost zone firN object-to-zone plate distance S1 zone plate-to-detector distance s2 a ratio of S2 to s^
thickness of zone plate
4.2.2 Relations Among Parameters
Consider first the relationship of the detector parameters. Because contains a number of resolution 51 elements of size Ax^, the number of resolution elements per side is
N 1 ' 25^ <4'2>
Consider next the relationship of the object plane parameters, T and At. The detector data is an by matrix of gamma counts. According to sampling theory
(Chapter 3, section 3.3.1), the maximum image information is contained in an by matrix of image data. Then the size of the object plane resolution element is
At » jj— . (4-3)
The zone plate parameters have the following relation
ships. Combining Equations 3-1 and 3-2, and using an approximation which is valid for N greater than 4, we obtain
Dn « 4NArN . (4-4)
The thickness of the zone plate is dictated by the energy
of the incident gamma radiation. A lead zone plate used with 140 keV gamma rays should have a thickness of at
least 1 mm. 52
The object plane parameters are related to the detector parameters as follows. In order to resolve two distinct point sources, the outermost zones of their zone plate shadows must not overlap, but must be separated by at least one detector resolution element. Hence the centers of the two zone plate shadows on the detector must also be separated by Axj. The similar triangles of
Figure 4.9 illustrate that to obtain a shadow center separation of Ax1# the two object points must be separated by At, wheTe
Ax 1 At *» (4-5) a and a is defined in Equation 4-1. Then by combining equations 4-2, 4-3 and 4-5 we obtain an expression for
the field of view in terms of the detector size, that is,
(4-6)
Having related object plane parameters to detector parameters, we consider next the relation of zone plate
parameters to detector parameters. To obtain maximum
system resolution, the width of the outermost zone of
the zone plate, Ar^, should be such that a point source in
the object plane fully irradiates exactly one resolution 1 1 At Ax, T T
Figure 4. 4 The relationship between object plane resolution At and detector plane resolution Ax^.
c/i u 54 element in the detector plane, as shown in Figure 4.5.
If Ar^ is smaller, the outermost zones of the shadow will not be resolved by the detector; these unresolved zones will not contribute to image improvement, but rather will constitute noise which would degrade the image. If is chosen too large, the resolving ability of the system will not be fully realized. From the similar triangles of Figure 4.5 and Equation 4-1, we obtain
Ax, ArN ■ . (4-7)
In order for the detector to receive the complete shadow of the outermost zone cast by a source at the
center of the field of view,9 • the diameter of this shadow Dg must be less than or equal to D^. But the diameter of the shadow is equal to B^Cl + a), and hence
d n ± r r r ( 4 - 8 )
An alternate expression for DN can be found by substituting
Equation 4-7 into Equation 4-4, giving
N D1 dn * 4 RJ inr-sr • C4-®) 55
1 Axj Ar, T
object zone plate detector plane plane plane
Figure 4.5 Radiation from a point source in the object plane passing through the outer most zone. 56
Comparison of Equations 4-8 and 4-9 yields the result that
N 1 N<-^ . (4-10)
For detectors having a small value for it is desirable to choose N to be the largest integer less than or equal to Nj/4.
The relationships developed in this section arc summarized by the diagram in Figure 4.6. The diagram assumes that D^, Ax^ and a are known first; the arrows and associated equations show how the other parameters arc determined.
4.2.3 Method of Choosing Parameters
Several practical considerations affect the order of choice of the parameters discussed above. In Figure
4.6, it is assumed that a is known before the parameters
Ar^ and At are determined. In practice, constraints on these parameters can limit the choice of a.
The choice of a affects the magnification of the system. If a » 1 (that is, s^ * s2^ » s*ze °f
field of view equals the size of the detector, as in the case of the multihole collimator. However, as the zone plate is moved closer to the detector and a decreases, the < Di D1 N — 1 * a T = 4 D, T
Dn = 4NArN N < h T = NxAt ■ 4 D 1 ~ N j AX j
V ArN < - Ax. ■> At Ax. Ax. At = ArN = 1 + a
zone plate detector object plane parameters parameters parameters
Figure 4.6 Relationships among the zone plate, detector and object plane parameters. C/1 ^■4 system is able to image a field of view larger than the detector. The penalty for this ability is a loss in system resolution. In the limit that a equals zero and the zone plate is pressed against the detector face, the image of the object is reduced to a single point. Conversely as the zone plate is moved toward the object and a increases, the ability of the system to resolve detail in the object improves. But at the same time the field of view decreases. It is important to keep in mind that as the zone plate is moved between the object and detector its dimensions must be changed in accordance with Equations
4-7 and 4-8, if optimum imaging ability is to be realized at each new zone plate position. The inability to construct a zone plate of extremely small dimensions is a limitation on the resolving power of the system.
Another limitation, however, is imposed by the thick ness of the zone plate. Gamma rays passing through a zone at an angle with the axis of the system are partially attenuated because of the finite thickness of the plate.
A minimum value for Sj must be maintained to keep this attenuation from becoming severe. A consideration of this problem for the case a > 1 is presented here. When a is greater than one the attenuation is most severe for gamma rays from the center of the field of view passing through the outermost zone. Figure 4.7 illustrates (a)
(b)
Figure 4.7 Attenuation of gamma rays passing through the outermost zone, (a) gamma rays from a point source at the center of the field of view are emitted toward the outermost zone; (b) the angle 0 is such that half the gamma rays enter ing the zone emerge from the other side. 60 the problem and assumes for simplicity that the plate material is completely opaque to radiation. In Figure
4.7a radiation from a point source on the axis of the system travels toward the outermost zone in a direction that makes an angle with the axis having a tangent equal to D^/Zs^. Figure 4.7b shows this radiation passing through the outermost zone. If we impose the constraint that at least half of the radiation incident at the entrance to the zone must leave the zone without passing through opaque material, then the tangent of the maximum allowable angle of incidence is Ar^/2tz. Therefore, we require that
Dm Ar„
! C 4 ' U ) and hence a minimum value of s^ is established. Substi tuting Equation 4-4 leads to the result that
s. • - 4Nt_ . (4-12) l,min z '
When a is less than one, sources at the edge of the field of view are attenuated most severely; by an argument similar to the one above, it is readily shown that
4Nt_ s a _____ 2 l.min a (4-13) Since the zone plate system is capable of magnifica
tion, it is worth inquiring about the best resolution obtainable in the center of the object field of view by a camera with given parameters. According to Equation
4-5, the object field resolution improves with increasing a. Then if s^ is fixed at Sj min, by increasing s2
resolution can be improved. But increasing s2 reduces
the gamma ray collection efficiency and increases the
importance of small angle scattering in the zone plate.
More important is that an increase in s2 requires a
decrease in zone plate dimensions in accordance with Equa
tion 4-8. The mechanical inability to construct a zone
plate of small precise dimensions while maintaining a
minimum thickness imposes the basic limitation on system
resolution. By substituting Equation 4-7 into Equation
4-S and rearranging, we obtain
At-(l*i)ArN , (4-14)
which shows that as a increases to infinity the best system
resolution obtainable is equal to the width of the outer
most zone. However, for a system in which the camera
resolution AXj is greater than the minimum attainable
value of Ar^, substituting Equation 4-7 into Equation 4-5
and eliminating o results in a more severe restriction on
system resolution, namely, 62
If, for example, Ax^ is 2 mm and the minimum value of
ArN is 1 mm, then the best object resolution is 2 mm.
In choosing the parameters of a zone plate system, one must decide whether to design for the best object resolution or to enlarge the field of view in order to form a complete image of a large object. In most cases of interest in nuclear medicine in which a germanium gamma camera would be used, the decision would be to optimize system resolution. Table 4-2 presents the system parame ters, their order of choice and equations for determining these parameters for the case where system resolution is to be optimized. The table assumes that D^, Ax^, Ar^ and t are given.
The table also assumes that s^ should be set equal to its minimum value in order to achieve maximum gamma ray collection efficiency, emflX. Expressed in terms of
the given parameters of Table 4-2, the efficiency is
approximately
em a x ” 7 ( * t ^ ) (4-16)
If the source activity in the field of view is q gamma
emissions per second, the camera must have a count rate
capability C.R., at least as great as 63
For Optimum System Resolution
Given: D ^ t AXj * ArN» tz
Parameter Equation Equation No.
N1 = D i 4-2 N i AXj
Ax, a a « 4-7 1
Ax, i At At » a 4-5
T T = N j A t 4-3
N i N N 4 x 4-10
Dn Dn “ 4NArN 4-4
Sj Sj ■ 4Nt2 4-12
s2 s2 - 0.5J 4-1 64
C.R. = qemax (4-17)
If the count rate capability of the camera is'exceeded, then either q can be reduced or can be chosen greater than the minimum allowable value. Although the second alternative reduces the efficiency, the attenuation of gamma rays passing at an angle through the zone plate is also reduced and system imaging capability is enhanced.
If the system parameters are chosen so as to obtain a specified field of view, the order of choosing the parameters is that given in Table 4-3. When T is specified,
At will not necessarily be optimum and Ar^ may be greater than the minimum width one is capable of constructing.
4.2.4 Parameters for the Germanium Camera
Let us next consider the choice of system parameters for the germanium camera described in Section 4.1. In this case is 15.3 cm, the FWHM electronic spatial resolution is 2.4 mm, and so N1 is 64 resolution elements per side. The radioisotope most often used in nuclear medicine is technctium-99m with a gamma ray energy of
140 keV; if the system is optimized for use with this
isotope and if the zone plate thickness is chosen to attenuate 95% of the incident radiation, then t£ is 1 mm.
It is extremely difficult to machine zones in lead which 65
Table 4-3. System Parameters and Equations
for a Specified Field of View
Given: Dj, AxJt t2, T
Parameter Equation Equation No.
D1 N1 N1 ■ -SJ- 4 ' 2
At t “ 4-3
X1 4-5
ArN 4rH ■ 4-7
N1 N N * 4-10
Dn Dn » 4NArN 4-4
Sj Sj ■ 4Ntz 4-12
s2 s2 “ aSj 4-1 66 are spaced closer than 1 mm; and in addition, if the width of the opaque zones is less than the zone plate thickness, gamma rays making an angle with the system axis will penetrate these supposedly opaque zones. Hence a minimum practical value for Ar^ is also 1 mm.
Table 4-4 gives the parameters of a zone plate-germanium camera system for three cases. In case 1 the parameters are chosen to optimize system resolution according to the method of Table 4-2. The minimum practical value of
Ar^ is chosen, the magnification parameter a is maximized, and a spatial resolution of 1.7 mm for the object plane is predicted. Notice that the system is able to resolve objects smaller than the spatial resolution of the camera alone. However, the field of view is only 11 cm, while the camera detector is a square of side 15 cm.
In cases 2 and 3 the parameters are chosen so as to achieve a specified field of view, according to the method of Table 4-2. In case 2, the field of view is chosen equal to the detector dimensions; a is unity, and the area imaged is the same as would result from using a multihole collimator. In this case the system resolution is equal to the detector spatial resolution. In case 3 the field of view is chosen larger than the detector dimensions.
More specifically, the area of the square field of view is chosen equal to the circular usable area of the standard 67
Table 4-4. Zone Plate-Germanium Camera
System Parameters
Parameter Case 1 Case 2 Case 3
Dj (mm) 153 153 153
AXj (mm) 2.4 2.4 2.4
Ni 64 64 64 tz (mm) 1 1 1
At (mm) 1.7 2.4 3.7
T (mm) 110 153 236
ArN (mm) 1 1.2 1.5
N 16 16 16
Dn (mm) 64 77 93 a 1.4 1 0.65
Sj (mm) 64 64 100
S2 (mm) 90 64 65 68 scintillation camera in Figure C.2. The parameters were chosen as in case 2 except that s^ was determined according to equation 4-13 instead of equation 4-12. Although a
larger area is now imaged, the system resolution is degraded to 3.7 mm. Nevertheless, it is considerably better than the resolution obtainable with a scintillation
camera using a multiholc collimator.
4.2.5 Mechanical Components
A supporting structure is needed to position the
zone plate with respect to the camera head, as shown in
Figure 4.8. The center of the zone plate pattern must
be aligned with the center of the camera detector.
Hoivever, small deviations from perfect alignment will
not affect image quality. Grooves in the supporting
structure allow precise placement and replacement of
removable zone plates. The figure shows three such
grooves to provide for three zone plates, one designed
for each of the three cases of Table 4-4.
Bach plate consists of 1 mm thickness of lead
bonded to 1 mm thickness of aluminum to provide structural
support. The transparent zones are formed by machining
away lead, leaving the aluminum which is relatively
transparent to gamma radiation. camera supporting zone camera head frame structure plate window
Figure 4.8 Mechanical components of the zone plate system. Not drawn to scale. The zone plate is in the second of three grooves in the supporting structure. Figure 4.8 also shows a lead frame aligned with the' zone plate and the detector. The frame is to be placed against the patient's body and simply consists of a lead sheet with a square window equal in size to the field of view. Three frames must be provided to match the three fields of view obtained with each of the three zone plates. The frame is important for coded aperture imaging and extremely so when computer reconstruction is used with a germanium gamma camera. The small value of this camera necessitates choosing N as large as possible, which requires that be chosen as large as possible.
Then the shadow of a centrally positioned point source fills the detector face. Under these circumstances a point source at the edge of the field of view casts a circular shadow centered on the detector edge, so that half the shadow falls on the detector and half off. For sources outside the field of view, less than a half shadow falls on the detector. Since these sources will not appear in the reconstructed image, the counts they contribute to the detector constitute only noise. Hence it is essential that these sources should be screened off from the detector by a frame. 71
4.3 Image Processing and Display
Systems considered for digital data processing include a large-scale integrated (LSI) semiconductor chip, a minicomputer and a large computer system. The present design calls for processing by a large computer system.
In the future, LSI chips may be used extensively for data processing. The most important algorithm in the data processing routine to be discussed in Chapter 5 is the fast Fourier transform (FFT) algorithm. FFT chips are currently under development (35J which can process data faster than a large minicomputer but which cost much less: a chip measuring 0.27 in. by 0.35 in. can perform a 12-bit, 1024-point FFT in only 2.5 milliseconds. These chips are not available at the present time, however, and no chips are available to perform the image enhancement algorithms. Although such chips are not likely to be available in the near future, designers of an imaging system such as outlined here should keep aware of advances in LSI technology.
Minicomputers purchased solely for use with a gamma camera have several disadvantages compared to large com puter systems used in the time-sharing mode. A sophisti cated minicomputer system could add as much as $200,000 to the cost of the complete imaging system. Apart from 72
the cost of the central processing unit, a minimum of
10,000 eight-bit bytes of memory is needed to process the
data without image enhancement; this memory alone would
cost about $10,000. Considerable extra core space and memory would be required for image enhancement processing.
Also, extra personnel would be required in the nuclear
medicine laboratory for programming, operation, and
maintenance of the minicomputer system.
A large computer system, on the other hand, is
normally present in hospitals large enough to have a
nuclear medicine laboratory and is available on a time
sharing basis. The data processing to be described in
Chapters S and 6 can be carried out on the IBM 370 at The
Ohio State University Hospital for about $10 per patient.
Also, the availability of a large computer system permits
quantitative analysis of the image as described by
Deutchman (36).
A data acquisiton system for putting gamma camera
raw data onto magnetic tape for off-line digital processing
has been designed and built in our laboratories (37).
The raw output of a gamma camera is x and y positional
voltage pulses which can be applied to the deflection
plates of an oscilloscope to cause a positionally correct
light flash; the integration of these light flashes by 73
Polaroid camera film produces the final image or the coded image, depending on the choice of aperture. The data acquisition system digitizes the x and y pulses and records this positional information on computer com patible tape. The tape serves both as the data storage medium and the computer input medium. It can be hand carried to the hospital computer center, read in, and analyzed by software programs kept on tape at the computer center. The positional data is recorded in real time in the same order as the information is received. Hence dynamic function studies are easily performed: for example, if data is collected for 30 seconds, each
5-second segment of the tape can be processed independently, providing six sequential time-elapsed images.
Use of this system with a scintillation camera at
University Hospital has shown that results are available
in about two hours. The system hardware can be built for under $11,000. Although the present system accepts only
20,000 counts per second, one accepting 100,000 counts per second can be built for the same cost (38).
Display of the final image can be on film if the
hospital data processing system has an adequate computer-
interfaced CRT display. Our own University Hospital has not, but image printout on a standard line printer has
proved satisfactory for diagnostic purposes. 74
4.4 The Complete System
A block diagram of the complete imaging system is presented in Figure 4.9. The radioisotope distribution to be imaged is framed in the field of view. Gamma radiation passing through the on-axis Fresnel zone plate creates a coded shadowgram pattern on the camera detector.
The germanium camera translates this pattern into a
series of voltage pulses. The information contained
in these pulses is recorded on magnetic tape by the data
acquisition system. The tape is processed at a computer center. The computer output is an image of the radio
isotope distribution. On-Axis Germanium Radioisotope Fresnel Gamma Distribution Zone Plate Camera
Image of Computer Data Radioisotope Acquisition Processing Distribution System
Figure 4.9 Block diagram of the complete imaging system. CHAPTER 5
DIGITAL RECONSTRUCTION OF SHADOWGRAMS
The Fresnel zone plate, germanium, gamma camera imaging system described in the previous chapter promises a significant potential improvement in diagnostic ability in nuclear medicine. The present chapter assesses the imaging capability of this system by digital simulation.
The physical model on which the digital reconstruction algorithm is based is outlined first. This model was used as a basis for developing a computer code for digital reconstruction. The effectiveness of the code was established using digitized data derived from optical shadowgrams. For the majority of the studies, however, computer generated shadowgrams proved more useful. The essential content of this chapter is the discussion of the results obtained from the simulation of shadowgrams and their digital reconstruction.
S.l The Reconstruction Model
A description of the optical model on which the theory of digital reconstruction is based is presented
76 77 here. The detailed derivation of the theory is given
in Appendix D.
The shadowgram pattern which falls on the detector
face is a continuous intensity distribution which is
sampled by the camera. A camera with by resolution
elements results in an by matrix of sampled values,
each value being the number of gamma ray counts per
resolution element. When the system parameters are chosen
according to the principles of Chapter 4, the highest
spatial frequency content in the continuous intensity
distribution is exactly equal to half the sampling fre
quency. Therefore, the requirements of sampling theory
for acquiring complete information about the continuous
shadowgram intensity distribution are satisfied.
The reconstruction model assumes that a shadowgram
transparency is illuminated with a planar beam of coherent
radiation. The shadowgram transparency is considered
to transmit radiation only through an Nj by Nj array
of equally spaced points in which a point is associated
with each camera resolution element. In effect, the
transparency becomes an array of point sources of coherent
radiation. The intensity of each point source is propor
tional to the number of counts received in the associated
resolution element. Immediately following the array of
point sources is a converging lens which focusses onto a screen the diffraction pattern resulting from the mutual interference of the many coherent point sources. The diffraction pattern consists of a radiation intensity distribution, determined by adding vectorially the radia tion amplitudes received from each of the point sources and taking the square modulus of the resultant amplitude.
