THREE-DIMENSIONAL ULTRASOUND IMAGING OF THE EYE DONAL B. DOWNEY1,3, DAVID A. NICOLLE2, MORRIS F. LEVIN1 and AARON FENSTER1,3 London, Ontario 6 SUMMARY sound imaging is cost-effective, we believe technical We assessed whether an inexpensive, three-dimen­ improvements are needed before its full diagnostic sional (3D) ultrasound (US) imaging system could potential is realised. produce clinically useful 3D images, without causing Though orbital anatomy and pathology are three­ patient discomfort. Five patients were examined. The dimensional (3D), the B-mode ultrasound data are 3D US system consisted of a transducer holder presented to the examiner in a two-dimensional (2D) containing a mechanical motor, and a microcomputer. format. This is regardless of whether the information During data acquisition the transducer was mechani­ is viewed directly from the ultrasound monitor, from cally rotated for 22 seconds, while 200 two-dimensional videotape, or from photographic film or paper. To (2D) US images were collected and formed into a 3D interpret the data, the diagnostician mentally inte­ image by the computer. The 3D image was viewed on grates multiple 2D images into a 3D impression of the computer monitor. The 3D images correlated with the anatomy and pathology being examined? While the clinical and radiological findings. The new perspec­ experienced examiners are often extremely skilful at tives were helpful in diagnosing eye abnormalities and developing this mental 3D representation of the no patient discomfort occurred. The device was easy to anatomy, the process itself is inefficient and requires use. It is concluded that, as good-quality 3D and 2D US considerable learning on behalf of the operator. No images were produced quickly, with no patient dis­ matter how assiduously one performs the ultrasound comfort, and the device is inexpensive, uncomplicated, examination, or how well trained one is, it is and easily attached to existing ultrasound machines, it probable that at least some of the available will probably be useful in clinical practice. diagnostic information is lost during this process. For example, when the scan is performed by a Conventional ophthalmic ultrasonography is an technician and only a few 2D ultrasound photo­ established diagnostic technique that is essential for 1 graphic images are later reviewed by the clinician, the clinical practice of ophthalmology. Both bright­ ness mode (B-mode) imaging, which gives informa­ inaccuracies may occur because out-of-plane features tion about the topographic nature of ocular and have not been recorded. In addition, it is often orbital lesions, and amplitude mode (A-mode) difficult to localise the exact plane in which any 2D imaging, which gives information about the size and ophthalmic image was obtained. This makes locating nature of lesions, are widely used? Increasingly, the exact position of an anomaly difficult. To colour Doppler imaging is being utilised to assess reproduce a particular image plane at a later time, blood flow in the blood vessels of the eye and or to follow up some orbital pathologies with 2D US S orbit?- Although it is generally agreed that ultra- is complex, especially if the abnormalities have 8 tapered or indistinct margins. Standard 2D images are usually obtained in an I From: Department of Diagnostic Radiology and Nuclear antero-posterior direction, either in a transverse, 2 Medicine, and Department of Ophthalmology, University axial or longitudinal plane, or in a plane parallel or Hospital, University of Western Ontario, London, Ontario N6A 3 ? 5A5, Canada; The Tom Lawson Family Imaging Research slightly oblique to, one of these planes Some ocular Laboratories, The John P. Robarts Research Institute, 100 Perth pathologies may be more appropriately viewed from Drive, London, Ontario N6A 5K8, Canada. a mediolateral orientation or a more oblique Correspondence to: Dr D6nal Downey, Department of Diagnostic Radiology and Nuclear Medicine, University Hospital, orientation, and these perspectives may not be 339 Windermere Road, London, Ontario N6A 5A5, Canada. adequately achieved using conventional 2D US. Eye (1996) 10, 75-81 © 1996 Royal College of Ophthalmologists 76 D. B. DOWNEY ET AL. Many authors have concluded that a 3D display of available, end-firing, sector type mechanical transdu­ the ultrasound information would correct some of cer (Ophthascan Mini B 10 MHz transducer, Bio­ these weaknesses and improve diagnostic ability. physic Medical, Clermont-Ferrand, France), coupled Many attempts have been made to develop 3D with a Macintosh Quadra 840A microcomputer systems suitable for use in different parts of the (Apple, Cupertino, CA). To obtain a 3D image, the 9 12 body. - Some endeavours have been reported in ultrasound transducer was inserted into a specially 13 17 the ophthalmology literature, - though none, as designed motorised transducer holder (Fig. 1). The yet, have been widely implemented. These systems holder was held by the examiner in a manner that were generally unsuccessful either because the 3D