INVITED COMMENTARY Adding Structure to Function

and staging of malignant neoplasm by rate results are obtained with attenua w.hen cross-sectional anatomic im FDG PET are technically demanding. tion-correction versus emission-only aging emerged over two decades ago, A large portion of the body must be images is lacking. Centers continue to Henry N. Wagner, Jr., recognized that imaged quickly, small deposits of neo use noncorrected images for whole- the nuclear medicine community should plasm detected accurately, and the PET body oncology FDG PET, and when focus on tissue and organ function findings interpreted in the context of attenuation correction is performed, rather than anatomy (7). The value of the patient's corresponding normal and noncorrected images are often con functional imaging has not only be abnormal anatomy. At present, 2 practi sulted, in addition to corrected images, come clear since that time but is now cal, related considerations drive the before the final interpretation is ren recognized as the future of diagnostic merging of PET and CT: the need for a dered (7). imaging. Structural and functional im rapid, noise-free transmission scan for In response to the above needs, rapid aging are also increasingly understood attenuation correction of the PET emis advances have been made in the past 5 as complementary rather than compet sion data and a need for an anatomic years in the attenuation-correction hard ing imaging modalities. Over the past framework for the physiologic informa ware and software. The capacity to decade, manufacturers have considered tion provided by PET. perform transmission scans after tracer developing combined CT and PET or Without attenuation correction, PET injection has made routine whole-body SPECT imaging devices. CT, PET, and images are spatially distorted, and there (cf. torso) attenuation-corrected FDG SPECT are, in fact, closely related is marked radial nonuniformity in the PET feasible in clinical practice. High- imaging devices and share a common depiction of tracer activity (6). The activity single sources have been devel origin. In some respects, x-ray CT is application of a 511-keV transmission oped to improve counting statistics of the stepchild of transmission computed map to the reconstruction of the PET the transmission scan (11), and segmen tomography (2), which was first used emission data yields a spatially and tation methods allow elimination of the by Kühland Edwards (2) with a colli- quantitatively accurate depiction of the noise propagation, although with risks mated radionuclide source, as a direct tracer activity. High-quality, attenua of segmentation error (72). However, extension of their work in emission tion-corrected FDG PET images con these developments have not fully computed tomography (3). The 2 ar tain recognizable anatomic structure; solved the limitations of attenuation ticles by Beyer et al. (4) and Patton et yet, debate continues over whether the correction. Ideally, attenuation correc al. (5) in this issue of The Journal of use of current technology for attenua tion would add very little to the scanner Nuclear Medicine mark the beginning tion correction is truly advantageous time imposed by the emission-image of a new era in the growing integration over noncorrected PET for clinical on acquisitions, and there would be essen of functional and structural imaging cology diagnosis (7). At a practical tially no noise and no possible segmen into clinical practice. level, the additional scanner time tation errors propagated into the final The work reported by Beyer et al. (4) needed for the transmission scan using attenuation-corrected emission images. and Patton et al. (5) is part of a broad current technology can nearly double Kinahan et al. (73) have shown that it is trend toward the fusion of complemen total PET study time, imposing in possible to use CT imaging data to tary diagnostic imaging modalities. The creased likelihood of patient-move construct an attenuation map scaled to merging of biologic imaging, embod ment artifacts and decreased patient 511 keV. This essentially noise-free ied in PET, with the rapid and detailed throughput, ultimately decreasing the attenuation map, in which there is no structural imaging provided by x-ray diagnostic accuracy and increasing the need for segmentation assumptions, can CT may well be among the most fruit cost of PET imaging, respectively. Be then be used for attenuation correction ful direction of this trend. The impetus cause the typical transmission scan is of the PET imaging data. The time to build combined scanners has emerged very count poor, substantial statistical imposition for this approach is limited