The continuous radiation intensity distribution on the
screen is the true (as compared to the conjugate) image.
The image plane in focus is changed in a real
optical system by either moving the screen or moving the
lens. Mathematically each of these alternatives is
equivalent to changing the focal length of the lens. The
focal length parameter is variable in the digital recon
struction algorithm.
In a digital reconstruction algorithm one must deter
mine how to sample the continuous radiation distribution
so as to obtain the complete image with maximum resolution.
More specifically, the size of the area in the image plane
to be sampled and the spacing of the samples must be
determined. When the sample spacing is chosen optimally,
a continuous image distribution can be generated by inter
polating between sampled points. The continuous image is
then complete in the sense that the maximum information
has been extracted from the shadowgram data. Samples 79 spaced too far apart result in loss of information; samples spaced too closely result in excessive computer time, yet provide no additional information beyond that obtainable by interpolating between optimally sampled points.
A set of dimensionless variables for characterizing the optimal spacings in the shadowgram and image planes has been defined, relations between these variables have been derived, and a set of reconstruction equations incorporating these variables has been developed. These results are presented in Appendix D, and summarized here.
Maximum information from an by matrix of shadowgram data is contained- in an by matrix of image data.
The size of the image is inversely proportional to the size of the shadowgram resolution elements. The size of the image resolution element is inversely proportional to the size of the shadowgram. In both cases, the propor tionality constant depends only on the.shadowgram plane to image plane distance.
5.2 Reconstruction of Optical Shadowgrams
Initial testing of the reconstruction algorithm was carried out by reconstructing images of point sources.
Shadowgram data was generated by accurately drawing overlapping zone plate patterns on fine mesh graph paper, ' dividing the graph paper into a 32 by 32 grid pattern, and counting the number of times each small square within a grid element was within a darkened zone. The resulting
32 by 32 matrix of data was used as input for the recon struction computer code. The ability of the code to reconstruct images of several points in various parts of the field of view and in different tomographic planes was established. Once the code was tested in this way, the ability of the code to reconstruct images of extended sources had to be established. To accomplish this, an optical apparatus was constructed to record on film optical shadowgrams of extended sources.
5.2.1 Experimental Apparatus and Procedure
Optical shadowgrams were formed using the apparatus
shown in Figure 5.1. An optical bench was constructed
from a section of an aluminum duct. A 75-watt floodlamp was mounted behind a ground glass plate. Light passing
through this plate fell on a second ground glass plate
masked by heavy black construction paper. The source
distribution pattern was cut from the construction paper.
The purpose of the second ground glass plate is to diffuse
the light emerging from the source pattern. Ideally,
light from any point in the source pattern should emerge optical large 75-watt Fresnel zone bench box floodlamp plate
ground sheet glass source power small film plate pattern screen cord box holder
timer
Figure 5.1 Experimental apparatus for casting optical shadowgrams. 82
with isotropic intensity. In practice, the intensity
is greatest for light emerging along the axis of the
optical bench. Use of the first ground glass plate close
to the floodlamp was found to reduce the nonisotropy.
The floodlamp and glass plates were entirely enclosed in
a small box with a window having an area equal to the
intended field of view of the source.
To use the apparatus, light from the source pattern
was allowed to pass through a Fresnel zone plate pattern.
This pattern was a photoreduccd transparency of a photo
graph of a large zone plate pattern, constructed with
black India ink on transparent mylar and mounted on
white cardboard. The transparency was a 32-zoned plate
with the central zone opaque. Light passing through the
zone plate was allowed to fall on sheet film loaded into
a sheet film holder. A screen with a window was placed
immediately after the zone plate to insure that only the
light passing through the zone plate fell on the film.
.A movable mount was constructed for each item on the
optical bench. The entire apparatus was contained in a
large box painted black on the inside to avoid having
light from the source reflect from nearby metallic fixtures
in the laboratory onto the film.
The apparatus was used to simulate the shadowgram
intensity pattern which would be formed on a detector by 83 a source of gamma radiation and a lead zone plate; however, there were difficulties with the simulation. The noniso tropy of the source has already been mentioned. In addi tion, some light from the source was reflected by the large box to the zone plate. Some light passing through the zone plate was reflected by the box to the film, and some was also reflected by the black surface covering the opti cal bench between the zone plate and the film. Also, the transparent zones of the zone plate transparency caused some dispersion of the light passing through.
To expose a shadowgram, sheets of 4-inch by S-inch
Kodak film No. 6127 were loaded into lightproof film holders in a photographic darkroom, carried to the labora tory and mounted on the optical bench. A timer controlling the illumination time of the flodlamp was set. After the laboratory lights were turned off, the film holder cover was raised, and the timer was turned on, causing the floodlamp to expose the film. After the timer shut off the floodlamp, the film holder cover was lowered and the laboratory lights were turned on. The film was then ready for development in the darkroom. For each experi mental setup, four shadowgrams having four different exposure times were made.
In order to maintain uniform results from one experi ment to the next, a standard film developing procedure was 84 devised. The film was processed in a sheet film tank at
68°F. Processing began with Kodak developer HC-110,
dilution D, for 2 minutes 42 seconds, followed by 8 minutes
in a fixer. After washing for 30 minutes in water and
rinsing for 30 seconds in a nonwetting agent, the films
were dried for one hour. Of the four exposures, the
one having the greatest contrast was selected for scanning.
A microdensitometcr was used to scan the processed
film and provide numerical data on film exposure. A 64 mm
by 64 mm square area of each shadowgram was scanned; this
required 64 scanning runs. On each run a square 1 mm by
1 mm beam of light passed through the film as the film
traveled across the beam. The output of each run was a
graph of the film optical density versus distance. The
graph density axis was calibrated in terms of a standard
optical density scale; the distance axis was divided into
64 segments, each corresponding to one millimeter of film
traversed. The 64 scanning runs were parallel and spaced
1 mm apart. The resulting 64 graphs provided density data
for a 64 by 64 matrix of resolution elements. Film
density is proportional to the logarithm of film exposure,
and hence it is proportional to the logarithm of the number
of photons incident at each resolution element of the
film. Therefore, ignoring a proportionality constant, one can determine an input data matrix giving the number of photon counts per resolution element. In this way
the number of gamma counts per resolution element
recorded by a germanium gamma camera can be simulated.
Even allowing for the simulation difficulties mentioned
above, the results obtained by processing such data can
reasonably be expected to simulate results obtainable
from the system outlined in Chapter 4.
5.2.2 Results
Because it was known that the reconstruction code was capable of reconstructing the shadowgram of a point
source, the apparatus and procedure for forming optical
shadowgrams were tested by forming a shadowgram from
a small circular source and using the code for recon
struction. The source was a 1/8-inch diameter circle
at the center of a S-inch field of view. The first
trial involved a 32 by 32 scan, and successful
reconstruction of this data lead to a 64 by 64
scan of the same shadowgram. The parameters of the
system for the latter case are given in Table 5-1. The
system parameters were not chosen according to the optimum
method discussed in Chapter 4 because at the time of this
experiment this method had not been developed. In fact,
it was consideration of how to define the field of view
and how to select the area of the film to be scanned, for 86
Table 5-1. System Parameters for 1/8-Inch Diameter
Circular Source Shadowgram
Parameter Value
64 mm Di
AXj 1 mm
ArN 0.4 mm
64 mm N 1
a 0.5
At 2 mm
T 128 mm
N 32
48 mm d n
S1 406 mm
203 mm S2
4. Definitions and explanations of these parameters are given in Table 4-1 and the accompanying text. 87 larger extended sources and hence wide area shadowgrams, which prompted development of the method. The distances s^ and s2 were chosen large enough so that the effect of source nonisotropy was minimal. The thickness of the zone plate can be considered zero. The reconstructed image for this small circular source is shown in Figure 5.2.
Before the image itself is discussed, a brief explanation of the method of image display is required.
The line printer code used for image display is given in Table 5-2. It must be emphasized that, although an attempt was made to correlate roughly the appearance of the line printer image with the calculated image intensity, the data presented in this form can only be properly evaluated by using the code given in the table. The calculated image is a matrix containing 4,096 values.
In the studies to be discussed, these actual values were used to evaluate the success of the reconstruction, whereas the line printer display of the kind shown in Figure 5.2 was used primarily to provide an overview of the image data and direct attention to interesting regions of the matrix.
In Figure 5.2 the reconstructed small circle is present at the center of the image. The diameter of the source circle was 1/8 inch, or 3.2 mm, in a source field 88
•**•*# 0 "M* *
• • • • • • •
•• 9 9 •• 9 * t t • 9 . * !• 9 m 9 9 9 • • • • • 9* • • • • ♦ t • • • • •• • • *1 • • • 1* • • • 9 • • • • ■ • • . • « • • • • • • I* :• • ♦ • « - • • • • • • • • • • • |* •| m • •• • • • m • • • • •
Figure 5.2 Image of a point source from an optical shadowgram. 89
Table 5-2. Line Printer Code in Which A is the
Greatest Intensity Value in the Image
Intensity Value, I Line Printer Code Symbol
0.9 A < I < A # overlaid with * f
0.8 A < I < 0.9 A # overlaid with + #
0.7 A < I < 0.8 A # overlaid with : H
0.6 A < I < 0.7 A H overlaid with . H
0.5 A < I < 0.6 A H V
0.4 A < I <0.5 A A A
0.3 A < I < 0.4 A + +
• » 0.2 A < I < 0.3 A • •
0.1 A < I < 0.2 A ••
I <0.1 A blank for which parameters were chosen so that the resolution
At was 2 mm. In the image the circle appears as a single brightest point surrounded to a distance of one resolution
element by somewhat less bright points. These results
are consistent with the analysis presented in Chapter 4.
Some noise is present in the image. At a few isolated
points the noise intensity is over 30% of peak image inten
sity; however, at most points the noise is less than 10%
of this value. In view of the simulation difficulties
discussed above, this reconstruction was considered
acceptable.
The main conclusion drawn from this experiment is
that the apparatus and procedures for forming optical
shadowgrams are workable. Hence use of the same apparatus
and procedures for an extended source should result in
a practical shadowgram which can be used to test the
ability of the reconstruction code to reconstruct the image
of an extended source.
Following this experiment, an extended source having
the shape of the letter "B" was used to form an optical
shadowgram by the method described above. The parameters
of the imaging system are given in Table 5-3, and the
reconstructed image is presented in Figure 5.3. Of the
three horizontal arms of the "EM the lower two are close Table 5-3. System Parameters for Shadowgram
of the Letter "E"
+ Parameter Value
Di 64 mm
ixi 1 mm
0.7 mm ArN
N1 64
a 0.38
At 2.6 mm
T 168 mm
N 15
43.5 mm
443 mm S1
s2 164 mm
^Definitions and explanations of these parameters are given in Table 4-1 and the accompanying text. 92
#•«»* *####• • : : : : : : t: • •:♦♦♦♦♦♦::!. • ft t + . ft# » • <++♦«•++:: ft *:♦♦♦♦♦* • • • * JJ4+4S: :: :: •ft ft* * *ft ft • ft • • • • *ft ft* *f ft • ft •» • ft ft ft 1 • ft ft ft
• ft ft# ft# ft ft ft ft ft »t ♦♦$$*>♦ • :• :* . ♦: •:: • : :i * : : :. 1♦*#♦«>♦+::..
• ■»•••» •»
• »•••••» • • >s t• * • • •
Figure 5.3 Image of the letter "E" from an optical shadowgram. 93 to maximum intensity but the upper arm has only about half maximum intensity. Also, the vertical part of the letter is better defined along the lower half than along the upper half. Background intensity between the horizontal arms varies from 204 to 504 of maximum intensity. Back ground intensity in the immediate vicinity of the ME" is in the range 104 to 304. Background intensity drops to less than 104 away from the letter but increases again toward the edge of the reconstruction. This background increase is not uniform in all directions; in one corner the background intensity is close to 504 of maximum intensity.
Two important results can be drawn from this experi ment. First, the reconstruction code is capable of reconstructing an image from digitized data derived from a shadowgram of an extended source. Although the recon structed letter is not of uniform intensity and the edges of the letter are not sharply defined, a clearly recog nizable "EM appears in the reconstruction. Therefore, the physical model for digital reconstruction presented in Section 5.1 and the theory of digital reconstruction in Appendix D are demonstrated to be serviceable.
A second result is the verification of the parametric relations developed in Chapter 4 for determining the resolution obtainable by a zone plate imaging system. The' optical source was a letter "E" of overall height 30.8 mm and overall breadth 17.6 mm. The horizontal arms and the vertical part of the letter had a width of 4.4 mm. As shown in Table 5-3, given the chosen system parameters, the resolution element in the field of view was 2.6 mm.
Consequently, the letter had a height of 12 resolution elements, a breadth of 7 resolution elements, and was composed of segments having a width of 1.7 resolution elements. In Figure 5.3, each symbol corresponds to one resolution element of the reconstructed image. With the assumption that image points having an intensity less than
40% of maximum are background noise, the reconstructed image is 12 resolution elements in height and 7 resolution elements in breadth. The width of the arms of the "E" is less clearly defined but is approximately 2 resolution elements.
This lack of definition is predictable for an arm having a width less than several resolution elements.
With an arm width of 1.7 resolution elements, there are several ways in which an arm could be aligned with a grid of resolution elements, as shown in Figure 5.4 for three cases. In case (a), the middle resolution element should ideally result in reconstruction with maximum intensity 95
one 1.7 resolution element resolution elements
resolution one resolution elements element
one resolution resolution elements clement
Figure 5.4 Alignment of ME" arms with rows of resolution elements. The width of the "E" arm is 1.7 resolution elements. while the lower element is reconstructed with 70% intensity*
In case (b), both the upper and middle elements reconstruct with an intensity equal to 85% of maximum. In case (c),
the middle element has maximum intensity while each of
the upper and lower elements has 35% intensity. Of course, cases intermediate between these extremes are more probable
in practice. The upper arm of the "E" in Figure 5.3
appears to correspond to case (c) because of the symmetry present. The lower arm, composed of two horizontal rows
of elements having an intensity above background, corre
sponds closely to case (a) in that the upper row is near
maximum intensity while the lower row is near 60% intensity.
The middle arm appears to correspond to a situation
intermediate between cases (a) and (c). The vertical
part of the ME" in Figure 5.3 also corresponds closely
to case ( c ) .
Although these two positive results were established
by the experiment, the cause of the noise present in the
reconstruction was not unambiguously identified. The
noise either could be introduced in the illuminating,
developing and digitizing of the optical shadowgram or
could be an inherent feature of the reconstruction code.
Any of the difficulties with optical shadowgram formation
previously discussed could result in the unevenly distributed 97 noise observed. In addition, the cause of the uneven intensity of the image cannot be determined with certainty, although it may be due to an anomalous background density gradient which appeared in the shadowgram transparency.
In order to determine to what extent these difficulties may be due to the reconstruction code alone, noiseless input data is needed. Por this reason, the computer generation of noiseless, digital shadowgrams was under taken next.
5.3 Reconstruction of Computer-Generated Shadowgrams
The initial motivation for generating shadowgrams by computer was to provide idealized input data for testing
the reconstruction code. However, another advantage of
computer generation is that it allows some testing which would be difficult or impossible to perform accurately with optical shadowgrams, for example, the ability to
image eithor an extended source of nonuniform intensity
or a three-dimensional source. In addition, a time
limitation is removed because the procedure for exposing
and digitizing optical shadowgrams required nearly two
weeks to complete. Therefore, once the ability of the
code to reconstruct images from actual laboratory data
was established, computer generation of shadowgrams was
substituted for optical generation. 98
5.3.1 Generation of the Shadowgrams
The shadowgrams were generated so as to satisfy the maximum resolution criteria of Chapter 4. The number of zones in the zone plate, 16, was chosen equal to one-fourth the number of detector resolution elements per side, 64, in accordance with Equation 4-10. The diameter of the zone plate shadow was chosen equal to the length of the side of the square detector; hence the equality of Equation
4-8 holds, and Equation 4-7 is automatically satisfied.
A source at the midpoint of one side of the square field of view will cast exactly half a zone plate shadow onto the detector. No values for the parameters s^, or a need be specified in order to achieve these optimal conditions.
To form a shadowgram, the source distribution was divided into an assembly of point sources. The shadowgram generation code provided an option for allowing each object field resolution clement to contain one, four, or sixteen point sources; in most studies one point source per object resolution clement was used. The intensity of each point source could be specified individually.
Each point source cast a zone plate shadow pattern of proportional intensity on the detector. A point source not at the exact center of the field of view cast only a 99 partial zone plate shadow because the shadow center is displaced and only part of the complete shadow should fall on the detector. The shadowgram was formed from the distribution of point sources by adding in each detector resolution element the contributions of the overlapping zone plate shadows.
5.3.2 Point Sources
The ability of the computer code to reconstruct the shadowgrams of point sources without noise was tested first. For a point source in the center of the field of view, the code resulted in a single point image at the center of the image field. When displayed using the line printer code of Table 5-2, the reconstruction appeared perfect in that it consisted of a single point of maximum intensity with all other points coded as blank.
A point source positioned halfway between the center and the right edge of the field of view was then imaged.
The resulting shadowgram, consisting of a partial zone plate shadow, had symmetry about a horizontal axis but not about a vertical axis. As a result, the image of the point source lacked symmetry. Because the shadowgram had a full complement of zones in the vertical direction, the vertical position of the image point was well defined, 100 that is, confined to a single horizontal set of points.
The image consisted of a correctly positioned point of maximum intensity flanked on each horizontal side by one point of intensity between 10% and 20% of maximum. In effect, the image was slightly out of focus in the hori zontal direction because the zone plate shadow did not have a full complement of zones in the horizontal direction.
Similarly, when a point source was displaced verti cally, the reconstructed image consisted of a single brightest point, slightly out of focus in the vertical direction. That is, both.above and below the brightest point was another point having much less intensity.
Next a point source positioned at the center of one quadrant of the square detector was imaged; that is, the source was displaced from the detector center both verti cally and horizontally. The reconstructed image was a point of maximum intensity surrounded by four nearest neighbor points of intensity between 10 and 20%.
These results indicate that when system parameters are chosen for maximum resolution, a source near the center of the field of view is imaged with the best resolution.
The resolution is degraded as the source moves away from the center of the field of view. It should be noted that a better uniformity of response can be achieved at the expense of overall resolution. By not choosing the zone 101 plate large enough to fill the detector, that is, by choosing it small enough so that as the source point moves around in the field of view, a small but complete circular shadow moves about on the detector, the recon structed image point will have a more nearly constant resolution throughout the image field. However, with this smaller zone plate shadow, few zones can be resolved by a detector of given resolution. By employing a zone plate of fewer zones, less resolution would be achieved
at all image points including those at the center of
the field of view.