allowed the transducer tip to be closely coupled with imaging devices were too slow due to immature the closed eyelid of the supine patient.u; Copious technology, or because they required complex amounts of ultrasound coupling gel facilitated good instrumentation. Previously, 3D ultrasound images visualisation of structures in the orbit (Figs. 3, 4). The were acquired slowly and the structures that were 20 B-mode ultrasound image was optimised by clinically reviewed were generated manually by altering the time-gain curve, field-of-viewand overall tracing the contour of the organ on each 2D image. gain settings. When a focal abnormality in the eye These images were stacked on top of each other to was being viewed (e.g. melanoma) the lesion was create a 'wire frame' representation. These wire positioned in the centre of the field of view of the 2D frames were then assembled by computer into a 3D image. This ensured that the ultrasound beam was image and shaded to produce a surface-rendered 13 15 perpendicular to the object being examined, allowing representation. This technique is labour-inten­ optimum measurements to be made. sive and excludes important ultrasound information The patient kept eye and body still, and the about the interior of the organ. The images appear examiner held the transducer holder as steadily as artificial, making interpretation difficult. possible (Figs. 3, 4). The motor was activated for 22 Recently it has become possible to purchase fast, seconds, causing the transducer to rotate about its powerful and inexpensive microcomputers, and we own axis through 2000 (Fig. 2, top). After each l' believe that now is an appropriate time to re­ angular movement, the resulting ultrasound B-mode evaluate methods of 3D ultrasound information video image was captured by a frame grabber display. The following were considered important (Precision Digital Images, Redmond, WA) and criteria in the design of this new system. All stored in the microcomputer (Fig. 2, bottom left). diagnostically relevant ultrasound information from Care was taken to ensure that the cable from the the original 2D images should be preserved in the reconstructed 3D image. It should also be possible to ultrasound transducer never became twisted, and view and manipulate the 3D volume interactively in that it remained clear of the patient's face at all distinct 2D planes, orientated in any direction. The times. 2D 'slices' obtained from the 3D image should be Immediately after the scan, the 200 B-mode similar to those obtained in a standard 20 examina­ images were reviewed on the microcomputer moni­ tion. 'Slicing' of the 3D image should occur rapidly, tor sequentially, and evaluated to ensure that each data acquisition and reconstruction times should be had been accurately recorded and that no patient or short, and the system should be relatively inexpen­ examiner motion had occurred during the examina­ SIve. tion. If any motion was evident, another data On the basis of this approach we constructed a acquisition was obtained and the original one was prototype 3D US imaging system. The purposes of discarded. Two satisfactory 3D data sets were this preliminary study were twofold: (1) to evaluate obtained on each patient. whether this new system could produce accurate The acquired images were reconstructed into a 3D images in vivo, and (2) to ensure that the technique image (Fig. 2, bottom right) using a computer would be acceptable to patients. program developed in our laboratory. Data recon­ struction time was 140 seconds. The viewing software, also developed in our MATERIALS AND METHODS laboratory, allows the data to be viewed in different Five patients with different ocular pathologies - a ways. The simplest involves presenting the acquired choroidal melanoma, a post-traumatic vitreous haem­ 200 B-mode 2D US images to the examiner on the orrhage and retinal detachment, a simple retinal computer screen. This is the easiest method for most detachment, a treated retinal detachment and a retro­ beginners to interpret, as the data are exactly the orbital tumour - were evaluated with the 3D same as they would see in a standard examination, ultrasound imaging system. All gave informed con­ except that there are many more 2D images sent and the study received approval from the presented for review. The 3D display software allows university human ethics committee. The prototype the examiner to rotate the 3D image in any direction, 3D US system consisted of a standard, commercially and 'slice' it interactively to obtain the desired 3D ULTRASOUND OF THE EYE 77 Fig. 1. Motorised transducer holder. The white ultrasound Fig. 3. 3D patient scanning. The coupled holder and transducer has been coupled with the black motorised transducer are applied to the closed eyelid llsing ultrasound holder. coupling gel. perspective. Viewing is also possible with other mouse, so the impression the evaluator gets is very commercially available 3D viewing software. The similar to that obtained while scanning a patient in microcomputer monitor (Figs.