with the rapid growth of whole-body noise is propagated into the recon only by the speed of contemporary oncology PET imaging. The diagnosis structed attenuation-corrected emission spiral CT technology, which is mea scan, resulting in apparent decreases in sured in seconds for each PET emis Received Dec. 16,1999; revision accepted Feb. lesion contrast and small-lesion detect- sion bed position. Thus, PET/CT 16,2000. ability (8-10). Anecdotal evidence of presents a plausible solution to the For correspondence or reprints contact: Paul D. Shrove, Division of Nuclear Medicine, University the value of attenuation correction in need for a very rapid, low-noise, and Hospital, University of Michigan, B1G 505-UH-Box FDG PET imaging is widespread, but quantitatively correct method of PET 2008, 1500 E. Medical Center Dr., Ann Arbor, Ml 48109. generalized evidence that more accu attenuation correction. Furthermore,

1380 THEJOURNALOFNUCLEARMEDICINE•Vol. 41 •No. 8 •August 2000 anatomic information in the CT images performed in response to a finding, or dural planning, and communication of can be incorporated into the image lack of findings, on CT. For larger findings to practitioners. reconstruction process, potentially en neoplasms (i.e., > 1.5-2 cm), anatomic There are arguments against the hancing the speed and accuracy of correlation often can be performed by merger of PET and CT in the form of a numerical reconstruction algorithms visual reference within the attenuation- single scanner. If PET is expensive, and scatter and partial volume correc corrected PET images and to the CT would not the addition of CT make it tions (14,15). images. For example, fused PET and more so? First, PET is not expensive. Many technical hurdles still remain. CT images added little to diagnostic Even high-end PET scanners have been, The rapid acquisition capabilities of accuracy in an early study of mediasti- and remain, priced in the mid to upper contemporary spiral CT can effectively nal lymph node staging in lung cancer range of MRI scanners. Second, PET freeze respiratory motion in the chest (17). As the size of detected neoplasm scanners already have transmission CT and upper abdomen, resulting in misreg gets smaller, however, "eyeball" corre capability built in them in the form of istration of the transmission scan with lation becomes increasingly inadequate, the transmission scan sources. Substi the PET emission data, for which respi particularly in the neck, abdomen, and tuting a widely used CT tube and ratory motion is averaged. Intravenous pelvis. Data showing that fused PET detector, for which development and contrast when given as a bolus can and CT images improve accuracy com manufacturing costs are distributed over result in intravascular attenuation dur pared with separate interpretation of many units, for the transmission scan ing the CT acquisition, which inadver the FDG PET and CT images in these sources and related mechanicals results tently scales to bone rather than soft more challenging situations have yet to in a modest increase in cost for a new tissue, creating attenuation-correction be published, because the hardware generation PET scanner. If the incorpo errors. These are solvable problems, and software for true image fusion is ration of CT reduces overall scan time however. CT for radiation therapy plan only now becoming available. to roughly 30 min for imaging the ning is conducted without the person Clinical PET imaging in oncology is torso, patient throughput becomes com holding their breath and with slowed already moving beyond cases requiring parable with MRI and effectively re table motion to average respiratory straightforward interpretive skills, such duces the cost of PET imaging. Incorpo movement. Currently, it is not uncom as pulmonary nodules, enlarged medi- ration of CT into the next generation of mon to perform spiral CT sequentially astinal lymph nodes, and well-circum PET should be seen, then, as an up during 2 or 3 phases of intravenous scribed masses. Increasingly, practitio grade in the transmission scan capabil contrast enhancement in some body ners are asked to identify and locate ity over that of the current generation applications (e.g., precontrast, arterial deposits of malignant neoplasm that PET scanners. phase, venous phase). Thus, a CT of are not detected on CT scans, evaluate The same concept applies to SPECT the whole torso for PET attenuation the precise location of recurrent neo and coincidence-SPECT devices. The correction (with lowered beam current plasm for biopsy, and provide accurate work by Patton et al. (5) demonstrates to reduce dosimetry) could be followed delineation of true neoplasm extent