A special problem is encountered when imaging a
point source exactly at one edge of the field of view.
It is an unavoidable fact that the digitized input data which is Fourier transformed behaves as one period of
an infinite array of periodic input data. Likewise the
calculated output is one period of an infinite array of
periodic output data. Hence when an image point on the
edge of the field of view is reconstructed, because it
is slightly out of focus it will be flanked on either side
by low intensity points. The low intensity points which
should appear just outside the edge of the image field
will appear on the opposite edge of the image field because
of the periodic nature of the output. This feature is an
inherent feature of the code. The problem can be dealt 102 with in one of two ways. First, the frame for defining the
source field of view (cf. Section 4.2.5 and Figure 4.12) can be designed to limit the field of view just enough
so that edge points are not imaged. Secondly, the input data matrix can be increased in dimension by the code from
x to (Nj + 1) x (Nj + 1) so that the x matrix
of true input is surrounded by a field of zeros.
These difficulties notwithstanding, it must be concluded
that the imaging system successfully reconstructs the image of a point source.
Also, a small group of two or three separate point
sources is successfully imaged. The resulting image is
very close to a superposition of the images which would
result from the sources imaged individually. However, if
the individual sources are of the same intensity, the
resulting image points will not necessarily have the
same intensity. Sources farther from the center will
cast less complete zone plate shadows which will in effect
diffract less light to the image point, and hence the
images of these sources will have less intensity than
centrally located sources.
5.3.3 Small Extended Sources
The ability of the imaging system to reconstruct
images of small extended sources was tested next. A 103 small extended source means here a source confined to one region of the field of view and occupying no more than
100 resolution elements.
The results obtained for imaging the letter "E" are typical of results obtained for all small extended sources.
The source pattern for the "E," Figure 5.5, was used to digitally generate a noiseless shadowgram, Figure 5.6, from which the image, Figure 5.7, was computed. The source pattern is composed of 64 point sources located near the center of the field of view. All source points are of equal intensity. The image is reconstructed with correct size, shape and position. Background noise is present throughout most of the image field; however, its intensity at most points is less than 30% of the maximum image intensity. At a few remote points and at most points close to the "E," the noise is a much as 40% of maximum.
Similar studies were performed for other small extended sources, including squares, triangles, and curved lines of irregular shape. These were positioned in various parts of the field of view. In general the results were the following. The system is able to form a clearly recognizable image of a small extended object. The size,
shape and position of the object are correctly imaged.
The image exists in a large field of low background
intensity. The size and intensity of the background field .104
#»*«#»# *# M M
*!» •0 M
Figure 5.S The source pattern for the "E". 105
•••••••••ttttltttlllttttt** ’••••••• ..... • tit: m i 44444444 it**::::rttt ••::::♦♦♦♦♦♦♦♦♦♦♦♦♦♦♦♦♦♦♦♦♦♦♦::t ::::••• ••:::44444444::*444444444444444444: • ••I t 444444444444******44444444444444: t t .• • •.:t44*44444******444*****4444 4-444444-: I : t . » • • : t :s••• • •it t44+44***********000000**************4444i :••• • •t it 11 !+4**«****00000{f3000900**********+4*4444-:: ••• •••t is: 1 *4*****0000****0#****#♦##! 99#******* 4444: : : : • « • • •••I II t44**0A0000*************4**0000040****4444:t I t . . . • ••: : t4444**000900*0000*000*004 00***9**0000*****4444:I I.. • • • 11 1444***# 000000000**************00<) *00000 H ****444441 I • •• :444*****009000#****0000fl000900***00*000000******44441.. • •I S44*******00***00»000rffl00000fle000*90*4*00*0***3**4444: { • • • I 1444*0*********0000***4444444**00009 0 ***0 ******* 1444-4441 . . • • . I 444*****0**0*00900**44444444444**000099*000*******44441 I • • •• 1144*********000000************44***00 f'0;l**0 0********4444 I •• • •I t44***0**0**00*fl00****ff00000000*******flt!l'0**0 0***0*****44! I. • I I 4**90900 0*0 90000***0000tit}9«fl#tt000*****fr«00*0000fl09*0**444t . • i:4**0*0**0**00*****0000000000000<| Figure 5.6 The shadowgram for the "E". « 106 ••••••■•••••SSS.SSSSSS..... •••••• • • • • • t ts • • • • •• •• • • • • • •:.a.sst:tsstsstssssstss4s s s • • •• •» • * • • • • •SI.. .t::::::t:::::::st::4S::.•••. •••••• • s st s 4 s t:••••••••••••••.ss 4i • •4 4f •» . •• • • SSSSS4SSSS..•..•....••.sss • *• •0 •4 *• ••••••• • » . • a#*.*.. » t . • »«**»^..»»* .s •. • * • m • •• • • • s s* * • ...... SSSSS. .t mm mmm 0 • • * *. .• • _ • • •• ss s st st ssssssss::::: t ::++«>: ss ss • • • • 4. . • • • • • m m ..... • » • • • 0 0 m • • * • ...... SSSSsssss:s:sst:s::::t:444 0 4 4 4 4s t 000mm mm 0 0 •• ■ s s s SSSSssssss:::::...ssssss 4444sss 0 0 mm 0 0 • •• • t ••:::: stss•T.....t...... »*••••»» 4444 S 0 0sss.. mmm ...... *• 0 0 0 • ••St••s ts:stss.*..»•». .... ».».*•••.. SS44S 0 0sssss 0 m • * * • • • * • • •t:: sss stt sss:: is:tssssss:::ssssss::tsss:• +0 0 0 0 0 mmm 4 44*4 0 0 0 0 0 0 • ttt:ssssssstss::tstt4S4S44s:4s::::s: • • » • 0ts9 0 0 0 0 0• m 0 0 0 0 0 • • st t:sss sss ss ssSSSSSS 4444444444444S SS tsss:: t0 0 0 0 ss0 mmm m • ss tssststsss ssSSSSS S4444444444444S SS stss:: stsss: s •• . » a * 0 * • ss sssssssssss:s:t4444ss:stsss:ssssss ssssst:00000 ssmmm .... • • • • • 0 0 0 0 +,• 0 • • ss s:ssssss* * • . :S:S444:44444444SS SSSS • • • • •: s■ * * * T 0 0t •• • * • m a 0 0■ • t ts: sssssssSSSS::: S444 4 *0 tJ* 304:4:1: 4*444SSSS 8440 0 • * * # • • a • s s+stsssssstssssss: 44440*0000094$ SSSS• • 0 0 •tsS S S44sss:.. 4 4 4 44 • • • t :♦:: ssssss• ts:t SSS44440O**«*444444SS to 0 • * * 11 • SS44• 0ts. • 4 4* • 0 s ts ••ts:sss• ss:SIS 4444400**4***4444 SS • 0 * * •• • • St S4 0 mSt. * • • 0 • st •• sss s.. :::44444004*4**44444:s tats:: s. t:: 4 0 m:•• ... 0 • s st s. •.«. S::44*440*******4444S$ ..SSSSS.8844 0m::• •».. 0 0• 0 s st s s♦♦t sss .... S t444*4*0*000900*444SS s.::: 4 S• S 544 0 000 mm • t • • • .... 0 0 * i • • • sss .... S t444444>0OO*OO40*444St :•::: 4 S• S 5 44 0 0::•• 0 • t IS s. 14- SSSS• ts:SSS 44*4400******4444$: ..ttss:• 8 S44 0 mSt. 0 0 0 • ts ..SS S t.»• s tsS *144444004****4440011 0 0. st r 4 0 m: • • • 0 0 ttt..SSSSSS• • • •S S S444440******44444S: s: 0 • • : s :4 0 0St. 0 0 • s : SSSSSS• t ssS S tS444400******0444SS sssssi:• S S44 0 0::• 0 0 • 0 XS4-SSssssssstss:st:44449O9tfoo004t:::s tsss;s:t S 844 0 ** • m m 0 0 0 * ts: t:sssssssss:::t:4444O00000O04S4t:: ss: t: t: s SS44• • *mm .. t . • • 0 0 • • • • stss» » . . « .stssSt::444S44 444444:SSSSS sssss 0 0::::♦ • 0:•• • a • 0 •• 0 • st ts ...... :: ss ts:4444tsssss:tts:stss sss st ss* a • • 00 0mmm 0 » • ss s t..ii.fi... SSSSS84444444444444S St tsss:t:sssss m m • •* • s: tssssssssss:SSSS S S 4444444444444SSt :: s ts :: sssss 0 0mmm • f * . • ». . . » 0 0 • st ss• * • . • .. . . *:ss:tt::s:44:s4:s:::t: ssssss:sssss 0 0m m • » •. • a • * • • a • ••tttttstss .... ssssss::::48sssssssttssssss• a 0 0 0 0 0 • •m m 0 0 0 0 0• • ••st ••stssSSSS t::::t::tt:»..ttrt44s 11 0 0 0 0 ■0 •• 0 0 0 • •• • t ••SSSS stss4 4 .r: 4444S s: 000 mm •••: ••»ssssss:s;:::ss:ts:t.::::::t44 44ss: ::::••• mm 0 0 0 • • • • ••SSSS ss tssssssstsss:t sss t:: S444ssssst:1 0 0 00 mm 0 0 • • • • t t: sssssstsssstss ss+ssss sssss 0 0* t m mm mm • •• • • • s s: s ts::ss::s:ss:t:s::4est:t rss s sss:•••• m • •••••• ssss11ttstss tts.00mm X X X ♦ X S X X X t (t « #•• *»••*£ t * i>a*»«»tstss::::::t s s*« *•»»*•»» tsssttss Figure 5.7 The image of the "H". 107 increases with the number of resolution elements contained in the source. This result is consistent with the theory presented in Section 3.2. 5.3.4 Large Extended Objects Large objects are not imaged well by the system. When the number of source resolution elements making up the object becomes more than 100, the background noise intensity at some points becomes equal to the image inten sity, and the noise tends to obscure the image. For a source composed of several hundred resolution elements, the image was almost completely lost in the background. An attempt was made to determine if the ability to image objects composed of many resolution elements depended on the spatial frequency content of the object. A series of computer simulations was carried out in which the object field was a bar pattern consisting of alternate strips of bright and dark bars. The spatial frequency varied from the brightest possible, in which alternate columns of resolution elements were bright bars, to the lowest, in which half the field of view was bright and half dark. In each case the number of source (that is, bright) resolution elements was 2,048. The bar pattern was not reproduced at any spatial frequency. However, for the lowest frequency, the bright half of the source 108 field was imaged as a bright area of indefinite boundary 1 in one half of the image field. The conclusion is that whenever the source occupies more than roughly 101 of the available source field resolu tion elements the image is lost in the background noise. 5.3.5 Extended Sources with Cold Spots The ability of the system to image areas emitting no radiation within a larger area of high intensity was investigated next. The results obtained with large extended sources indicated that if the field of high intensity is too large the resultant noise would obscure the cold spot, just as the dark bars of the bar pattern were completely obscured by the noise from the bright bars. In studying the ability of the system to image cold spots therefore, the scope of the study was restricted to cases in which the high intensity area was composed of a small number of resolution elements. The first class of objects to be studied was squares, composed of four line segments each of one resolution element thickness. A small square, four resolution ele ments on a side, was imaged well. All points on the square were reconstructed with maximum intensity. Inside the square were nine resolution elements, five of which imaged with less than 10$ of maximum intensity and four, 109 in the corners, with less than 20%. The noise outside the square was almost everywhere less than 10% of maximum and never more than 20%. However, a small amount of noise appeared in a regular pattern far from the image of the square. Hence the system is able to reconstruct cold spots in very small objects. However, as the size of the square increases this ability is less apparent. The small square above was outlined by 16 resolution elements. A similar square composed of 48 resolution elements was imaged. All points on the square reconstructed with maximum intensity, and all points not on the square had less intensity. The appearance of the background outside the square was very similar to that around the letter "E" in Figure 5.7, that is, it filled most of the field of view but was generally under 50% of maximum intensity. Inside the square the noise varied approximately bewteen 35 and 75% of maximum. The noise pattern was not random. Because of the symmetry and absence of noise in the input data, the reconstruction image noise was symmetrically distributed in a regular pattern inside the square. Another similar 80-element square was imaged with similar results. Background was greatest near the line segments making up the square and extended for an average 110 of eight resolution elements on each side of the line segments with a fairly uniform intensity of 60%. With the imaging of a 112-element square, some change in response begins to appear. The size of the square has become large and its corners are distant from the center of the field of view and hence they recon struct with intensity as low as 70% of maximum. The noise level outside the square is low, lower in fact than with the 80-element square. Inside the image square, however, there is considerable background noise which is no longeT of greatest intensity closest to the line segments of the square. Instead, the noise is greatest in an approximately annular region inside the square. At a few points the noise intensity exceeds the intensity of the reconstructed corners of the square. These new features were more pronounced in a 144-element square. The points of maximum intensity in the recon struction were in a centrally located, approximately circular group of background points; the diameter of this circle was equal to half the diagonal of the square. The image of the square was clearly visible amid the low (up to 40%) background in the neighborhood of the line segments. But because the line segments had intensity values of only 40 to 70% of maximum, the image contrast H i was low. It should be pointed out here that although the number of elements in the square was not very large, the positions of the source points making up the square were well removed from the center of the field of view. An interesting response was obtained for two larger squares. A 176-element square was not imaged at all. The reconstruction consisted of a small bright noise formation having radial symmetry, and no hint of the line segments of the square appeared at the correct positions even though the noise level at these places was not excessive. However, a larger square composed of 208 resolution elements resulted in a very low contrast, but nevertheless visible, reconstruction of the square. Because the inverse of the distance between opposite sides of the square is the fundamental spatial frequency of the reconstruction, this result suggests that the imaging system may have a spatial frequency response which is not monotonically decreasing with increasing spatial frequency. Studies performed on the hollow squares described above indicate that the smaller sources containing relatively large cold areas are well imaged. Consider next small sources with cold spots covering only a few resolution elements. 112 A hollow 12 by 12 resolution element square was gradually filled in from the outside so that a cold area of diminishing size was at the center of the square. Initially the square was composed of sides two resolution elements thick and so contained a cold area measuring 8 by 8. In the image this cold area was reconstructed with a noise level ranging from 40% to 70% of the true image. The highest noise level occurred in the neighbor hood of the center of the cold area, and the lowest toward the outside. As the square was filled in with sources, not only did the size of the cold area decrease, but also the number of point sources in the object increased and hence background noise increased. With a 6 by 6 cold area the noise level in the cold area of the image grew to 50-80% of maximum image intensity. For a 4 by 4 cold area, the range was 60-80%. A 3 by 3 cold area was imaged with a uniform intensity equal to 70% of maximum. The same result was obtained for a 3 by 2 cold area and also for a 2 by 2 cold area. However, with an increase in the number of resolution elements in the source, the intensity of the reconstructed square decreased at the edges to 80% and at the corners to 70%. Hence although the cold area was visible with low contrast, the reconstructed intensity of the cold spots inside the square equaled the 113 intensity of reconstructed corners. For cold spots con sisting of three, two, and one resoltuion element, the results were the same: the cold spot images were recon structed with less than 70% of the intensity of the surrounding true image but the corner points of the square had an equally low intensity. However, even with no cold spots in the 12 by 12 square, the decrease in intensity at the edges occurred. Hence small cold spots in sources can be imaged, but the contrast obtained is poor. It is difficult to dis tinguish a true cold spot of zero intensity from an area of reduced intensity.■ All the studies described to this point have dealt with sources of uniform intensity. 5.3.6 Sources of Nonuniform Intensity Studies were undertaken to examine the ability of the system to reconstruct sources of nonuniform intensity with correct relative intensities. Initially two point sources of different intensities were used. The sources were spaced 10 resolution elements apart and each had an equal displacement from the center of the field of view. Several relative intensities were used, and the results are presented in Table 5-4 in terms of the intensity of the less intense source relative to that of the more intense source. The table shows that 114 Table 5-4. Relative Intensity of Weaker Source and Weaker Image Relative Intensity of Relative Intensity of i, Less Intense Source Less Intense Image (4 of more intense source) (4 of more intense image) 90 80 80 65 70 SO 60 37 50 27 40 18 30 11 20 6 *These results are average values derived from a graph of computer-generated data. 115 the less intense image is always reconstructed with less than the correct intensity value. This effect occurs because the strong noise introduced by the presence of the stronger source interferes with the reconstruction of the weaker image more than the weak noise of the weaker source interferes with the reconstruction of the stronger image. In imaging two sources of unequal intensity, the imaging system will tend to suppress the image of the weaker source. This result was checked in tests involving two extended sources of uniform, but unequal intensities. The less intense extended source was usually imaged with correct relative intensity at its centermost image points but with less than correct intensity at those image points along its border. The relative intensity of the less intense image was correct and most nearly uniform when the two extended sources were continuous at the center of the field of view. As the sources separated, and especially when they moved toward diagonally opposite corners of the field of view, the uniformity of intensity in each of the images was lost. These results were obtained for a wide range of values of relative intensity. The results appear to be inconsistent with the low intensity suppression observed for point sources. However, interpretation of the results is complicated by the fact that the system does 116 not produce uniform image intensity even for a single uniform extended Lource, and by the fact that the back ground of the stronger image adds to the intensity of the weaker image. 5.3.7 Small Strong Sources in a Weak Source Field The ability of the system to image a small but relatively intense source in the presence of a much weaker source field was examined for two cases. First, the entire object field of view was filled with a back* ground source field of constant, low-level intensity, I . Secondly, the entire object field was filled with back ground sources such that each source intensity was chosen randomly between zero and a given upper limit, Iu , this limit being of low intensity compared to the strong source. For both cases the source was a hollow four by four square. The case for which the weak background source was uniform was examined for various values of I. For an Ic value equal to 25% of the strong source intensity, no image of the square was formed. For I equal to 10%, an extremely low contrast image of the square was formed; the image background intensity in the region of the image of the square was 75% of the square's image intensity. For I_ B 5%, the corresponding background value was 65%. 117 Similarly, for Ic values of 2.51 and II, the image back ground values were 551 and 251, respectively. The case for which the weak background sources were of random intensity was tested for various values of Iu . As in the previous case, for an Iu value equal to 251 of the strong source intensity, no image of the source was formed. For Iu a 101, the image was present amid an image background intensity equal to 651 of image intensity. For Iu equal to 51, 2.51 and 1%, the image background values were 451, 351 and 151, respectively. No significant background occurred inside the small square in the absence of the background source field. These results show that for a Fresnel zone plate imaging system an extremely high contrast between a strong source and widely distributed, weak background sources is needed to achieve acceptable contrast between image and image background. This difficulty can be lessened by reducing the viewing area defined by the mechanical frame (Figure 4.12) to a smaller value than that required by the para metric considerations of Chapter 4. 