Diagrammatic representation of 3-dimensional image formation (from Downey et al20).
A cutaway illustration of the relative positions of a standard Collaborative Ocular Melanoma Study–type ophthalmic plaque and seeds secured to the episclera beneath an intraocular tumor.
In vitro images of a 14-mm Collaborative Ocular Melanoma Study–type gold eye plaque in 7% glycerin in water (room temperature). Left, A transverse section of the plaque contains small arrows that indicate the location of the seeds within the plaque. The 2 large arrows indicate the edges of the plaque. The small black arrowheads illustrate the location of the surface of the acrylic fixative and the plaque's posterior surface. Right, A coronal reconstruction of the plaque demonstrates the suture eyelets (white arrows) and the edge of the plaque (black arrow).
A 3-dimensional ultrasonographic reconstruction of an episcleral radioactive plaque in vivo. Characteristics that can be appreciated include the highly reflective back and edge of the plaque and shadowing behind the plaque (arrow).
Left, A 3-dimensional reconstruction of a radioactive plaque beneath a juxtapapillary choroidal melanoma. This computer-generated reconstruction demonstrates the relationship between the optic nerve (arrow) and the radioactive plaque. The posterior (nasal) edge of the plaque tilts as it rests against the optic nerve. Right, A coronal section demonstrating the plaque (P) next to the optic nerve (ON).
Finger PT, Romero JM, Rosen RB, Iezzi R, Emery R, Berson A. Three-dimensional Ultrasonography of Choroidal MelanomaLocalization of Radioactive Eye Plaques. Arch Ophthalmol. 1998;116(3):305-312. doi:10.1001/archopht.116.3.305
Copyright 1998 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.1998
To evaluate the use of 3-dimensional (3D) ultrasonography for the localization of episcleral eye plaques during the treatment of choroidal melanomas.
A series of 13 patients with choroidal melanoma were treated with radioactive palladium 103 seeds affixed into gold eye plaques. During surgery, 3D ultrasonography was performed with a commercially available system to evaluate the relative position of radioactive plaques secured beneath their intraocular tumors. This system consists of an automated, rotating, handheld, B-scan ultrasonographic probe operating at 10 MHz, a personal computer, and 3D imaging software.
We measured the margins of the plaque extending beyond the tumor and the distance between the radioactive seeds and the tumor apex. We also evaluated the relationship between the plaque edge, the episclera, and the tumor's edges. While the plaques were well centered over the tumor in all cases, the plaque margins around the tumor were found to be variably sized. When comparing measurements taken at the time of plaque insertion with those taken at the time of plaque removal, we noted changes in the apical tumor height and in plaque centration. In the 1 patient with a juxtapapillary tumor, the posterior margin of the plaque was found to be displaced away from the sclera, or "tilted."
Three-dimensional ultrasonography offers a new method for ophthalmic plaque localization. Unique perspectives can be visualized through the use of computer-aided 3D reconstructions that permit the assessment of the relative position of the plaque to the optic nerve and the measurement of the distance between the in vivo radioactive seed and the tumor apex. Our experience suggests that when compared with 2-dimensional ultrasonography, 3D ultrasonography offers new capabilities that can be used to improve plaque placement and radiation dose calculations.
EPISCLERAL plaque brachytherapy is the most frequently used "eye-sparing" treatment for choroidal melanoma.1- 7 In North America, eye plaques primarily consist of iodine 125 seeds affixed in a bowl-shaped gold shell. The gold of the eye plaque blocks most irradiation traveling in any direction other than toward the episcleral surface. Thus, gold eye plaques are safer, with less irradiation of normal ocular structures and less exposure to operating room personnel.