for how a low-cost CT tube and detector by a rapid intravenous contrast-en radiation therapy planning. Experience system incorporated into a dual-head hanced CT as needed. In any case, the at the University of Michigan is that SPECT gamma camera can provide need for the aggressive use of intrave true image fusion would be highly relatively rapid low-noise transmission nous contrast in CT for oncology diag valuable and possibly indispensable for scans and provide useful anatomic in nosis may need reconsideration in light interpretation or subsequent procedural formation for interpretation of the func of the superior ability of FDG PET to applications in up to 20% of our cur tional images. This innovative ap detect malignant neoplasm; the FDG is rent whole-body oncology clinical FDG proach not only could solve some of the "smart" contrast. Finally, the abil PET studies. Because of the perceived the current limitations of sealed source ity of the CT-derived transmission scan need for fused PET and CT images for transmission devices applied to attenu to freeze respiratory motion may be interpretation and the demand for fused ation correction of cardiac SPECT but useful in the attenuation correction of PET and CT images from practitioners may aid interpretation of gallium-, anti respiratory gated PET emission data in radiation oncology, surgery, and in- body-, and peptide-based single photon (76), a strategy used to improve small- terventional radiology, image-fusion al tracers. The incremental cost of the lesion detection in the lung bases and gorithms are now being applied that are tube and detectors used in their system upper abdomen. capable of warped geometric deforma over the cost of a sealed source trans The second, and in some ways, more tions (18), although for the abdomen mission device for SPECT attenuation compelling driving force for the merger and pelvis, the algorithms are expected correction is modest considering the of PET and structural imaging is the to be inaccurate at times. It is becoming potential advantages for both single need for an anatomic framework for clear that the use of the powerful photon SPECT and coincidence FDG the physiologic information provided metabolic images provided by PET images. Although the structural images by PET. For cancer diagnosis and stag will require, at times, an anatomic are not of diagnostic quality, they may ing, FDG PET scans are nearly always scaffolding for interpretation, proce well suffice for many of the SPECT

ADDINGSTRUCTURETOFUNCTION•Shreve 1381 and coincidence applications; regard power of PET imaging becomes more 3. KühlDE, Hale J, Eaton WC. Transmission scan less, diagnostic CT images can be appreciated, the need for exquisite struc ning: a useful adjunct to conventional emission scanning for accurately keying isotope deposition aligned and fused with the structural tural image quality may diminish some to radiographie anatomy. Radiology. 1966:87:278- images as needed. what in the minds of diagnostic radiolo 284. Methods for fusion of PET and struc gists and referring physicians. 4. Beyer T, Townsend DW, Brun T, et al. A combined PET/CT scanner for clinical oncology. J NucíMed. tural images of the brain performed on The trade-offs of cost versus perfor 2000:41:1369-1379. separate scanners are well developed. mance will be addressed over the next 5. Patton JA, Delbeke D, Sandler MP. Image fusion The application of such rigid realign 5 years, with increasing experience in using an integrated, dual-head coincidence camera with x-ray tube-based attenuation maps. J Nucí ment fusion methods to nonrigid body clinical practice. Combined PET/CT Med. 2000:41:1364-1368. structures (19) has advanced to image scanners with contemporary diagnostic 6. Zasadny KR, Kison PV. Quint LE, Wahl RL. fusion with warped geometric deforma CT capability are currently under devel Untreated lung cancer: quantification of systematic distortion of tumor size and shape on non- tion capability (18). For the chest and opment. The approach used by Patton attenuation corrected 2-[fluorine-18]fluoro-2-deoxy- neck, this may be useful for interpre et al. (5) is already a marketable prod D-glucose PET scans. Radiology. 1996:201:873- tive applications; however, in the abdo uct and is being disseminated into the 876. 7. Wahl RW. To AC or not to AC: that is the question. J men and pelvis, movement of bowel clinical work environment. These differ NuclMed. 1999:40:2025-2028. and bowel gas between the structural ing approaches of adding structure to 8. Bengel FM, Ziegler RI, Avril N. Whole-body and PET imaging sessions could result function will be shaped by clinical positron emission tomography in clinical oncology: experience, even as clinical practice is comparison between attenuation corrected and un- in interpretive difficulties. To take full corrected images. Ear J NucíMed. 