5.3.8 Sources in Different Tomographic Planes The ability of the system to perform tomographic studies was tested next. Objects in different tomographic planes cast shadow patterns of different radii. As shown in Appendix D, the image plane in focus in the reconstruc tion depends on the choice of the parameter Af in Equation D-44, which in turn depends on the choice of the parameter R^ in Equation D-26. R^ corresponds to the radius of the first zone of the Fresnel zone plate which would focus light in the chosen image plane. Therefore, if a source is positioned so that it casts a shadow of first zone radius Rs , and if R^ in the reconstruction code is chosen equal to Rs, then that source's image will be in focus in the reconstruction. A source in another plane will appear out of focus to the extent that this other plane is distant from the first. In the following studies the source plane and the image plane will be specified by the value of Rg and Rj, respectively, measured in units of number of resolution elements in the detector plane. In all the studies discussed above, Rg and R^ were both equal to 8.0. Initial tomographic studies were performed on point sources. One point source was positioned in plane 7.5 (that is, it cast a zone plate shadow having a first zone radius of 7.5 resolution elements in the detector plane), and another source was positioned in plane 9.5. Reconstructions were computed for planes 7.0, 7.5, 8.0, 8.5, 9.0, 9.5 and 10.0. In all planes except 7.5 and 9.5, 119 the images were severely out of focus, and background "noise" at a level generally below 201 of maximum was * distributed throughout most of the image field. This "noise" was actually just the diffraction pattern calcu- lated at an out-of-focus plane. One point source was reconstructed in each of the planes, 7.5 and 9.5. The resolution was good and the noise level was everywhere well below 10% of maximum. Another experiment placed one point source in each of the seven planes mentioned above. In this case also each point was reconstructed in the correct image plane by being the only point of maximum intensity in that plane. However, in each reconstruction, noise was present due to the out-of-focus sources. The reconstruction in plane 8.0 had the least noise; the out-of-focus sources of planes 7.5 and 8.5 appeared as small rings of intensity less than 20% of the focussed image point intensity. For other reconstruction planes the noise was greater, becoming considerable in planes 7.0 and 10.0. The reason is that the plane 8.0 is the source plane for which the system is designed to give the best resolution; it is the plane for which the zone plate shadow has parameters satisfying the maximum resolution criteria of Chapter 4. Hence the system is fully able to reconstruct tomo graphic images of point sources. However, out-of-focus 120 images of points in other planes can appear; and without ' a priori knowledge that all sources are points and are of equal intensity, the true positions of these out-of-focus images is not well defined. Images of small extended sources in different tomo graphic planes were considered next. A 16-resolution element triangle in plane 8.0 and a 16-resolution element square in plane 9.0 were used as sources to form a shadow gram. Image planes 7.0, 7.5, 8.0, 8.5, 9.0, 9.5 and 10.0 were reconstructed. This procedure was followed for various separation distances of the triangle and square, where separation distance means the center-to-center distance when the two sources are projected onto a single source plane. The results obtained were virtually inde pendent of separation distance, including the case where the two source patterns overlapped. The results illustrated in Figures 5.8, 5.9 and 5.10 are typical. Figure 5.8 shows plane 8.0. As expected, the low intensity background noise associated with small extended objects exists in most of the image plane. The triangle is in focus and lies to the upper left of center. Its intensity varies from maximum to about 551 of maximum. To the lower right of center is the out-of-focus square. Its intensity at several points is as high as approximately 65% of maximum. 121 • ft# • • ft • : » * • s • ftftftftftftftftf ?* ft ftft ft ft ft ft ft • • ft ft t • • • • ft • • • • ft ft ft • • • • • » « » ftft ft ft ft »• ft ft ft ft ftft ft ft • ft • • • • • * • • • * •• • • • ft ft ft ft ft* ft ftft ft • • ft ft • • • • • • •# • • » ••••*•••••#• ft ftft ft ft ft ft • ft* * • • • • * • • * • • ft S * • * # • ftft • • ft ftft ft ft ft ft ft • • ft • • • • • » • • • • • ♦«»•#••••»#• ft ftft ft ft ft ft • • • • •• ••••••** ft* ft ftft ft ft ft ft ft ft ft* ft ft ft ft ft ft ft ft ftft ft ft ft ft ft ft ft ft ft ft ft ftft ft ft ft ft ft ftftft • • • * • • • • ft ft ft ft ft ft ft ft ft ftft* *••••••••* ftftftftftftftftftftftftft • • • • • ft ft ft ft • ft ft ft ft ft ft* •••••••••* * * • •••••••••• • • • • • • ft •• ft ft ft ft ft • ft •• • • • • • • • ft ftft ft ft ft ft •••::• •ftftfttetftf*.::**+:*:•••••• • • •• • • • • ft ftft ft ft ft • ••* t. •••tuts***:•••••• • • • •• ■ ft ft ft ft ft ft • • • • • •• ® **•••••••»#•• • • • • • • • ft ft ft ft ft • • • • • •••••••tfttS***+++«ftftftftftftft • • • • • • * ft ft • • • • *••:•••**:*:*::♦♦+*•••••:• • • • • • • ft ft • ft • • • • •« • ••• ***44*:••••*• tft ftft ftft ft• • • • ••.j.;: : : : • * s* • ft ft ft ftft ft ft• • • • ••ftftftfttftSS »• ** •• • • •• • • ft ftft ft• ftft* • • • •• •••: .:::•* ♦*###$*♦* •*: • • • • • • • • • • • ■ • • • • • • • • • • • • • • • • • • • • •••••••••••••***••• • • • • • • • ft ftft ft ft ft ft • • • • •• • •••••••»•***•••• • • • • • • • ft ftft ft ftft ft • :*: :• ft ftft ft ft ft • • * * * ■ ft ft ft ft ftft ft • ••••••••••••••••• •••••••••••• • • • • • • ft ftft ft ftft• ft ft• ft i «ftftftftftftftftftft* **•••••••••• • • • ftft ft ft ft ft ft ft ftft•••••••••••••••••••••••• • • • • • ftft ft • ftft • ft ft ft ft ft ft ft ft ft ft ft ft ft ft ft ft ftft ft • • • • • • *•••••# •••••••••ft*** • ft • • • ft ft ft ft ftft Figure 5.8 Image of triangle and square focussed on plane 8.0. 122 • • • * * * * • • • • • > • • * * ••••••• +fttt++: ••••*•• ss#ss ********** ••*22** * • • • •••: *•••*:::• • * * * * • • • • • • • * • •• • • *• • *• • * • • ***** * # * * • • • • «••*< • •••••• **•* Figure 5.9 Image of triangle and square focussed on plane 8.5. 123 ftft ft ft ft ft ft ft ft ••••••••••••ft ft ft •• ■ ••••••••• • •• ••• • • • • • • ft ft •* * * • •••••••• ♦ ♦ • • • • •• • • • ft ft • • • • •♦•••■•ftft• • • • • ••• • * • ft ft • •• • • ■ • • • • •■ • • • • • ft ft •• ••• • •♦ ♦ ♦ ♦ ■ ♦ ♦ tS••ft•• • •• • ft ft • • • • • • • • • # • •* t ft ft • ft• •» ♦ • ♦ ■ ■ • ft ft • • •• * • ■ ft • ft ft ft ft• •• ■ * ♦ • ft ft• • • • •»• •»++**++:* • * • 9 *♦ ♦ • • • • • • ft ft *• ••»•»• ;+*+♦:: £ S • ft ft ft• •• • • • • • ft ft • • • • • » • • $ «• • ♦• »• ft •• ■ ♦ • • • • • ■ ft ft • • •• • • •••••• • • • • • • • • • ft ft » * •» • •* £♦ s: r:: • ■ ♦ •• •• • ♦ :• • • •• ••• • • •■••ftftSft+**#•• ■ • • ♦ • • ♦ • • • • • • • •• • • • $:♦+:+» tfl**.ft • • •• • • • • •»«• ■ # :#40ftft • • • • • • • • • • ftftftSft*ft• ••••• • • • * • • • • • t ti • • • ♦ ft • ■ ♦ * ■ ft ft k ft < ft • ft • ft ft •••I ft • ft • • • ft •••••••••••ft • • • • • ft ftft ft • • ft ft Figure 5.10 Image of triangle and square focussed on plane 9-. 0. 124 Figure 5.9 shows plane 8.5. Neither source is in focus. In Figure 5.10 the square is in focus in image plane 9.0. However, only a portion of the complete square is imaged with high intensity. The intensity of the out-of-focus triangle becomes at one point as high as about 55% of maximum. These results show that an extended object does not disappear from the reconstruction if the plane it exists in is not precisely in focus. Instead, the image of the object may appear, somewhat out of focus and at somewhat reduced intensity. In further experiments the distance between the tomo graphic planes containing extended sources was varied. The ability of the system to distinguish tomographic planes increases as the distance between the planes increases. This is because any figure, a triangle for example, if only slightly out of focus, still appears as a triangle in the reconstruction. However, the ability of the system to reconstruct a high resolution image of a source decreases as the source becomes more distant from the source plane for whcih the system is designed to achieve its best resolution. Attempts to image small cold spots at the center of three-dimensional sources were not successful, with the exception of the simplest case. A small cube, three resoltuion elements on a side, with hollow center was 125 imaged. When the image plane passed through the center of the cube image, the resulting two-dimensional image was a three by three square with a central resolution element having only S0% the intensity of the points making up the square. 5.4 Summary of Results The optical simulation experiments yielded two basic results. First, the reconstruction code is capable of reconstructing an image of an extended optical source using digitized data obtained by scanning an actual experi mental shadowgram. Second, the parametric relations of Chapter 4 for determining the resolution characteris tics of a Fresnel zone plate imaging system correctly relate source field parameters to image field parameters and correctly predict the zone plate parameters required for maximum system resolution. The computer simulation experiments examined the ability of the reconstruction code to reconstruct shadow grams of a variety of idealized source distributions. The system reconstructs images of point sources with good resolution, but the resolution and intensity of the image point depend on the position of the point source in the field of view. The system forms images of small extended 126 sources but not of large ones. Image background intensity fills most of the image field except for the smallest f sources. As source size increases, the background inten sity also increases. When the source covers 10% of the resolution elements in the source field the image becomes totally obscured by the background. The ability to image cold spots in an otherwise intense source depends on the size of that source. The image background, which increases with source size, tends to fills in the cold spots in the image. This makes it difficult to distinguish a true cold spot of zero intensity from a source region of reduced intensity. On the other hand, with separated sources of unequal intensity, the imaging system will tend to suppress the image of the weaker source. When a strong source is imaged in the presence of a weaker, but widespread, field of background sources, this background intensity must be no more than a few percent of the strong source intensity in order to obtain an image with acceptable contrast. The ability of the system to distinguish tomographic planes increases with the distance between the planes. However, the plane to be imaged cannot be too far distant from the plane for which the system gives maximum resolution without seriously degrading image quality. CHAPTER 6 IMAGE ENHANCEMENT STUDIES The computer studies of the previous chapter were intended to simulate the reconstruction results which would be obtained with ideal optical reconstruction in the absence of noise introduced by photographic and optical irregularities. In the present chapter, methods for improving the image by digital processing are considered. Although these methods can in theory be employed with optical reconstruction, in practice they require such time and painstaking attention to detail that they could not be used routinely in a nuclear medicine clinic. 6.1 Introduction Numerous image enhancement techniques have been developed for application in such fields as television reception, electron microscopy, military reconnaissance, and space exploration. Application of these techniques to the present study is limited by four considerations. First, a technique may not apply because the noise source is not found in the Fresnel zone plate system. Second, 127 128 the amount of computer time and storage may be excessive • in relation to the expected image improvement. Third, enhancement processes intended to improve image resolution can sharpen features only at the expense of the overall signal~to~noise ratio (SNR). Such processes may have a net disadvantage in a system with a low SNR. Lastly, the linearity and spatial variance characteristics of the Fresnel zone plate system may differ from those of other systems. Many image enhancement techniques were not considered for the reason that the noise source was not present in the Fresnel zone plate system. For example, in vidicon tubes there is often significant frame-to-framc image retention which requires removal. Space satellite images require removal of periodic noise arising from the coupling of periodic signals related to the rastcr-scan and data sampling devices. In some systems the noise pattern varies only slightly from one image to the next. This allows identification and removal of the noise pattern. In the Fresnel zone plate imaging system, as discussed in Chapter 5 and Appendix B, the noise pattern depends on the size and shape of the object to be imaged. A digital masking technique was studied but was found to be of questionable value because it often degraded the SNR. In the Fresnel zone plate imaging system, image' points near the edge of the field of view are reconstructed with about half the intensity of points near the center because the corresponding point sources cast half and full zone plate shadows, respectively, onto the detector face. A computer code was written to correct for this situation. Although the code worked well with a few isolated point sources, it has little practical advantage for extended sources. Large extended sources were not imaged well by the system with or without the code. Small extended sources were imaged best near the center of the field of view, in which case the code would reduce the intensity of the central image and increase the intensity of peripheral noise. Although the code could result in image improvement in some cases, in practice the clinician will normally attempt to align the organ of interest with the center of the field of view. Hence this code does not appear suitable for use as part of a standard image enhancement package to be applied to every image. 6.2 A Fourier Transform Technique Several image enhancement techniques rely on the existence of a transfer function for the system. Components 130 of a possible transfer function for shadowgram formation can be identified. The following definitions will be useful. circ(r) * (6 -1) rect(x) (6-2) r ^ comb(x) » ) 6 (x - n) (6-3) n*»-» s m n x sinc(x) (6-4) TTX A uniform field of gamma radiation is transmitted by the zone plate according to the transmittance function, g(r), given by 3 g(r) - \ - 3- Y, p ©xpUnpr2/^2) (6_s) p»*® 131 For a finite zone plate of radius R^ the transmittance is gfy/r) “ g(r)circ(r/RN) (6-6) The gamma distribution arriving at the plane of the detector face is given by gd (r) - ** . . J 2 S r n Cr) (6-7) This distribution is sampled by the rectangular array of resolution elements, each of dimension AxQ , giving the following sample values 8 s Cx,y) - gd (x,y)comb comb (j£-) (6 -8) where gd is transformed to rectangular coordinates. Finally, because the detector is a finite square, with the length of each side equal to D, a*finite.number of sample values results, given by gD (x,y) - gs (x,y)rect (jj ) rect . (6-9) 132 The spatial frequency content of the sampled shadow gram is related to the spatial frequency content of source distribution by the Fourier transforms of these functions. In the frequency domain, the functions in Equations (6 -6) through (6-9) must be replaced by their Fourier transforms, and each multiplication must be replaced by a convolution integral. The Fourier transform of g(r), given in Appendix B, involves an infinite summa tion of terms, each containing integrals of exponentials. The Fourier transform of circ(r/R^) is JjUnRjjf) v where Jj is a Bessel function of the first kind, order one, and f is the two-dimensional spatial frequency equal to (fx 2 + fy2)1^2 . The ratio (Sj + s2)/s1 is simply a multi plicative scaling factor. The Fourier transform of comb ( g~-) *comb (^T") is (Ax0 )2comb(Axofx)*comb(Axofy). Finally, the Fourier transform of rect ( jj ^ rect ^ jy ) is sinc(Df^)sinc(Dfy). The expression resulting from successive convolutions of these functions is at best difficult to work with. Such an expression would relate 133 the Fourier transform of the source intensity distribution to the Fourier transform of the shadowgram intensity distribution. Relation of shadowgram intensity to image intensity involves more than simple multiplication of functions. In Appendix D it is shown that each sampled shadowgram intensity value must be assigned a phase value depending on the position of the resolution element recording that value and on the distance to the image plane. The resulting array of complex values is Fourier transformed, and the image intensity is found by taking the square modulus of each element in the transformed array. These considerations lead to the conclusion that relating source and image spatial frequency contents in the frequency domain for the purpose of deriving a system transfer function in closed form is not possible. It was shown in Chapter 3 that a transfer function exists only for a system which is linear and spatially invariant. A system is invariant if the image of a point source changes only in location, not in functional form, as the point source moves about in the field of view. In Chapter 5, we found that this condition does not hold for the Fresnel zone plate system. For example, a point source moving vertically up from center casts a partial zone 134 plate shadow, and the moving image point loses definition in the vertical direction. Hence the functional form of the image point is position dependent, and the system is spatially variant. A system is linear if the principle of superposition holds. We found in Chapter 5 that a low intensity image point is suppressed by the system in the presence of a high intensity point. Then the image resulting from two point sources of different intensities is not simply the superposition of images for the two points taken separately. Hence the system is nonlinear. These considerations indicate that no transfer function exists for the Fresnel zone plate imaging system. Computer studies confirm this result. The method for determining the transfer function of a system, discussed in Chapter 3, was applied to the computer model described in Chapter 5. This method consists of taking the Fourier transform of the image .resulting from a point source input. If a transfer function exists for the system, the result is independent of the point source position. With the computer model used in this study, the result was found to vary with the position of the source. With the source in the center of the field of view, the results from the transfer function calculation show a relatively flat frequency response, as depicted in Figure 6.1. The complete result is a 64 by 64 array covering positive and negative frequencies and having rough radial symmetry. The solid line shows the response to a test pattern with test lines parallel to the rows (or columns) of resolution elements in the detector array. The dashed line is for the case in which test lines are oriented diagonally. For intermediate orientation, the frequency response is between the two lines shown. There is a small spike at zero frequency indicating that the mean value of the transmitted intensity is nonzero. Other spatial frequency components are transmitted with good response: the response to the highest frequency is never less than one-third that of the lowest nonzero frequency. With the source in the corner of the field of view, the result is different, as shown in Figure 6.2. There is considerably less radial symmetry. The response to diagonal test lines is much poorer at all frequencies than the response to parallel lines, decreasing essentially to zero for the two highest frequencies. For parallel test lines the response at the highest frequency is one- tenth that at the lowest nonzero frequency. As the source point moves around in any given neighbor hood of the field of view, the results of the calculation 136 10 Modulus of the Frequency Response 10 10 10 20Af 30Af spatial frequency Figure 6.1 Results of transfer function calculation for a point source at the center of the field of view. Test lines parallel to resolution elements (solid line) and diagonal (dashed line). 137 10 Modulus of the Frequency Response 10 f 20 f 30 f spatial frequency Figure 6.2 Results of transfer function calculation for a point source in a corner of the field of view. Test lines parallel to resolution elements (solid line) and diagonal (dashed line). 