Directional, low-energy (125I and palladium 103) irradiation also requires the precise localization of the ophthalmic plaque onto the sclera, over the tumor's base.1- 7 To date, the accurate surgical placement of episcleral plaques (to cover their intraocular tumor) has relied on marking the transillumination shadow of the intraocular tumor, indirect ophthalmoscopy with scleral depression around the plaque, and 2-dimensional (2D) intraoperative ultrasonography.8- 12
In most cases, transillumination with visualization of an anterior tumor shadow can be performed with ease and offers an excellent method to ensure proper plaque localization. Small choroidal melanomas located in the posterior uvea are the most difficult to localize and are associated with the highest rate of irradiation failure.1,13,14 To ensure proper plaque localization of small posterior tumors, ophthalmic oncologists commonly use ophthalmoscopy with scleral depression, 2D ultrasonography (2D-US), or both. Magnetic resonance imaging has also been used to investigate plaque placement for radiotherapy.15- 17 Because plaque placement is critical to successful radiotherapy of a choroidal melanoma, and because reports suggest that failure of local control may reduce a patient's chance of survival, there is a need for improved methods to confirm proper plaque placement at the time of plaque insertion.18,19
This study investigates the usefulness of 3-dimensional ultrasonography (3D-US) for imaging radioactive plaques during radiotherapy. We describe how we measured the relative position of the eye plaques over their intraocular tumors and how computer-aided reconstructions can create new perspectives not previously available with 2D-US.20
Three-dimensional ultrasonographic studies were performed using a commercially available system (3D i-scan, Ophthalmic Technologies Inc, Toronto, Ontario). Three-dimensional ultrasonography uses a conventional brightness mode transducer combined with a motorized, rotating holder and computerized image processing (Figure 1). During data acquisition, the transducer is rotated while 1802D images are collected and processed by the computer to form a 3D image. After acquisition, it is possible to view and manipulate the 3D image interactively. Because the 3D image can be rotated and sliced, it reveals unique 2D images derived from new perspectives (Table 1).20,21
The B-scan ultrasonographic probe operates at a frequency of 10 MHz, with an axial resolution of 0.1 mm. The focal point is 25 mm, with a total image depth of 50 mm. For image processing, the proprietary software (3D i-scan) was run on a personal computer (Power MacIntosh 7500/100, Apple Computers, Cupertino, Calif). The acquisition and reconstruction times were 7.5 and 6.0 seconds, respectively. All 3D-US measurements were made at an almost equivalent sound velocity of 1532 m/s compared with 1550 m/s used for standardized A-scan ultrasonographic images.21 This difference was not thought to have an effect on this study in our measurements of heights, widths, lengths, or volumes (Table 1).
Fisher et al22 have studied the accuracy and reproducibility in vitro of measurements with this 3D-US system. The largest error (±SD) for repeated measurements was as follows: 0.02% (±0.06%) for linear measurements, 14.00 (±6.81) mm2 for area measurements, and 19.0 (±18.4) mm3 for volume measurements.22
For this study, one of us (J.M.R.) studied the intraobserver variability in vivo of measurements of intraocular tumors during radioactive plaque therapy. The SD for repeated measurements was found to be 0.2 mm for tumor diameter, 0.1 mm for tumor height, 7.8 mm3 for tumor volume, 0.4 mm for plaque margins around the tumor, and 0.1 mm for the distance between the radioactive seeds and the tumor apex.
Because 2D-US is commonly used to aid in plaque placement, and because the methods of accession of 3D-US were essentially equivalent, no international review board approval or informed consent was considered necessary.
In this series, 13 patients not included in the Collaborative Ocular Melanoma Study underwent ultrasonography. Patients were selected for 3D-US in addition to 2D-US based on the availability of the ultrasonograph. All tumors were treated at the Ocular Tumor Service, the New York Eye and Ear Infirmary, New York City, between October 1996 and August 1997; all patients underwent ophthalmic plaque brachytherapy, which was performed by one of us (P.T.F.). The following data were obtained from the patient medical records: affected eye, tumor location, shortest distance between the tumor and the fovea and between the tumor and the optic nerve, plaque size, and prescription point (tumor height measured as the distance between the tumor apex and the inner sclera, as measured by a preoperative A-scan ultrasonographic image) (Table 2). As possible, 3D-US was performed immediately after plaque insertion or just prior to plaque removal.
Standard eye plaques (Trachsel Dental Studio Inc, Rochester, Minn) were chosen (in part) to provide standard dimensions that could be compared with those obtained during our study and repeated in other centers. The radioactive plaque diameters in this series were as follows: 12 mm (2), 14 mm (6), 16 mm (1), 18 mm (3), and 20 mm (1). Plaques were made radioactive by affixing 103Pd seeds (model 200, Theragenics Corp, Norcross, Ga) onto the inner wall of the plaques with a thin layer of dental acrylic fixative (Figure 2).
The calculated seed-to-tumor apex distance (STAD) is defined as the distance between the central axis of the seed located in the middle of the gold plaque and the apex of the tumor. The actual dose of radiation is calculated based on the relative location of all the seeds affixed within the plaque. The calculated STAD for this study was determined as follows: the tumor height (determined by A-scan ultrasonography)+1.00 mm of sclera (the Collaborative Ocular Melanoma Study default value)+1.4 mm of acrylic fixative in front of the seed (measured value)+0.4 mm (half the thickness of the seed).