1997:24:1091- advantage of image fusion, especially being shaped by these technologies. 1098. if the structural imaging is used for Combined PET/CT scanners may be 9. Imran MB, Kubota K, Yamada S, et al. Lesion-to- a catalyzing development for clinical background ratio in nonattenuation-corrected whole- attenuation correction and related ana body FDG PET images. J NucíMed. 1998:39:1219- tomic information incorporated into PET. The combination has the potential 1223. PET image reconstruction algorithms, to solve some current limitations on 10. Bleckmann C, Dose J, Bohuslavizki KH, et al. the structural and PET imaging need to whole-body oncology imaging, with Effect of attenuation correction on lesion detectabil- ity in FDG PET of breast cancer. J NucíMed. be temporally and geographically as substantial shortening of imaging times 1999:40:2021-2024. close as possible. and possible improved overall PET 11. Karp IS, Muehllehner G, Qu H, Yan XH. Single There is general consensus that the image quality. Grafting the metabolic transmission in volume-imaging PET with a 137Cs information of PET onto the familiar source. Phys Med Biol. 1995:40:929-944. PET capability of a PET/CT scanner 12. Meikle SR, Bailey DL. Hooper PK, et al. Simulta should be high end. Debate remains, anatomic depiction of the body may neous emission and transmission measurements for however, over the required structural speed acceptance and use of PET by attenuation correction in whole-body PET. J Nucí referring physicians. Most important, Med. 1995:36:1680-1688. imaging capability that is appropriate 13. Kinahan PE. Townsend DW, Beyer T, Sashin D. for a combined scanner. Even low-end the integration of structural and bio Attenuation correction for a combined 3D PET/CT contemporary CT scanners produce im logic imaging will provide a bridge for scanner. MedPhy. 1998:25:2046-2053. the fundamental shift in medicine from 14. Ouyang X, Wong WH, Johnson VE, Hu X, Chen age quality that is suitable for most C-T. Incorporation of correlated structural images viewing disease as an overall anatomic diagnostic applications. High-end CT, in PET reconstruction. IEEE Tr Med Im. 1994:13: for the most part, translates into very process to a fundamentally molecular 627-628. process. If "form and function form the fast scan acquisition in an attempt to 15. Comtat C, Kinahan PE, Fessier JA, et al. Reconstruc future" (20), the functional capability tion of 3D whole-body PET using blurred anatomi squeeze some additional information cal labels. IEEE Tr Med 1m. 1999; 18:345-349. out of the intravenous contrast-en of PET will play a major role. 16. Miyauchi T, Raylman RR, Kison PV, et al. Can hanced studies. For oncology diagno respiratory gated FDG-PET improve image quality Paul D. Shreve in thoracic and upper abdominal tumor imaging sis, however, FDG PET renders the [abstract]? J NucíMed. 1996;37(suppl):36P. University of Michigan small incremental advantages of such 17. Wahl RL. Quint LE, Greenbough RL, et al. Staging high-intensity contrast exams largely Ann Arbor, Michigan of mediastinal non-small cell lung cancer with FDG PET, CT, and fusion images: preliminary prospec obsolete. The older spiral CT used in ACKNOWLEDGMENTS tive evaluation. Radiology. 1994:191:371-377. the prototype PET-CT scanner devel 18. Wahl RL, Quint LE, Cieslak RD, et al. "Anatometa- bolic" tumor imaging: fusion of FDG PET with CT oped by Beyer et al. (4) may have The author wishes to thank Richard Wahl, Jeffery Fessier, Charles Meyer, or MRI to localize foci of increased activity. J Nucí adequate image quality. The CT images Med. 1993:34:1190-1197. of these older scanners may not be and James Colsher for helpful discus 19. Meyer CR, Boes JL, Kim B, et al. Demonstration of entirely satisfactory to some diagnostic sions. accuracy and clinical versatility of mutual informa tion for automatic multimodality image fusion radiologists, but newer generation CT, REFERENCES using affine and thin-plate spline warped geometric even the lower priced systems, will deformations. Med Image Analysis. 1997; 1:195- generate better CT images at acquisi 1. Wagner HN. Images of the future. J NucíMed. 206. 1978:19:599-605. 20. Wahl RL. Imaging cancer in the new millennium: tion rates entirely suitable for a com 2. Friedland GW, Thurber BD. The birth of CT. A/A. form follows function [Eugene P. Pendergrass New bined scanner. In any case, as the 1996:167:1365-1374. Horizons Lecture). Radiology. 1999;213(P):25.

1382 THEJOURNALOFNUCLEARMEDICINE•Vol. 41 •No. 8 •August 2000