138 change very little. This suggests that image enhancement ‘ may be possible for small extended objects confined to one neightborhood. Although it is possible to calculate an "average" transfer function for each neighborhood of the field of view, only the result for the neighborhood of the center is useful in practice. For other neighborhoods the frequency response is so low at high frequencies that division by the transfer function results in great ampli fication of the high frequency content of the image. In a neighborhood where the system has very poor ability to transmit high frequencies, the high frequency content of the image consists mostly of noise. Then the result is to amplify greatly the noise content. The applicability of the "average" transfer function, derived for a central point source, to zone plate system image enhancement was studied. The technique was discussed in Chapter 3. Briefly, this technique consists of calcu lating the Fourier tranform of the image and dividing it by the system transfer function to obtain the Fourier transform of the source distribtuion, and then taking the inverse Fourier transform of the latter to recover the source distribution. The study began with an attempt to improve the image of a corner-located source using the transfer function for a central source. A significant 139 increase in the resolution of the image resulted. This result suggested that the method may be useful for improving the resolution of any image formed by the system and particularly useful for small extended sources in the center of the field of view. The image enhancement technique was tested for various small extended objects including curved lines and hollow squares. In all cases, excellent results were obtained. A typical result is the improvement of the image of the letter "E", which was shown previously in Figure 5.7. The processed image is shown in Figure 6.3. The absence of significant noise around the image is striking. However, the intensity of the various parts of the image is not uniform. These results applied to most processed images of smaller extended objects. For the smallest objects, the intensities were also uniform. For intermediate sized objects, some noise was present. Images of somewhat larger objects were also improved. Although significant improvement occurred in all cases tested, the quality of the improved image still decreased with increasing object size. This effect was demonstrated by studying the processed images of hollow squares of various sizes. The unprocessed images were described in Chapter 5. For a square containing four resolution ele ments on a side, processing removed the low-level, periodic #* Figure 6,3 Enhanced image of the letter 141 noise appearing around the image. The final image was a clean image of a square of uniform intensity. The image of a square of side nine units showed excellent improvement. Whereas before processing almost the whole field of view contained background noise at intensities up to 30% of maximum, after processing only one point (inside the square) had a noise level above 101. For a hollow square of length 17 (having 68 point sources altogether), some low-level, isolated, periodic noise appeared after image enhancement. In the case of a 25- element length, much of the inside of the square was filled with noise, usually under 201 of maximum, but occasionally up to 30%; nevertheless, the image was quite distinct and the unit width of the sides was well resolved. Before processing the thickness of sides of the image square was uncertain because of the noise level. For larger squares the results were less acceptable, even though some improvement was obtained. The image of the square of side 37 was almost completely lost’in background noise before enhancement. After enhancement the image was clearly present but of very low intensity, and the corners of the square did not appear. The most intense portion of the reconstruction was the noise, consisting entirely of an intense round artifact located 142 at the center of the hollow square. Virtually identical results were obtained for all larger squares. For very large objects the system formed no recog nizable image, and use of this technique to enhance the nonexistent image produces no result. It may be appro priate to reiterate here a point discussed in Chapter 3. Image enhancement for nonlinear and spatially variant imaging systems is a relatively new field of study. Although the image enhancement technique discussed above leads to good results in some cases, further research in this area could lead to better results for more, and possibly for all cases. The image enhancement processing described here leads to excellent results for small objects and improved results for intermediate sized objects. In no case is the image degraded by the processing. Hence this technique should be included as a standard part of the digital reconstruction code. CHAPTER 7 CLINICAL ADVANTAGES AND LIMITATIONS OF THE SYSTEM In this chapter the potential clinical applications of the Fresnel zone plate, germanium gamma camera are discussed. The applicability of the system depends on the size of the radioisotope distribution to be imaged. In nuclear medicine images are formed either of the radio isotope distribution in a whole organ of the body or of the radioisotope uptake by a lesion in an organ. The applicability of the system for each of these studies is discussed. 7.1 Limitation on Source Size The digital simulations of Chapters 5 and 6 demon strated that the Fresnel zone plate system forms good images of small sources, noisy images of intermediate sized sources, and no images of large sources. In the nuclear medicine clinic a germanium camera with a Fresnel zone plate aperture would have an advantage over one with a conventional collimator for imaging small 143 144 structures. The advantage consists of better spatial resolution and the choice of either a greater signal-to- noise ratio (SNR) or a shorter collection time. On the other hand, the system is limited to imaging structures of small size. For larger structures, a conventional collimator is to be preferred. In addition, a conventional collimator is better for imaging a cold spot within a large area of higher activity. The size of a structure must be considered relative to the area of the camera detector. Because of the present cost of high purity germanium, the first germanium gamma cameras produced for clinical use will be relatively small. However, as the technology of high purity german ium crystal growth advances and as the market for germanium increases, the cost is expected to decline and larger germanium cameras may become available. Then the size of a structure imageable with a zone plate system would increase. The size of the detector in the system described in Appendix C is a square with a length of about 15 cm on each side, containing 4,096 resolution elements. In Chapter 5, we found that this system produces no useful image if the number of resolution elements contained in the source exceeds 101 of the total number contained in 145 the field of view, that is, about 400 resolution elements.' Another limitation on the size of the source is that size at which the SNR for the zone plate system falls below the SNR obtainable with a conventional aperture. Appli cation of Equation 3-S gives the result that for point source imaging the system considered here has a SNR 20 times that of a pinhole aperture. The SNR's for the two apertures become equal when the number of resolution elements in the source equals 400, by coincidence, the same result as before. The size of a resolution element in the source field of view depends on the magnification parameter, a. Assuming that a is chosen to obtain optimum system reso lution, the theoretical resolution of the system, given in Table 4-4, is 1.7 mm. Then 400 resolution elements would be contained in any source distribution having a frontal area of 11.6 cm ; for example, a square of side 3.4 cm or a circle of diamter 3.8 cm. Structures smaller than these would be imaged better with a zone plate system than with a conventional collimator. For larger struc tures, the conventional collimator is better. A larger structures can be imaged by the zone plate system by decreasing a, but the spatial resoltuion of the system is degraded so that less detail of the larger structure can be imaged. 146 7.2 Imaging Whole Organs In the nuclear medicine clinic, gamma cameras are used to form images of the distribtuion of a radiopharma ceutical in many organs of the body for the purpose of assisting in the diagnosis of disease. The organs normally imaged are the brain, thyroid, lungs, heart, liver, pan creas, spleen, kidneys and bone. Various kinds of studies can be performed. For example, to study the ability of a kidney to filter waste products from the blood, a pharmaceutical normally filtered efficiently with three or four passes through the kidneys is tagged with a radioisotope and injected intra venously. The field of view of the gamma camera is set to include the kidney and the bladder. Sequential images show the efficiency of the kidney in collecting the radio pharmaceutical and delivering it to the bladder. In this example, the clinician is interested in examining the functioning of the organ as a whole. In other cases, it is desirable to image the whole organ in order to find a pathology in part of the organ. For example, because the normal lung is richly supplied with blood, the lung can be imaged by adding a radiotracer to the blood. If some part of the lung is receiving little or no blood, a cold spot appears in the lung image. 147 The cold spot shows the existence of disease, and the appearance of the cold spot may indicate the nature of the disease. Typically, a large cold area of the lung is associated with pulmonary edema, an effusion of fluid into lung tissue; and a small cold spot indicates a lesion, an abnormal discontinuity in the tissue of an organ. The smallest organ normally imaged is the thyroid. Although there is considerable variation in thyroid size and weight, the average human thyroid weights 20 to 25 grams and has a thickness of 1 cm and a frontal area of 2 20 cm . The largest structure which can be imaged with the zone plate system considered here, when that system is set for optimum spatial rcsoltuion, has a frontal area of 12 cm . Assuming the distribution of the radiopharma ceutical is approximately uniform throughout the organ, the thyroid is too large for imaging with this system. In addition, the tomographic capability of the system within a one-centimeter thick organ is not impressive. Hence a gamma camera with a 15 cm by 15 cm germanium detector using a conventional collimator would have a performance superior to one using a zone plate aperture in all applications requiring imaging of a whole organ. However, this conclusion might not apply to a camera with a larger germanium detector. 148 7.3 Imaging Small Lesions Gamma cameras are also used to detect lesions within organs. Two types o£ lesions important in nuclear medicine are neoplasms and infarctions. A neoplasm is any abnormal growth such as a tumor. An infarction con sists of dead tissue resulting from the cutoff of the blood supply due to either a blood clot or a hemorrhage. The size of both types of lesions can vary from micro scopic size up to 3 cm or more in diameter. The location of a lesion within an organ, as well as its size, will determine whether a patient notices symptoms of his condi tion. As a result, there is not a standard size at which lesions may produce those clinical symptioms which bring the patient to the nuclear medicine clinic. Lesions of any size may exist in the diseased organ. In practice, more large lesions are detected because they are more imageable. One important goal of nuclear medical research is to increase the probability of detection of the smaller lesions. Lesions can appear on gamma camera images either as cold spots or hot spots, depending on whether the abnormal tissue has a greater or lesser affinity for the radio pharmaceutical than the surrounding normal tissue. In some organs such as the lungs, liver and kidneys, lesions must alwyas appear as cold spots because pharmaceuticals 149 with greater affinity for the abnormal tissue of these organs have not yet been developed. For the brain, the heart and bone, such pharmaceuticals have been developed and are preferred for lesions detection because hot spots are imaged with better resolution than cold spots. The ability of the Fresnel zone plate system to detect small lesions may be superior to the ability of a system with a conventional collimator for some applications. Three possible applications have been identified. The first application is the detection of tumors in bone tissue. Patients with diagnosed malignant tumors in any internal organ are routinely given a bone scan to determine if the disease has spread to the skeletal system. The radiopharmaceutical, 99m -technecium pyro phosphate, is administered to the patient because it is absorbed faster by neoplastic bone tissue than normal bone tissue. The ratio of neoplastic to normal tissue activity is high, rougly an order of magnitude. Gamma camera images are then taken of the various parts of the skeletal structure. Tumors appear as intense hot spots against a weak background. Small tumors, detected early, may be successfully treated. The Fresnel zone plate system could offer the improved spatial resoltuion needed to increase the probability of detection of such small tumors. 150 In addition, if the germanium camera were used to scan the whole skeleton system, the reduced collection time of the zone plate would minimize the total time required per patient. The second application is the detection and location of myocardial (heart muscle) infarction. Once again, 99m technecium pyrophosphate is used because it tends to localize in the infarcted tissue. A major problem in locating the infarction results from the uptake of the same radiopharmaceutical by the sternum and rib cage overlying the heart. Fortunately, the uptake by the infarction is faster. Detection of a small infarction may be easier with a zone plate system not only because of its better spatial resolution, but also because of the tomographic ability of the system. If the infarcted tissue is separated from the overlying bone by several centimeters, such separation is well within the tomographic resolving power of the system. A third application involves the detection of small intracranial lesions. At present such lesions are very difficult to detect unless they lie close to the skull, for two reasons. First, images of deeper lesions are degraded because the source lies deeper in a scattering medium. Secondly, with a conventional, parallel-hole 151 collimator, radiation from a small source far from the collimator passes through many of the collimator holes. As a result, the size and location of the lesion is not well defined. This combination of the size effect and the collimator effect often makes detection of small, deep lesions impossible. A germanium detector would greatly reduce the scatter effect. A Fresnel zone plate aperture would eliminate this collimator effect and would identify the tomographic level of the lesion. Although the germanium camera is relatively small, several images could be taken to image the entire intra cranial region. In practice, however, the physician can often associate a patient's symptom, such as loss of feeling in the left arm, with a localized region of the brain, so that only a single image might be needed to con firm the presence of a lesion. 7.4 Conclusions Although the particular zone plate system design described in this dissertation appears inferior to a con ventional system in clinical applications requiring the imaging of whole organs, it may be superior for those clinical applications requiring the imaging of small lesions. 152 A nuclear medicine clinic often has several gamma cameras. For example, the clinic at The, Ohio State Uni versity Hospital has six scintillation cameras. There fore, it is not necessary for one camera to be useful for all applications. A germanium camera would complement, rather than replace, the scintillation camera. If the scintillation camera image of a large organ indicates the possible existence of a small lesion, the germanium camera could then obtain a high resolution image of the region of interest to confirm the existence of the lesion. The framing device defining the field of view of the zone plate system, discussed in Chapter 4, can be adjusted so that only activity from the small region of interest is imaged. If this region has a frontal area 2 as small as 12 cm , then a very high resolution image of the region would be obtained. Such an image would not only increase the probability of confirming the existence of a lesion, but might also provide a much greater ability to diagnose the nature of the lesion. CHAPTER 8 SUMMARY OF RESULTS AND CONCLUSIONS This study has examined the feasibility of using an on-axis Fresnel zone plate aperture and digital reconstruc tion with a high purity germanium gamma camera in nuclear medicine. In this chapter we summarize the research per formed and the results obtained. Possible extensions of the research are suggested and the major conclusions of the study are presented. 8.1 Summary of the Research The major areas of the research are listed here. 1. A theory was developed for determining the optimum position and parameters of the zone plate in terms of the camera characteristics. 2. The physical system required to implement the zone plate concept was designed. 3. A computer code was developed for reconstructing an image from a shadowgram, the coded data received by the camera. 4. An optical system was designed and constructed for forming shadowgrams with ordinary light in order to simulate 153 154 the forming of shadowgrams with gamma radiation and a germanium camera. A method of sampling the shadowgram data for digital processing was developed. 5. A computer code for generating shadowgrams was developed to obtain idealized data for the reconstruction code and to generate shadowgrams of sources difficult to simulate optically. 6 . Digital enhancement of Fresnel zone plate system images was investigated. 7. Clinical applications of the system were investi gated. 8.2 Summary of the Results The major results of the research are summarized here. 1. When the parameters of the Fresnel zone plate system are chosen optimally, the theoretical spatial resolution of the system exceeds the theoretical spatial resolution of the camera. The validity of the theory for optimally choosing system parameters was established by optical simulation of shadowgrams. 2. A shielding frame must define the field of view to maximize the signal-to-noise ratio. 3. Images found by digital reconstruction, when they are not lost in obscuring noise, correctly reproduce the size, shape and three-dimensional position of the source. 155 4. Digital simulation of shadowgrams established the kinds of source distributions which are imaged well by the system. Images of point sources are excellent. Small objects are imaged well with widespread, but low intensity noise. The images of larger objects are obscured by noise when the number of resolution elements in the object exceeds 10% of the total number in the field of view. Cold spots, although they can be imaged under special circumstances, are not imaged well in general. For sources of nonuniform intensity, the relative intensities are not well preserved in the reconstruction. Strong sources in the presence of a background source field are not imaged with good contrast unless the background field is very weak. The system has tomographic capability, but the ability to distinguish the tomographic levels of two sources depends on their size, shape, and the degree to which they overlap. 5. Image enhancement by digital processing is possi- » .ble. The results for small sources are excellent because much of the low.level background is removed. The technique . which removes noise for small sources docs not work for Inrgc sources. Few standard image enhancement techniques are applicable because the system is nonlinear and spatially variant. 156 6 . In clinical application the system cannot be used to image whole organs because of the limited frontal area of the germanium detector. The system has signifi cant potential application for imaging small lesions. 8.2 Intensions of the Research In the study of coded aperture imaging of gamma ray sources with a germanium camera and computer reconstruction, further research is possible in the following areas. 1. The clinical applicability of a system with a germanium detector of large frontal area may be suitable for future study when the cost of high purity germanium is reduced. 2. Research in the area of nonlinear, spatially variant optical systems could lead to an image enhancement technique useful for large sources. 3. The present study was aimed at maximizing the spatial resolution of gamma camera images by use of an on-axis zone plate. .A valuable future study would examine the clinical usefulness of an off-axis zone plate with a germanium camera, a system which would sacrifice spatial resolution in order to image larger sources. 4. Other coded apertures recently introduced, espe cially the nonredundant array and the time-varying array, 157 deserve future consideration if their initial apparent disadvantages are overcome. 