All patients underwent ophthalmic plaque insertion while under general anesthesia. After the patient was draped and the affected eye was isolated, indirect ophthalmoscopy was performed to confirm the presence, location, and condition of the eye and tumor. Then, a 360° conjunctival peritomy was performed at the corneal-scleral limbus, the Tenon fascia was opened with curved Steven scissors, and the rectus muscles were isolated as needed. In some cases, muscle disinsertion was necessary. In most cases, transillumination could be used to delineate the anterior tumor margins. Direct, transpupillary, and transocular transillumination techniques were employed. A surgical marking pen was used to mark the edges of the transillumination shadow on the episclera. Then, a 2- to 3-mm shadow-free margin was marked on the episclera around the tumor's base. A minimum of 4 interrupted episcleral sutures were used to anchor the plaque to the sclera, with its edges covering the tumor and its shadow-free margin (Figure 2). At this time, epibulbar contact 3D-US was performed. In this series, the information obtained by the 3D-US examination was not used to center the plaque. At the end of the procedure, the muscles were affixed to the sclera in positions and at tensions thought not to affect the plaque position. Last, the conjunctiva was closed. After radiotherapy, the plaques were removed while the patients were under local anesthesia.
At the time of scanning, the patients were on the operating table and in a supine position. Intraoperative localization involved placing the probe into a sterile plastic bag containing coupling gel. Saline solution was irrigated onto the surface of the globe. Postoperative confirmation of plaque position usually involved direct placement of the ultrasonographic probe on the eyelid and coupling gel. Because eyelid tissue is not a refractive surface for ultrasonography, we did not expect this difference to notably affect our measurements. A minimum of 2 sets of data were acquired for each eye plaque. One was oriented in a transverse direction, and a second was oriented in a longitudinal direction; both were centered in the meridian of the location of the intraocular tumor and plaque.
In vitro imaging of an eye plaque with dummy (nonradioactive) seeds was performed in 7% glycerin at room temperature (Figure 3). This phantom material was used because it was acoustically equivalent to vitreous (the speed of sound in vitreous=1532 m/s). During this study, we were able to define 3 echogenic interfaces corresponding to the surface of the acrylic fixative, the seeds, and the posterior wall of the plaque. The plaque's edges could also be identified and measured to be equivalent to the known diameter of the plaque. This in vitro experience confirmed our opinion about which linear echodensities represented the radioactive seeds in vivo (Figure 3).
One hundred eighty sequential, stored, 2D images were used to reconstruct the 3D image (Figure 4). Like a 2D image, the 3D image demonstrates the high reflectivity of the gold plaque. Like a 2D image, the highest reflectivity is noted at the plaque's edges and posterior face. These surfaces generate edge distortions as well as shadowing of the structures posterior to the plaque (Figure 4). Unlike a 2D image, the 3D reconstructions demonstrate the volume of plaque-induced shadowing. Edge reflections were found to be important in determining the relative position of the plaque in relation to the tumor and the adjacent structures (such as the optic nerve).
Because the plaque diameter is known prior to insertion, its size was found to be a useful reference to determine the position of the plaque in relation to the tumor's margins. The plaque diameter in the horizontal and vertical planes perpendicular to the plaque circumference was displayed. Measurements of the margin between the plaque's edge and the tumor's borders were determined in each quadrant. A replay of 3D reconstructions allows for the visualization of all the plaque edges and an assessment of the plaque's contact with the sclera. Unlike 2D-US images, 3D-US reconstructions can be visualized from unique perspectives. To demonstrate this property, we have rotated the reconstructed eye to view it from behind and show a unique coronal perspective of the relative positions of a juxtapapillary eye plaque and the optic nerve (Figure 5). This view is not possible with standard 2D-US because ultrasonographic waves cannot pass through orbital bones.
Three-dimensional ultrasonography was performed immediately after plaque insertion in 6 cases and immediately prior to plaque removal in all 13 cases. Measurements of tumor length, width, height, and volume were obtained. Tumors were located in various anterior-posterior and clock positions (Table 2). Case 9 was a juxtapapillary tumor. Tumor shapes were noted to be either dome or collar button. The prescription points ranged between 2.0 and 10.6 mm. Three-dimensional ultrasonographic apical height measurements were measured in a masked manner immediately after plaque insertion and were found to be essentially equivalent (P=.85) (Table 3).