8.4 Conclusion The use of an on-axis Fresnel zone plate and digital reconstruction with a germanium gamma camera is both possible and practical. Although conventional collimators are more suitable in the nuclear medicine clinic for imaging whole organs, the Fresnel zone plate system provides superior resolution in imaging small lesions. APPENDIX A OPTICAL THEORY OF FRESNEL ZONE PLATES AND HOLOGRAMS - AN OVERVIEW A.1 The Fresnel Zone Plate A.1.1 Introduction The zones of an advancing wavefront were first described by Fresnel, and the zone plate for spatially filtering the advancing wavefront was first constructed in 1871 (39). In 1875 Sorct published the first paper describing experimental work with zone plates. The literature in English refers to Fresnel zone plates; the literature in French refers to the same devices as Soret zone plates. The first extensive investigation of zone plates was carried out by R. W. Wood (40). Wood constructed zone plates by various methods and performed a series of experiments using visible.light. Following Wood's work, which began in 1898, zone plates became a standard topic in optics texts (4j_,42^) and were used in optics courses to demonstrate the diffractive properties of light. It appears that only one paper (43) was published on zone 158 1S9 plates in the 30-year period before 1950, and no practical applications for them were found* Fresnel conceived the theory of zones to aid in the understanding of wave diffraction. Diffraction effects are * classified as either occurring because of the interference of plane wavefronts from a source at infinity (Fraunhofer diffraction) or occurring because of the interference of spherical wavefronts from a source at a finite distance from the diffracting aperture (Fresnel diffraction). The latter kind of diffraction was originally described in terms of the somewhat obscure method of lunes and the Cornu spiral. The method of zones, introduced by Fresnel, is an easier approach to Fresnel diffraction theory. A.1.2 Basic Zone Plate Theory Let 0 be a point source of waves (such as light waves) of wavelength X; see Figure A.I. Any sphere about 0 is an instantaneous wavefront, all points of which are in phase. Let us determine the amplitude of the light at the point P under various circumstances. Let Mq be the intersection of the wavefront with OP. Let a » OMq and b « Choose the point Mj on the wave- front such that MjP = rt » b + j . (A-l) 160 Figure A.l Spherical wavefront expanding from point 0 . % 161 If the line M,P is rotated about the axis ht P then M, 1 o 1 describes a circle on the wavefront. The area inside the circle is the first Fresnel zone. Choose the points Mn so that (A-2) then the area between the n-th circle and the (n-l)-th circle is the n-th Fresnel zone. If a and b arc much greater than X, then these zones occur on a small enough portion of the wavefront so that the approximation R >> e will hold; see Figure A.2. Then a2 - (a - e)2 + R2 s a2 - 2ac + R2 (A-3) and (A-4) Let MP » r ° b + 6 . Then r2» R2 + (b + e )2 (A-5) s R2 + b2 + 2be Substituting the above expression for e, and rearrang ing, gives * Figure A.2 Wavefront geometry 163 R = 2ab 6 \fa + b (A-6) Since 6 ** n and defining abl a + b * (A-7) wc obtain Rn B K /n* (A-8) where Rn is the radius of the outer boundary of the n-th Fresnel zone. Hence = K. The area of the n-th zone is (A-9) - *Ri Hcncc all zones havo ,the same area. Consider now the light amplitude at P. The light passing through the first zone may be considered of posi tive phase in that every point within the zone has a phase within 180° of the phase of light at the point MQ . In the second zone the path length r differs from the 164 path length b by more than a half wavelength. The waves arriving at P from the first zone interfere destructively with the waves arriving at P from the second zone. Since the zones are of equal area, the cancellation is exact, provided the zones are small enough with respect to r so that the obliquity factor and the 1/r factor may be neglected. Let In be the light amplitude at P due to light passing through the n-th zone only, and consider an opaque screen placed between the points 0 and P at Mq and having a pinhole of radius R centered at MQ . If R B Rj the light amplitude at P is 1^. If R ** R2 then the amplitude at P is I » Ix - I2 ** 0 (A-10) If R a R3 then 1 ’ h ’ h * h = H ' h ( A - 1 1 J and so on. It can be easily shown that if the opaque screen is entirely removed, the amplitude at P is 1/2 1^, or half the value due to the first zone alone. Suppose, however, a zone plate is constructed so that the odd numbered zones are transparent and the even numbered zones are opaque. The first few zones of such 165 a plate are shown in Figure A.3. Then only waves of positive phase arrive at P; the resulting light amplitude at P is much greater than if no zone plate were present. The amplitude resulting from a plate with N zones (half of them transparent) is ^ 1^. The amplitude is not infinite for an infinite number of zones because the 1/r factor becomes significant for a zone of very large radius. In effect, the zone plate acts as a lens to focus the light from 0 onto the point P. Alternatively, the same effect can be achieved by making the odd zones opaque and the even zones transparent. Although the above theory applies only to an on-axis source at 0, an off-axis wave source at O' will give rise to.an off'axis image at P ' , as shown in Figure A.4. A small object such as a light bulb filament centered at 0 may be considered a large number of point sources. Hence a white screen in the plane at P will receive an image of the filament. Zone plates have multiple focal points. If the screen is moved closer to the zone plate, a position is found at which a second image occurs. At this position each zone of the plate allows three Fresnel zones of the wavefront to pass through to the new point P2 . Since two of these zones interfere destructively at P2 and only one contributes * 166 Figure A.3 A Fresnel zone plate I zone plate Figure A.4 Imaging an off-axis point 168 to the image, this image is less intense than the primary* one. By moving the screen increasingly closer to the plate, more images of decreasing intensity can be found. The primary image may be made more intense by a zone plate in which both odd and even zones are transparent but in which the even zones introduce an optical path length increase of one-half wavelength. The light from all zones interferes constructively at the image point. Such a plate is called a phase-reversal zone plate, or simply a phase zone plate. To find the primary focal length of a zone plate, consider the radius of the first zone; it is CA-12) Rearranging gives CA-13) If we define f ■ -j- , m e n the zone plate conforms to (A-14) the standard lens formula (A-IS) The other focal lengths are f/3, f/5, f/7, .... 169 Zone plates behave as lenses even to the point of producing virtual images as well as real images. They have not replaced lenses in optical instruments because 1. The focal length is inversely proportional to wavelength. The focal length for red is half the focal length for violet, and good images can be obtained only for monochromatic light. 2. The primary image is degraded by the presence of the defocusscd secondary images. 3. The primary image is degraded by the presence of the virtual images. 4. Accurate construction of zone plates with many zones is difficult. The ability of the plate to obtain high resolution depends on the number of zones in the plate. In 1951, 0. E. Myers (44) published a study of the resolving power of zone plates and considered their possible use for focussing infrared and X-radiation. He considered the effect of approximations used in zone plate theory and concluded that for a point object on-axis the maximum allowable number of transparent zones for good focus (the exact optical path length should not differ from the cal culated path length by more than 1/4 X) is given by 170 N - \/& - VsxtS-tt (a'16) This limitation becomes significant only when the object is very close to the zone plate. The maximum allowable angle a between the principle axis and the parallel rays of light from an off-axis point object at infinity for a plate of N zones is given by I ±/2fNX NX/f sin a - NX sin2 a /< 1/4X (A-17) The resolving power of a zone plate may be defined analo gous to that of a lens: two distant point objects arc resolvable if the central maximum of the diffraction pattern of one falls at or beyond the first minimum of the diffraction pattern of the other. For a zone plate of diameter d the minimum angular resolution 3 is the same as for a lens of aperture d, viz, 0 « I-.,22 , radians (A-18) This is also the angular resolution of image points with respect to the zone plate. The result is valid for large N. Experiments by Myers show the existence of an image at f/2 due in part to imporfections in plate construction. In 1967 Stigliani, et al. , (4S) extended Myers' work on the resolving power of a zone plate to consider the case of a small number of zones. The results for a positive zone plate (even zones being opaque) differ from the results for a negative zone plate (odd zones opaque). The resolv ing power of each converges to the resolving power of a lens as N increases. The results indicate that a nega tive plate has greater resolving power than a lens of the same aperture while a positive plate has less. How ever, the negative plate has a brighter ring of first order in the diffraction pattern of the image of a point source, while the positive plate has a less bright ring, when both are compared to the ring for a lens. Hence for closely clustered points there may be no real advantage to using a negative plate. Studies of the diffraction pattern produced by a zone plate were begun by Boivin in 1952. He extended the classical theory of diffraction by a circular slit to the case of several concentric ring-shaped apertures, and he considered Fresnel zone plate diffraction as an example (46). Boivin, et al., in 1956 published further theo retical and experimental work on zone plate diffraction patterns (47). In 1958, Dyson (48^) derived the light intensity distribution in the image formed by an optical 172 grating consisting of equally spaced concentric circles. Apart from a proposal by Murty (49) to manufacture a zone-sphere, the zone plate diffraction problem was not discussed in the literature for some time. In the latter half of the 1960's zone plate diffrac tion studies were renewed. An exact calculation of the diffraction pattern due to a single Fresnel zone was first carried out in 1965 (50_). Four years later the complete zone plate diffraction problem was solved for the case of an image of an on-axis point source. The solution to the problem of focal spot dimensions and energy distribution was derived by Childers and Stone (51) as a function of wavelength and number of zones. The diffraction pattern of off-axis images was determined by Bottema by summation of the patterns due to the first 10 zones of a zone plate (£2). A study of zone plate aberrations concluded that good images could be obtained only from highly monochromatic light but that the zone plate, like the pinhole aperture, produced images having no linear distortion, even for a very wide field of view (SjS). Useful approximations to the exact solution of the diffraction problem were worked out by Lehovec and Fedotowsky (54) in a study of image formation by a general ized zone-plate-sensor system. 173 A.1.3 Applications of Zone Plates The focussing ability of zone plates has suggested their use for imaging infrared, ultraviolet and soft x-ray sources (55). The use of zone plates for ultraviolet and x-ray telescopes mounted on satellites above the earth's atmosphere seems practical. Sets of zone plates have been used to determine the locus of points having equal optical path lengths through an optical system (56); and zone plate can be used for precision alignment of points along an axis, with accuracy of a few microns over a 12 meter distance (57). The ability of a zone plate to focus light of various wavelengths at spatially well-separated points has suggested the use of a zone plate in combination with a pinhole stop as a band-pass filter (58). A.1.4 Manufacturing of Zone Plates Manufacturing of a zone plate by precision ruling is difficult if hundreds of zones are to be produced. Other methods have been treated by Rogers (5_9) , Chau (60) , and New (61) . 174 A. 2 Holography A. 2.1 Basic Holography Holography is the recording of the interference pattern formed by the intersection of two wave trains. Consider two wave trains intersecting at a photographic film, as in Figure A. S. Figure A. 5 shows the .interfer ence fringes that are formed at the film; in Figure A.S the solid lines represent wave peaks, and the dotted lines wave troughs. Constructive interference occurs along the locus of points at which peaks intersect peaks and troughs intersect troughs. Destructive interference occurs where troughs intersect peaks. Figures A. 5(a) and (b) show that as the angle 0 between the two wave trains increases, the spacing between the fringes decreases. The wave tTain perpendicular to the film surface can be called the reference beam and the other train may be called the signal. If the reference beam were not present the signal would uniformly expose the film; with the reference beam present the film records information about the angle of the signal beam. The information recordod cannot, however, distinguish between the signal of Figure A.5(c) and that of Figure A.5(d): the two signals produce the same interference pattern. ( C ) ( d ) Figure A.5 Reference beam and signal arriving at the film, (a) and (b) show signals at different angles; (c) and Cd) show two signals at the same angle but different directions. 176 The above discussion assumes that the two beams arc ‘ of the same wavelength and come from a single source of waves; otherwise each wave train would be a superposition of randomly emitted wave trains from many small sources. Any single source of waves is made up of many atomic sources of waves and these must act in unison (i.e. coherently) to produce usable wave trains. If the film is developed and the resulting trans parency is illuminated with a train of coherent waves, the interference pattern acts as a diffraction grating: part of the incident wave is undiffractcd and travels straight ahead, forming the zero-th order beam; two first- order beams are diffracted at the angle 0 as shown in Figure A.6. In effect, the transparency reproduces the two beams of Figure A.5(c) used to expose the film and also another beam corresponding to the possible situation of Figure A.5(d). A true diffraction grating with sharply defined edges would give rise to higher order diffraction beams. However, because of the sinusoidal intensity variation of the interference fringes, only the zero-th and first order beams are formed. Notice that it makes no difference whether a positive or a negative transparency of the film is used. 177 first order beam incident beam O-th order reference beam first order beam (signal)transparency Figure A.6 The diffracted beams 178 Consider now a point scatterer S of waves at d as shown in Figure A.7 (a). At the film the unscattered plane waves act as a reference beam and record information about the signal which is the spherical waves scattered by S. Interference fringes are formed on the film as in Figure A.7(b). The fringe spacing decreases moving away from the center of the pattern as the angle 6 between the two interfering wavefronts increases. It is easy to show that the radius of the n-th fringe is proportional to /n. When the transparency in this case is exposed to coherent waves, the zero-th order beam goes straight through as before. The first order beam corresponding to the original signal leaves the transparency at such an angle that it appears to come from a point at a dis tance d behind the transparency. The point is the point where the scatterer was when the film was exposed. Hence a virtual image of the scatterer is seen through the transparency. The other first-order beam is diffracted to a point S2 in front of- the transparency. Since all fringes on the transparency diffract waves to this point, the waves converging at S2 form a real image of the original scattering object. See Figure A.8. A small object may be considered as a collection of point scatters. The interference pattern formed on the 179 (a) (b) Figure A.7 Hologram of a point scatterer. (a) waves interfering at the film plate; (b) the holographic pattern. 180 first order (signal) 0-th order ^ (reference) Figure A.8 Reconstructing a point image. 181 film may be an elaborate superposition of many such over- ‘ lapping circular fringes. In fact, for a complex object, even microscopic observation of the patterns of intensity variation in the transparency may reveal no clearly observ able fringe pattern. Such a transparency is called a hologram. When illuminated with coherent light the holo gram forms a real and a virtual image of the original object. The images formed by holograms are three-dimensional since the position of any point in the image varies according to the position of the original object with respect to the film. If the scatterer S is far from the recording film, the virtual image of S is formed far behind the illuminated hologram, and vice versa. Notice that any small segment of the hologram (so long as the segment is much larger than the fringe spacing) can produce the real and virtual image of S. But the image produced by a segment of the hologram has less intensity than that produced by the whole hologram. A.2.2 A History of Holography Dennis Gabor invented holography in 1947 in an attempt to remove the spherical aberration of the images formed by an electron microscope. Tho interference pattern between a plane electron beam and a scattered electron beam is recorded, but the scattered beam is not spherically 182 symmetric due to the aberration introduced by electron lenses. A transparence of the pattern is illuminated with coherent light and an image of the original object is formed. During the image formation an optical lens having a spherical aberration opposite to that of the electron lens produces an undistorted image of the original object. Holography was only of limited success in improving the resolution capabilities of the electron microscope. The main problem with the method was the deterioration of image quality caused by the out-of-focus real image being superimposed on the virtual image. During the early 1950's, Baez attempted to use the holographic method to form images with soft x-rays (62), and several other workers tried to separate the two overlapping images (63). None of the early efforts to remove the extraneous image were very successful. The effort most worthy of mention was a spatial filtering technique which blocked out the unwanted image but serious degraded the quality of the remaining image (64). One of the first successful applications of the holo graphic principle was the classified work on "side-looking radar" carried out at the University of Michigan beginning in 1956 (65). As an airplane flies past an object, the 183 interference pattern between a radar beam reflected from the object and a local oscillator is recorded. The pattern of maxima and minima recorded is a one-dimensional holo gram of the object: the pattern may be duplicated on a photographic transparency and a good image of the object is obtained when the transparency is illuminated with coherent light from a mercury arc lamp. Hence up until 1960 all the work on holography consisted of either (a) forming an interference pattern with electrons or X-rays and reconstructing the image with visible light having a much larger wavelength; or • (b) forming an inter ference pattern with radar and reconstructing with visible light having a much shorter wavelength. One of the members of the group which developed this technique was Emmett Leith who, along with Juris Upatnieks, established holography as a major field of purely optical research. Between 1961 and 1966, Leith and Upatnieks at the University of Michigan overcame all the previous stumbling blocks of holography. In 1961 they considered holography from a communication theory viewpoint (66): the photographic film is a square-law detection device receiving a modulated carrier frequency which has under gone a frequency dispersion; during the reconstruction, an inverse dispersion operation is carried out in which the carrier frequency (zero-th order beam) is separated from . the upper and lower side bands (the real and virtual images). Using the communication theory viewpoint, they determined the method of successfully separating the two images, as shown in Figure A.9. The reference beam is made to meet the beam diffracted by the object at an angle 0 during the hologram formation. During the image recon struction the two images will be separated by an angle 20. In 1962, Leith and Upatnieks introduced the laser as the source of coherent light for holography. Up to this time holograms had been made only of small scatterers mounted on a transparent background; in 1963 Leith and Upatnieks formed holograms of transparent objects (e.g. letters of the alphabet) on an opaque background, and also developed the ability to form holograms of continuous - tone objects, such as an ordinary black and white photo graphic transparency (67). In 1964, "diffused" holograms were first made by the same researchers; instead of simply allowing light to be scattered by semi-transparent objects, they formed holograms by recording the inter ference pattern between the reference beam and the light diffusely reflected from opaque objects, as shown in Figure A.10. 185 incident beam prism reference ^ beam signal film transparent object virtual image incident beam 0-th order beam hologram real image (b) Figure A.9 Hologram of a transparent object. (a) Making the hologram; (b) reconstruction. 186 opaque incident beam A objects /// signal mirror reference beam film Figure A.10 Hologram of an opaque object. Reconstruction is the same as in Figure A.9 (b). 187 After 1965, progress in laser technology and holo graphic image quality was so rapid and the applications of holography so many that it is not possible to cover them concisely. In 1966 an excellent summary of the differnt basic holographic techniques along with their advantages and disadvantages for various applications was given by G. L. Rogers (68). A. 3 The Relationship of Fresnel Zone Plates to Holograms Shortly after Gabor's original work on "the theory of reconstructed wavefronts," G. L. Rogers noted that the interference pattern formed by the waves from a point scatterer and a coaxial reference beam resembled the pattern of opaque and transparent regions of a Fresnel zone plate (69). If p^ is the radius of the first dark ring of the pattern then /n p^ is the radius of the n-th ring. The rings are alternately light and dark. Since any object may be considered as a large number of small light scatterers, a hologram may be considered a large number of overlapping zone plates. Conversely, the zone plate may be considered a special kind of hologram. The three-dimensional character of holograms and the existence of the two conjugate images is easily understood once the hologram is conceived of as a multiple-zone-plate (70). 188 When coherent light falls on the hologram, each zone plate forms one real and one virtual image point: the super position of all the virtual image points makes up the virtual image of the hologram, and similarly the real holographic image is formed by superposition of the many real image points. The analogy between Fresnel zone plates and holograms is far from exact, however, because ideal zone plates consist of perfectly opaque and perfectly transparent regions with sharp edges, resulting in an infinite number of higher order images, both real and virtual, as wc have seen. The hologram of a point source consists ideally of a sinusoidal variation of intensity from one fringe to another; this results in only first order images. This ideal hologram of a point source has been called a Gabor zone plate by Horman and Chau (71). An actual hologram of a point source does not produce an exact sinusoidal film exposure because the transmittivity of exposed film does not vary linearly with the intensity of the light exposing the film. The actual hologram of a point source has been termed a binary zone plate; when illuminated with coheront light it also gives rise to multiple images. The conceptual understanding of holograms as multiple zone-plates has made possible the theoretical calculation of the intensity variation patterns of the holograms of 189 some simple three-dimensional, geometrical objects. When' these patterns are plotted, photoreduced and illuminated with laser light, they give rise to the three-dimensional image of the intended object. The patterns have been plotted by computer point-by-point (72) (each point being fully opaque or fully transparent) and have also been created by sequential printing of overlapping zone plate patterns on a photographic plate (73). These studies were begun in the late 1960's. The similarities and differences in the mathematics used to describe Fresnel and Gabor zone plates has been the subject of some discussion (7£,75), and the discussion has led to a new way of describing the efficiency of the Fresnel zone plate for imaging a point source (76). A good summary article on the relation of zone plates and holograms is given by Ferrier (77). APPENDIX B THEORY OF ZONE PLATE IMAGING OF GAMMA RAYS The theory of zone plate imaging of gamma rays has been developed primarily by Barrett (78), and is summar ized here. Consider a source distribution given by d(r), where r is the position vector in the source plane. The Fresnel zone plate is located at a distance z » s^ from the source: coordinates in this plane are denoted by r'. Let the transparency of the zone plate be denoted by g(ir'). The detector plane is located at z « s^ + S2 » and its coordinates are denoted by r". •+■ 2 -f Let h(r") d r" be the average number of gamma rays 2+ falling on a resolution element d rM in the detector plane during time t . By a straightforward but tedious application of geometric shadowcasting principles-, Barrett * shows that h(rM) is given by h(?") - f (sl * s2 ' *)'ZS(r’) Using Fourier transform techniques it is easily shown that the source distribtuion is 190 191 d(r) » ^ exp[-2ni(ig.r)ld2l (B-2) c G(-f/b) where a » Sj/CSj + s2) (B-3) c = ( t / 4 it)/(S j + s2)2 (B-5) and H and G are the Fourier transforms of h and g respec tively, and ? is the two-dimensional spatial frequency. Equation B-2 shows that the source distribtuion can be recovered from a knowledge of h and g, provided that the Fourier transform of g is nonzero for all spatial frequen’ cies contained in d(r). For an on-axis Fresnel zone plate, g(r) is given by 1 , if sinOrr2/!^2) > 0 g(r) CB-6) 0 , if sin(Trr2/Rj2) < 0 where r * |r|, and Rj is the radius of the first Fresnel zone. The Fourier transform of g is given by 192 G(f) - I «(?) + Y, ( - T - ) ld2r e*p[2»i pCodd)'2* P / J (?-r + j pr2/ ^ 2)] (B-7) where 6 is the delta function and p is a dummy summation index. Computer calculation of G(t) shows that its absolute value is reasonably constant and nonzero for all 2 frequencies up to a cutoff frequency given by fc B » where is the radius of the zone plate having N zones. Beyond the cutoff frequency |G(?)| falls quickly to zero. Maximum contrast occurs in the image when the scaled spatial frequency of the source matches the cutoff fre quency of the zone plate, that is, f - — f . (B-8) source Sj c v * Barrett also considered the effect of quantum noise in Fresnel zone plate imaging (79). The only noise source considered was the statistical variation of the source radiation emission. The signal-to-noise ratio (SNR) in the reconstructed image was found to depend on the size and shape of the source. A single point source will produce a single zone plate shadow. A randomly sampled zone plate will, on the average, 193 focus as well as a full zone plate. However, noise repre sents deviations from average behavior. The ratio of the mean amplitude of the reconstruction wave at the image point to its standard deviation is given by where is the number of gamma rays detected coming from the point source. For a collection of M point sources, the emissions from each source are statistically independent. The mean and variance at each image point arc then additive. Let Nm denote the number of gamma rays detected from the m-th source and denote the total number from all M sources. Then the ratio of mean amplitude to standard deviation at the corresponding m-th reconstruction point is (B-10) if all point sources are of equal intensity and all cast full zone plate shadows on the detector, then this result reduces to 194 (B-11) where, on the average, Ramma rays are detected from each point source. In experimental work the observed quantity is not amplitude but intensity. The intensity SNR is given in each case above by . < IAI > Z/(AA)2 For the case of M equal point sources, this reduces to (B-13) For purposes of comparison, the SNR for a pinhole aperture with any source distribution is ^ N y ^ 1^2- For a single point source, as an example, if the open area of a zone plate is 500 times that for a pinhole, then for equal exposure times, the zone plate SNR will be about 10 times better than that for the pinhole. On the other hand, for a fixed SNR in both cases, the pinhole requires an exposure time a factor of 100 greater than that for the zone plate. However, for a source made up of M resolution ele ments, while the pinhole SNR remains constant, the zone -1/2 J plate SNR is reduced by a factor of M ' . These results apply to both on-axis and off-axis zone plates. A more extensive study of the effect of quantum noise in off-axis zone plate systems has been reported by Joy and Houle (80),(81). APPENDIX C THE GERMANIUM GAMMA CAMERA The imaging system considered here employs a high purity germanium gamma camera. A camera as large as the one described in this appendix has not been constructed at the present time, but several developmental efforts are underway. The camera design used in this study is based on developmental research in our laboratories in the Nuclear Engineering Department at The Ohio State University. The camera consists of an array of orthogo nal strip detectors with associated electronics and mechanical equipment. A detailed description of this design can be found elsewhere (82); an overview of the design with emphasis on those features most relevant for Fresnel zone plate imaging is presented here. C.l The Orthogonal Strip Detector The detection system is made up of an array of orthogonal strip detectors, each of which is cut from a single crystal of high purity germanium. The detector is fabricated (85) from p-type germanium by depositing an 196 197 n-type lithium contact on one face and a p-type palladium * contact on the opposite face of a rectangular planar crystal. Electrically isolated electrode strips are formed by cutting grooves into each face of the crystal. By cutting the strips on one face orthogonal to those on the opposite face, as shown in Figure C.l, an array of isolated detector elements is formed. Each strip is connected to a resistor network having a charge sensitive preamplifier at each end. This resistor network elec tronically identifies the detector location of the burst of charge released by a gamma ray interaction: a charge pulse emerges from the electrode strip on each face nearest to the interaction site and divides into the resistor network according to its location along that resistor network. The amount of charge, or Q2» arriving at either end of the upper network decreases proportion ately with an increase in the distance of the interaction from that end of the network. Similarly, Q3 and fix the position of the interaction in the orthogonal direction. The energy of the incident gamma ray is proportional to the total charge released, + Q2* The spatial resolution attainable by this detector is limited by the minimum center-to-center spacing of the electrode strips. Because of the practical problems of accurately mounting the detector strips onto metal contacts 198 CHARGE* SPLITTING RESISTOR NETWORK ELECTRODE STRIPS (n-TYPE) ^ p-TYPE r HIGH PURITY GERMANIUM ELECTRODE STRIPS fp-TYPE) Figure C.l Schematic of the orthogonal strip, charge splitting detector. 199 leading to the resistor networks, the minimum strip spacing achievable at present is considered to be about 1 mm (84). The design considered here has a strip spacing of 1.2 mm. C.2 Detector Array In addition to the constraint imposed by the minimum strip spacing, additional constraints are imposed by the limited size of available germanium crystals and by the cost of these crystals. The largest square detector that can be cut from available high purity germanium crystals is 3.78 cm on a side. Because the surface area of such a detector is so small, an array of these detectors is needed to achieve the capability of imaging the organs of interest in nuclear medicine. However, the present high cost of ultrapure germanium ($10-$18 per gram) limits the number of detectors in the array. The camera is designed to have 16 detectors in a four-by-four square array, as shown in Figure C.2. Each detector has 32 strips on each face, and hence the complete array contains 128 by 128 resolution elements. It is important that the center-to-center spacing of neighboring strips on adjacent detectors be the same as that spacing for neighboring strips on an individual detector. If it is not the same and a conventional collimator is used, then bright or dark lines will appear on the image 200 >. /— IU-I/Z10-1/2" DDIA. I* (USABLE) \ ' \ 12fl strips l.490H (3.70 cm) \ 15.3 cm . * . AREA" 36 In2 ( 2 3 2 cm2 ) Figure C.2 Illustration of the 4 x 4 detector array. The array is shown inside of the usable area (dashed circle) of a scintillation camera. 201 corresponding to the detector borders; but when a Fresnel' zone plate aperture is used, the result will be a decrease in the resolution for all points in the image. The dashed circle of Figure C.2 indicates the usable area of the standard scintillation camera, 86 square inches; the germanium camera area is 36 square inches. When used with a conventional collimator, the germanium camera can be used to image almost the entire vertex view of the brain and somewhat greater than half of the posterior, anterior, or lateral views; about half of one lung and one-third of the liver can also be imaged. Each detector in the array has a thickness of 2 cm. This thickness represents a compromise between high detector efficiency and minimum germanium cost. For 140 keV photons the detector efficiency is 86%, whereas the standard scintillation camera efficiency is about 95%. However, for gamma ray energies above 500 keV this ger manium camera is more efficient than the scintillation camera. C.3 Electronics The array of 16 high purity germanium detectors is divided into four electronically separate quadrants, each containing four detectors. The four detectors in each 202 quadrant are interconnected electronically forming, in effect, a single detector having a single charge splitting resistor network for each of the two sets of orthogonal strips. A complete set of filtering electronics is provided for each quadrant of the detector. The charge pulses from each quadrant, Q^, (J2, and as shown in Figure C.l, are received by the instrumentation diagrammed in Figure C.3. Each charge pulse is integrated by a separate preamplifier resulting in proportional voltage outputs V V2, V3 and V^, respectively. A summing Gaussian filter adds and V 2 , and puts out a Gaussian shaped pulse having a peak value proportional to the energy of the incident gamma ray. If the energy is outside the preset energy window, the system is cleared and reset to receive another gamma event. If the energy is inside the window, the summed pulse is stored in the peak detector for future processing. A summation Gaussian trapezoidal filter puts out a pulse having a peak value proportional to - V2, and hence proportional to the x-position of the gamma event. Another such filter finds Vj - and defines the y-position. After the peak value of each pulse is determined, the control circuit is designed to detect and reject HIGH VOLTAGE ANTISYMMETRIC SUMMATION GAUSSIAN X - CHANNEL PREAMPLIFIER TRAPEZOIDAL DRIVER FILTER PEAK DETECTOR PREAMPLIFIER SUMMING ENERGY CHANNEL GAUSSIAN FILTER DRIVER ENERGY DISCRIMINATOR PREAMPLIFIER ANTISYMMETRIC SUMMATION GAUSSIAN Y - CHANNEL TRAPEZOIDAL DRIVER FILTER CONTROL PILE-UP AND TO PROCESSOR PREAMPLIFIER OVERLAP REJECTION RESET T r OM PROCESSOR Figure C.3 Schematic diagram of the amplifier, noise filtering and control circuitry required to process signals Q., Q-, Q3 and Q* emerging from one quadrant 203 of the detector array. 204 two or more gamma events occurring so close in time that their individual energy and positional information is not accurate. The complete camera has a count rate capability of 100,000 counts per second. The time constants selected for the filtering systems determine the energy and spatial resolutions of the system. For time constants of interest, of the order of a microsecond, energy resolution is degraded as spatial resolution is improved. The goal for system resolutions, full width at half maximum, for 140 keV gamma rays is 2.4 mm spatial• resolution and 3.8 keV energy resolution (8j>). C.4 Mechanical Components The mechanical components of the camera consist of a camera assembly, a cryogenic system and a vacuum system. The camera assembly consists of a camera stand supporting the camera head and adjustable counterweights used to balance aperture systems of different weights. Figure C.4 shows this camera assembly with no aperture attached. The camera head, shown in cutaway view in Figure C.5, contains the detector array, preamplifier assembly, a cooling unit and a vacuum chamber maintained by a vacuum pump. The lower vacuum chamber plate is a thin aluminum window directly below the detector array. CAMERA HEAD R01ATI0N BRAKE ADJUSTABLE COUNTER-WEIGHT TWO-AXIS GIMBAL VERTICAL BRAKE 205 Figure C.4 The camera assembly consisting of camera stand, camera head,and adjustable counter-weight. 206 ROUOHINO PORT CLOSED CYCLE 30 L/S REFRIQERATION UNIT ION PUMP JEt PREAMPLIFIER MOUNTIN'! FLANGES SECTION FOR CAMERA STAND UPPER VACUUM 'CHAMSER PLATE k w A W W S W A 1 ALUMINUM NINDOW Figure C.5 Cutaway view of the camera head showing the the detector array, the preamplifier section, the cooling devices and vacuum equipment. 207 The cryogenic system maintains the detector array, the resistor networks and the preamplifier FET's at a temperature of approximately 130°K to minimize electronic noise. This refrigeration system is a closed cycle system containing an air expander housed in the camera head connected by flexible tubing to a remotely located compressor, not shown in the figure. The vacuum system helps to maintain the cold tempera ture and also prevents contamination of the high purity germanium detectors. A roughing pump is used to pump down to 10’^ Torr, and a 30 liter per second ion pump maintains the chamber pressure at 10’** Torr or less. APPENDIX D THEORY OF DIGITAL RECONSTRUCTION OF SHADOWGRAMS Gamma rays from a distributed source of Tadintion passing through a Fresnel zone plate aperture cast a shadow pattern on the detector which is a hologram of the source distribution. The resulting shadowcast hologram, like the interference hologram, contains coded information about the source distribution. The image of the source can be reconstructed by passing a coherent beam of laser light through a transparency of the hologram. The image is the resulting diffraction pattern. The difficulty with this optical method has been the obscuring of the diffraction pattern by the undiffracted beam of light. A solution to this difficulty is to calculate the diffraction pattern digitally. The theory of digital reconstruction is derived here. D.l The Diffraction Integral Suppose coherent radiation passes through a shadow gram of area S surrounded by an opaque screen, as shown 208 209 in FigUTe D.l. In the following development a vector valued function will be denoted by an arrow. Let the electric field vector at any point P in the electromagnetic field be given by U(P) = U0(?)exp[-i where the real valued function UQ (P) is the wave amplitude, u is a constant unit vector, and <|>(P) is the phase. Let UCPj) be the radiation distribution in the plane of the shadowgram. The radiation amplitude and phase at a point P is given by (86J) 1 UCPJ IT j j 9 ( ? l ) cos(n,rQ1)dS (D-2) where X is the radiation wavelength, i is the unit imagi nary number, n is the unit normal to the surface of the volume in which the radiation propagates, and and CD-3) r01 ’ l?0ll 210 shadowgram opaque screen Figure D.l Coherent radiation incident on a shadowgram. i 211 Several approximations can be made to change Equation D-2 to a more readily usable form. Figure D.2 illustrates the reconstruction geometry for an illumi nated shadowgram in the x ^ t y^^ plane forming an image in the X q , yg plane. Let the distance between these two planes* be z. Let a^ be the maximum distance of a point in the shadowgram from the origin in the x^, yj plane, and let a^ be the maximum distance from the origin in the Xq , y^ plane at which the image intensity is to be calculated. If rfll >> and rQ1 >> a^, then the following approximations are valid; cos (n, rQ1) = 1 (D-4) r01 “ 2 Assume the shadowgram is illuminated by a plane wave. Then all points in the shadowgram plane have the same phase, taken arbitrarily to be zero, and hence u(px) « ■ V xi*yi5" * (D_5) Equation D-2 becomes “ (x0 ,y0) " Uz JJ Uo(xX*yl) exp[2irir01/X]dx1dy1 . (D-6) Figure D.2 The reconstruction geometry 212 213 Notice that in the phase term cannot be replaced by z because it is divided by a small number, X. Hence small variations in r ^ produce large phase changes. D.2 The Fresnel Approximation The Fresnel approximation is often used to simplify diffraction calculations. It consists of replacing the spherical wavefront expanding from every point toward Pq by a parabolic wavefront whose axis is the z axis. The approximation is valid if Pj and Pq are close to the origins of their respective planes and if z is suffi ciently large. The value rni can be expressed as Using the approximation that for e << 1, (1 + e)1/2 = 1 + j e - g-e2+ ... we obtain Substituting this expression into Equation D-6 yields ®cx0,y0> - / / V W * p | & [(Xi - x0)2 + (y1 - y0)2]{ dxldyl (D'7) This approximation is valid provided the next term in the expansion has a value much less than one radian, that is, ^l[(^)4*(^)4]«d c d -.j Substituting a^ for x^ and y^, and a^ for Xq and yg, and rearranging, we obtain ^ (flj * a0)4 « zJ CD-9) Assuming a factor of 5 means "much greater than," a minimum value for z is obtained for which the Fresnel approximation is valid. 215 To check whether Equation D-7 is valid for a typical holographic reconstruction, the. following typical values are substituted into Equation D-10. For a helium-neon laser, X = 0.6328 microns; and typicall a^ and aQ arc a centimeter; hence zm^n is about 17 cm. Since z is typically 100 cm, the Fresnel approximation is valid for holographic image reconstruction. The shadowgram is a superposition of Fresnel zone plate patterns. In order to focus a plane wave at a distance z the focal length of the zone plate pattern (sec Equation 3-3) must be equal to z, that is, Z “ -j±-Ri (D-ll) where R^ is the radius of the central Fresnel zone. When an on-axis zone plate pattern is used, the shadow pattern must be smaller than the dimensions of the shadowgram transparency. The smallest practical zone plate is one with four zones, that is two transparent zones. For this pattern to exist .within the dimensions of the hologram, the condition 2Rj < ax (D-12) must hold. This condition places an upper bound on z, namely, 216 CD-13) zmax * VT Because in practice a^ and ag are approximately equal, let us replace both by the symbol a. Combining Equations D-10 and D-13, we obtain 2.52 a4/3 X_1/3 CD-14) This expression defines the limits on z values for which the Fresnel approximation is useful for the calculation of the image. D.3 The Fraunhofer Approximation The Fraunhofer approximation is often used to simplify diffraction integrals, but it cannot be used for the digital reconstruction of shadowgrams. The Fraunhofer assumption is << z CD-15) or equivalently CD-16) 217 Because Equations D-13 and D-16 cannot both be satisfied simultaneously, the Fraunhofer approximation cannot be used for digital reconstruction of shadowgrams. D.4 The Reconstructed Image as a Fourier Transform Equation D-7 is a special case of a two-dimensional Fourier transform. To see this, introduce a new unit of length, , defined by (D-17) Substituting this expression into Equation D-7 makes it independent of X and z and dependent only on their product. The plane of reconstruction is now specified by the choice of Rj. The equation can be simplified by introducing the following dimensionlcss variables. The dimensionless image coordinates are and (D-18) The dimensionless shadowgram coordinates are (D-19) 218 Then Equation D-7 becomes 8(x,y) = “°»Pj2»*?-/.U . / / u0 (£x ,« ).«p j «i [ffx - x)2 s ( + (fy ' y)2] I dfxd£y • (D'20) The intensity, I, of optical radiation detected (by the eye or by film) at a point in the image plane varies as the square of the amplitude. That is, I(x,y) - U(x,y)fi*Cx,y) (D-21) where the asterisk denotes the complex conjugate. There fore, the phase factor in front of the integral does not contribute to the image intensity, and it will be ignored. Expanding the exponential terms of Equation D-20 gives fi(x,y) » u f f U0 (fx »fy)exp|ni(fx2 ♦ fy2 ) Jexp £ iri S (x2 + y2)J exp [-2ni(fxx + fyy) J d f xdfy (D-22) 219 The phase factor depending only on x and y can be brought * outside the integral; hence it will not contribute to the intensity and can be ignored. Define a new complex amplitude in the shadowgram plane to be X(fx.fy) - U0Cfx,fy)exp[iil(fx2 + fy2 )]“ (D‘2S) Then Equation D-22 becomes S(x.y) - / J X(fx ,fy)exp[-2»i(fxx ♦ fyy)] dfxdfy . CD-24) s Equation D-24 shows that the complex amplitude in the image plane is the two-dimensional Fourier transform of the complex amplitude in the shadowgram plane. With this in mind, the shadowgram may be thought of as a spatial frequency plane (f„»fv); and the image, which is actually x y given by the square of the amplitude, may be considered a Fourier transform of the shadowgram. D.5 Application of Sampling Theory Sampling theory is a familiar part of communications theory concerned with transforming digitized information back and forth between the frequency and time domains 220 with minimal loss of information and minimal introduction of noise. The fundamental theorem of sampling theory is that a continuous waveform having no frequency content higher than F can be completely recovered from the inter polation of data sampled at a frequency greater than or equal to 2F. Consider the spatial frequency content of a shadowgram recorded by a Fresnel zone plate on an orthogonal strip gamma ray camera. The output of the camera is digital: the number of counts per detector resoltuion element is determined. The detector is a square of side with spatial resolution AXj and contains Nj by Nj resolution elements. The shadowgram can contain no spatial frequency higher than one cycle per pair of resolution elements: then alternate resolution elements are irradiated and the intervening ones are not. Then the highest frequency resolvable in the shadowgram is F - ■ (D-25> In dimensionless units, the dimensionless spatial resolu tion element in the shadowgram plane is Ax Af B h . (D-26) K1 221 Now F is given by F * • (D‘27) We see that the sampling rate, or a number of samples per unit distance, is 1/CAfRj); and the sampling rate is double the highest frequency, F. Consider next the spatial frequency content of the image reconstructed from this shadowgram. According to sampling theory, if samples of a frequency spectrum are taken at intervals Af, up to a maximum frequency F » NjAf, then the Fourier transform of the frequency spectrum gives the waveform in the time domain with time resolution At “ 1/F for a total time T ■ NAt. Let us apply this theorem to the case of the orthogonal strip detector. Equation D-24 gives the spatial distribution of the image amplitude. It follows that the quantities f and f in that equation must represent the spatial x y frequency content of the image in orthogonal directions. Hence the quantity Af of F.quation D-26, which is the dimensionless spatial resolution in the shadowgram plane, is numerically equal to the dimensionless spatial frequency resolution in the image plane. Therefore, by the sampling theorem stated above,the dimensionless spatial resolution in the image plane is or, in terms of dimensioned parameters, (D-29) The dimensioned spatial resolution of the image is given by AXq * RjAx (D-30) By substituting from Equation D-29 and using the relation D1 * N xAxi» we obtain CD-31) Ax0 " TJJ- • This equation shows that the size of the image resolution element depends only on the shadowgram size. We show next that the size of the image depends only on the size of the shadowgram resolution element. The size of the image is CD-32) Substituting Equations D-29 and D-30 gives 223 (D-33) The ratio of image size to shadowgram size is 2 (D-34) D.6 Digitizing the Reconstruction Integral The integration in Equation D-24 must be replaced by summation for digital reconstruction. For purposes of numerical integration it is convenient to begin with the integrand in the form given in Equation D-20. In this integrand we see that the phase difference for coherent radiation from the origin of the shadowgram plane to the origin in the image plane is zero. It is necessary to preserve this relation to avoid distortion of the image in digital reconstruction, and so the values of f , f , x y x, and y should be zero at the ori'gin. Because computer FORTRAN does not allow negative subscripts in an array, which are normally used to denote points left of center, an offset parameter s is introduced, defined to be the number (not necessarily an integer) of resolution elements between the center of the shadowgram and the center of the outermost resolution element, that is, 224 N, + 1 s » -±~2---- . (D-35) It is assumed that the resolution in the shadowgram plane, Af, is the same in the f and f directions and x y that the resolution in the image plane, Ax, is the same in the x and y directions. In order to digitize Equation D-20, the following substitutions are used: f ■ (m - s)Af ; m e 1, 2, ..., Nj fy - (n - s)Af ; n » 1, 2, ..., Nx (D-36) x ■ (j s)Ax i j a l, 2, ..., y » (k - s)Ax ; k » 1, 2, ..., Nx Ignoring phase factors outside the integral, Equation D-20 is written as a double summation as follows * 0 .1 0 - 5 : i uU0 (m,n)exp(i (D-38) 225 When this expression is expanded, terns not containing m or n nay be factored outside the double summation. Phase factors outside the double sum may be ignored since they will not contribute to the radiation intensity; then the remaining terms are * - "Tcm2 + n 2) (Af)2 ♦ 2s(m + n) (ji- - CAf)2)J (jm + kn) CD-39) Define the digitized complex shadowgram amplitude to be A(m,n) » uUQ(m,n)cxp iir[Cm2 ♦ n2) (Af)2 + 2s(m ♦ n) - (Af)2 )] j . CD-40) Then the digitized complex image amplitude is given by Ni Ni U(j.k) = Y , Y . ^(m,n)exp I" - O m ♦ kn) 1 . (D-41) mal n*»l L 1 J 226 The digitized complex image amplitude is a two-dimensional* discrete Fourier transform (DFT) of the digitized complex shadowgram amplitude. The intensity of a point in the image plane is given by I(jJO ■ U(jtk)u*(j,k) (D-42) Hence the Nj by array of shadowgram intensity values, U(m,n), is transformed into the by array of image intensity values, I(j,k). D.7 Reconstructing the True Image The model described above relies on the focussing ability of shadowgrams to form the real, conjugate image on a screen in the image plane. In addition to this conju gate image, the virtual, true image also exists in the diffraction pattern but requires a converging lens to focus it to a real image on a screen. Assume the shadowgram illuninated with a plane coherent beam of radiation forms the conjugate image at a distance, z. It is easily shown by geometric optics that placing a converging lens of focal length z/2 immediately after the shadowgram results in forming the true image at the same distance, z. 227 The effect of this lens, essentially in the shadow gram plane, is to modify the complex shadowgram amplitude with the phase factor Cxx * 8*ven by (®Z.) UfcCxi.Xi) = exp[- t ^ ( x j 2 ♦ yi2 )] (D-43) When substitutions are made using Equations D-17, D-19 and D-36, and when this factor is included in Equation D-41, the net result is merely to change the sign of the first term in the exponential function of Equation D-40. That is, the complex shadowgram amplitude becomes A(m,n) ** uU(m,n)exp -iirl(m^ + n^)(Af)^ (D-44) Equations D-41 and D-42 are unchanged. Different tomographic planes in the image can be brought into focus either by changing the position of the screen or by changing the focal length of the lens. Because we have related the lens' focal length to the image position, 228 these two alternatives are mathematically equivalent. The focal plane of the image is controlled by the choice of the parameter Af in Equation D-44. BIBLIOGRAPHY 1. N. F. Moody, W. Paul, and M. L. G. Joy, "A Survey of Medical Gamma Ray Cameras," Proc. IEEE, 58, 2, Feb. 1970, p. 217. 2. J. F. Dctko, "A Prototype, Ultrapure Germanium, Orthogonal Strip Gamma Camera," Proc. IAEA Symp. on Radioisotope Scintigraphy, IAEA/SM-164/135 , Monte Carlo, October 1972, 3. P. A. Schlosser, D. W. Miller, M. S. Gerber, R. F. Redmond, J. W. Harpster, W. J. Collis, and W. W. Hunter, "A Practical Gamma-Ray Camera System Using High-Purity Germanium," IEEE Trans, on Nuc. Sci., NS-21, 1, Feb. 1974, p. 658. 4. L. Kaufman, et al., "Delay Line Readouts for High Purity Germanium Medical Imaging Cameras," IEEE Trans, on Nuc. Sci., NS-21, 1, Feb. 1974, p. 652. 5. P. M. Will and K. S. Pennington, "Grid Coding: A Novel Technique for Image Processing," Proc. IEEE, 60, 6, June 1972, p. 669. 6. R. H. Dicke, "Scatter-Hole Cameras for X-Rays and Gamma Rays," The Astrophysical Journal, 153, August 1968, p. L101. 7. L. Chang, et al., "A Method of Tomographic Imaging Using a Multiple Pinhole-Coded Apertures," J. Nuclear Medicine, 15, 11, p. 1063. 8. II. H. Barrett, "Pulse Compression Techniques in Nuc lear Medicine," Proc. IEEE, 60, 6, June 1972, p. 723. 9. K. F. Koral, W. L. Rogers, and G. F. Knoll, "Digital Tomographic Imaging with Time-Modulated Pseudorandom Coded Aperture and Anger Camera," J. Nuclear Medicine, 16, 5, May 1975; p. 402. 10. H. J. Caulfield and A. D. Williams, "An Introduction to Holography by Shadow Casting," Optical Engineering, 12, 3; 1973. 229 230 11. H. H. Barrett and F. A. Horrigan, Applied Optics, 12, November 1973. 12. H. H. Barrett, et al., "The Use of Half-Tone Screens in Fresnel Zone Plate Imaging of Incoherent Sources," Opt. Comm. , 15, 398; 1972. 13. D. T. Wilson, et al., "Point Source Artifacts in Fresnel Zone Plate Imaging," Raytheon Report, T-945; April 2, 1973. 14. H. J. Caulfield and Sun La, The Applications of Holography, John Wiley and Sons, New York; 1970. 15. H. M. Smith, Principles of Holography, Wiley and Sons, New York; 1965T 16. J. B. DeVelis and G. 0. Reynolds, Theory and Applica tion of Holography, Addison-Wesley, Reading, Mass.; 1967. 17. H. H. Barrett, "Fresnel Zone Plate Imaging in Nuclear Medicine," J. Nuclear Medicine, 13, 6, June 1972, p. 382. 18. H. H. Barrett, et al., "Recent Advances in Fresnel Zone-Plate Imaging," IAEA Symposium on Medical Radio- isotope Scintigraphy aT Monte Carlo, Monaco; Oct. 23-28, 1972; iAe A/SM-164/11S. 19. W. L. Rogers, et al., "Application of a Fresnel Zone Plate to Gamma Ray Imaging," J. Nuclear Medicine, 12, 8 , p . 612. 20. W. L. Rogers, et al., "Imaging in Nuclear Medicine with Incoherent Holography," Optical Engineering, 12, 13; 1973. 21. B. Macdonald, et al., "Gamma Ray Imaging Using a Fresnel Zone Plate Aperture, Multiwirc Proportional Chamber Detector, and Computer Reconstruction," IEEE Trans, on Nuc. Sci.. NS-21, Feb, 1974; p. 678. 22. H. H. Barrett, et al., "Apertures, Images and Nuclear Medicine," Optical Spectra, Feb. 1974; p. 31. 231 23. M. H. Farmelant, "Improved Anatomical Definition by a Fresnel Zone Plate Imager," J. Nuclear Medicinet 14, 6; p. 393. 24. H. H. Barrett, et al., "Fresnel Zone Plate Imaging in Radiology and Nuclear Medicine," Optical Engi neering , 1 2 , 8; 1973. 25. J. W. Goodman and R. W. Lawrence, "Digital Image Formation from F.lectronically Detected Holograms," App. Phys. Letters, 11, 3, Aug. 1, 1967; p. 77. 26. Y. Aoki and A. Boivin, "Computer Reconstruction of Images from a Microwave Hologram," Proc. IEEE, May 1970; p. 821. 27. Y. Aoki and Y. Suzuki, "Hologram Conversion by Computer," Proc. IEEE, May 1974; p. 643. 28. B. Macdonald, et al., op. cit. 29. E. 0. Brigham, The Fast Fourier Transform, Prentice- Hall, Inc.; Englewood Cliffs, N.J., 19/^7 p. 25. 30. ibid. , p. 25. 31. J. W. Goodman, Introduction to Fourier Optics, McGraw- Hill, New York, 1075; p. 18. 32. G. W. Stroke and M. Halioua, "Attainment of Diffraction- Limited Imaging in High-Resolution Electron Microscopy by 'a Posteriori'Holographic Image Sharpening," Optik, 35 (1), 1972; p. 50. 33. T. C. Rindfleisch, et al., "Digital Processing of Mariner 6 and 7 Pictures," J. of Geophysical Research, 76 (2), 1971; p. 394. 34. A. W. Lohmann and D. P. Paris, "Space Variant Image Formation," J. Opt. Soc. Am., 55 (8), 196S; p. 1007. 35. "LSI Chip Handles Fourier Transforms," Electronics. March 6, 1976; p. 32. 36. A. H. Deutchman, personal communication, The Ohio State University, 19,75. 232 37. A. H. Deutchman, "Quantitative Radiotracer Imaging; The Development of a Clinically Practical Instru mentation and Analysis System," Ph.D. dissertation, The Ohio State University; June 197S. 38. ibid. 39. Lord Rayleigh, Theory and Sound, Vol. II, New York, Dover, 1945. 40. R. W. Wood, Physical Optics, Macmillan, New York; 1914, second ed., pp. 37-40, 217, 218. 41. C. F. Meyer, The Diffraction of Light, X-Rays and Material Particles, UniiT. of Chicago Press, Chicago, 1934. 42. F. A. Jenkins and H. E. White, Fundamentals of Optics, New York; McGraw-Hill, 1957, p. 360. 43. C. T. Lane, "The Theory of the Zone Plate Derived from Voight's Integral," Canadian Journal of Physics, 2, 26-30, 1930. 44. Ora E. Myers, Jr., "Studies of Transmission Zone Plates," Am. J. Phys. , 19, 359; 1951. 45. Stigliani, Mittra, and Semonin, "Resolving Power of a Zone Plate," J. Opt. Soc. of Am., 57 (5), 1967. 46. A. Boivin, "On the Theory of Diffraction of Concentric Arrays of Ring-Shaped Apertures," J. Opt. Soc. of Am., 42, 60; 1952. 47. A. Boivin, A. Dion and H. Koenig, "Etude Expdrimentale de la Diffraction des Micro-ondes par dcs Ouvcrtures a symdtrie de revolution," Canadian J. Phys., 34, 166; 1956. *“ 48. J. Dyson, "Circular and Spiral Diffraction Gratings," Proc. Royal Society (London) Ser. A . 248, 93; 1958. 49. M. V. R. K. Murty, "Spherical Zone Plate Diffraction Grating," J. Opt. Soc. Am., 50, 923; 1969. 50. B. A. Lippman, "Exact Calculation of the Field Due to a Single Fresnel Zone," J. Opt. Soc. of Am., 55, 360; 1965. 233 51. H. M. Childers and D. E. Stone, ’’Solution to the Fresnel Zone Plate Problem” Am. J. Phys., 37, 721; 1969. “ 52. Murk Bottema, "Fresnel Zone-Plate Diffraction Patterns,” J. Opt. Soc. of Am., 59, 1632; 1969. 53. M. Young, "Zone Plates and Their Aberrations,” J. Opt. Soc. of Am., 62, 972; 1972. 54. K. Lehovec and A. Fcdotowsky, Theoretical Studies of Zone Plate Monolithically Integrated with Sensors-; AFCft'tTT2 - 0'49'ffT~l972: ~ 55. A. V. Baez, "Fresnel Zone Plate for Optical Image Formation Using Extreme Ultraviolet and Soft X-Radiation,” J. Opt. Soc. of Am., 51, 405; 1961. 56. M. V. R. K. Murty, "Common Path Interferometry Using Fresnel Zone Plates," J. Opt. Soc. of Am., 53, 568; 1962. 57. G. Boivin and R. Gagnon, "Methods d' Aligncment an Moycn D'une Mire a Foyers Multiples a Phase Inversee," Canadian J. Phys., 49 , 1284; 1971. 58. P. N. Keating, et al., "Fresnel-Zone-Plate Spectrometer with Center Stop," J. Opt. Soc. of Am., 62, 945; 1972. ” ” 59. G. L. Rodgers, "The manufacture of a stepped zone- plate," J. Sci. Instrum., 43, 328; 1966. 60. H. Chau, "Zone Plates Produced Optically," App. Opt., 8, 1209; 1969. 61. B. M. New, "Design, Production and Performance of Circular Fresnel Zone Plates," App. Opt., 10, 498; 1971. 62. A. V. Baez, "A Study in Diffraction Microscopy with Special Reference to X-rays," J. Opt. Soc. of Am., 42, 7S6; 1952. 63. P. Kirkpatrick and H. El-Sum, "Image Formation by Reconstructed Wavefronts, I. Physical Principles and Methods of Refinements," J. Opt. Soc. of Am., 46, 825; 1956. 234 64. A. Lohmann, "Optical Single-Sideband Transmission Applied to the Gabor Microscope," Opt. Acta, 3, 97; 1956. 65. L. Cutrova, E. Leith, C. Palermo and L. Porcello, "Optical Data Provessing and Filtering Systems," IRE Transactions on Information Theory IT-6, 386; i960. 66. E. Leith and J. Upatnieks, "Reconstructed Wavefronts and Communication Theory," J. Opt. Soc. of Am., 52, 1123; 1962. “ 67. E. Leith and J. Upatnieks, "Wavefront Reconstruction and Continuous Tone Objects," J. Opt. Soc. of Am., 53, 1377; 1963. 68. G. L. Rogers, "The Design of Experiments for Recording and Reconstructing Three-dimensional Objects in Coherent Light (Holograph)," J. Sci. Instrum.,43, 677; 1966. 69. G. L. Rogers, "Gabor Diffraction Microscopy: the Hologram as a Generalized Zone Plate," Nature, 166, 237; 1950. 70. W. E. Kock, et al., "Holograms and Zone Plates," IEEE Proc. , 19, 1599; 1966. 71. M. H. Ilorman and H. H. Chau, "Zone Plate Theory and Holography," App. Opt., 6, 217; 1967. 72. K. Clifford and G. Waldman, "Comments on Zone Plate Theory Based on Holography," App. Opt.. 6, 1415; 1967. 73. M. H. Ilorman, "Reply to Comments on Zone Plate Theory Based on Holography," App. Opt., 6, 1415; 1967. 74. M. H. Horman, Efficiencies of Zone Plates and Phase Zone Plates," App. Opt., 6, 2011; 1967. 75. W. G. Ferrier, "The Zone Plate and Its Role in Holography," Contemporary Phsyics, 10, 413; 1969. 76. J. P. Waters, "Holographic Image Synthesis Utilizing Theoretical Methods," App. Phys. Letters, 9, 405; 1966. 77. W. J. Siemens-Wapniarski and M, Parker Givens, "The Experimental Production of Synthetic Holograms," App. Opt. , 7_, 535; 1968, 23S 78. H. H. Barrett and F. A. Horrigan, "Theory of Fresnel Zone Plate Imaging of Gamma Rays," Raytheon Technical Memorandum (T-926), 1973. 79. H. H. Barrett and G. D. De Meester, "Quantum Noise in Fresnel Zone Plate Imaging," Raytheon Technical Memorandum (T-972), 1973. 80. M. L, G. Joy and S. Houle, "The Potential Performance of Off-Axis Fresnel Zone Plate Imaging Systems on Arbitrary Objects," IRF.F, Transactions on Nuclear Science, Vol. NS-22, February, 197 5; p. 364. 81. S. Houle and M. L. G. Joy, "Small Signal Suppression in Coded Aperture Imaging in Nuclear Medicine," IEEE Transactions on Nuclear Science, Vol. NS-22, February 197 5; p. 369. 82. M. S. Gerber, "The Analysis and Development of Instru mentation for a Position Sensitive Semiconductor Gamma Ray Camera," Ph.D. dissertation, The Ohio State University; June 1975. 83. I. G. Zubal, "The Fabrication and Analysis of Orthogo nal Strip High Purity Germanium Detectors," M.Sc. thesis, The Ohio State University, August 1974. 84. P. A. Schlosser, personal communication, The Ohio State University, 1975. 85. M. S. Gerber, personal communication, The Ohio State University, 1975. 86. J. W. Goodman, Introduction to Fourier Optics, McGraw- Hill, New York, 1968; p. 45. 87. ibid.. p. 81