Although most tumor dimensions were noted to increase after radiotherapy, we found that only the changes in height and volume were statistically significant (P=.001 and P=.01, respectively) (Table 4). Without histopathologic correlation, we can only hypothesize that these changes may have been due to acute radiation angiitis or postsurgical inflammation with secondary tumor and scleral thickening.
In all cases, we found that the tumors were covered by the plaques at the time of imaging (Table 5). While the edges of the plaques were found to be in good contact with the sclera in 12 cases, the plaque covering the juxtapapillary tumor was found to be displaced away from the sclera at the optic nerve margin, or "tilted" (Figure 5). We also found evidence suggesting displacement or "movement" of the eye plaques during treatment (Table 5). In the 6 plaques measured at the time of insertion and the time of removal, the largest measured change at any margin was 2.1 mm (along the temporal margin of case 3). However, because the original margin was 3.1 mm, the plaque continued to cover the tumor. Table 5 also demonstrates that the mean changes were in a range from −0.9 to 0.4 mm, and the average of those changes was −0.4 mm. This negative number corresponds (in part) to our observation that the tumor width and length may have increased during the course of radiotherapy (Table 4).
Once sewed to the eye wall, we assessed the position of the plaque in relation to the tumor apex by viewing 2D planes sliced perpendicular to the tumor base. Then, a computer-generated perpendicular line was created to measure the distance from the tumor apex to the layer of seeds within the plaque.
In Table 6, we provide the "calculated" STAD prior to surgery (based on presumptions commonly used for plaque dosimetry), the STAD immediately after plaque insertion, and the STAD immediately prior to plaque removal. Unlike standard 2D or A-scan ultrasonographic studies, 3D reconstructions allowed for a measurement of the distance between the radioactive seeds at the middle of the plaque and the tumor's apex (Table 6). Because all the radioactive seeds are affixed to the inner aspect of the plaque's posterior wall, we could (for the first time) calculate the actual distance between the radioactive seeds and points within the tumor at the time of plaque insertion and removal.
Our measured STAD was found to be shorter than our calculated STAD. The calculated STAD was an average of 1.2 mm longer than our measured STAD at the time of plaque insertion and 0.8 mm longer at the time of plaque removal (Table 6). This means that even though the tumors, surrounding tissues, or both were found to enlarge during the course of radiotherapy, the actual dose to the tumor's apex was higher compared with what it might have been determined using the measured STAD at the time of either insertion or removal (Table 6). The increase in the STAD during the course of treatment could be attributed to acute vasocongestion and edema as a result of surgery, irradiation, or both. However, changes in the STADs of as little as 0.8 mm (as found on average at the time of plaque removal) can have a notable effect on the calculation of the tumor dose.
Standard 2D-US techniques have been used intraoperatively and postoperatively to assess episcleral radioactive plaque position.8- 10 Yet, the role of ultrasonographic imaging has largely been defined as an adjunct to transillumination or scleral indentation in cases of small posterior tumors. Although ophthalmic oncologists have used ultrasonographic information, most are uncomfortable in relying on 2D image "slices" to confirm plaque placement.
While 2D-US and 3D-US studies are dynamic, they differ in that the collected images from conventional 2D-US studies are usually stored in a 2D format (Table 1). Typically, the surgeon takes 5 to 15 minutes imaging the plaque and the tumor while performing a mental 3D reconstruction.
In contrast, with 3D-US, 180 images are acquired and stored with each transducer rotation. Software uses this information to reconstruct a 3D image (Table 1). Then, without further contact with the patient, the entire examination can be played back over and over again (with the same perspective) or the eye (its tumor and plaque) can be rotated and viewed from any orientation. For example, the tumor and plaque can be viewed from oblique, mediolateral, or coronal views that cannot be acquired by 2D-US because of the presence of orbital bones (Figure 5).
Previous 2D-US imaging studies have been performed for the localization of episcleral radioactive plaques. Harbour et al8 imaged 29 cases using intraoperative evaluations; in 4 (14%) of the cases, the plaque did not cover at least 1 tumor margin, and in 2 (7%) of the juxtapapillary tumors, the plaques were found to be tilted away from the optic nerve. Williams et al10 performed postoperative ultrasonography on 16 cases and found accurate positioning in all of them. Pavlin et al9 scanned 9 cases postoperatively and found that 1 (11%) of the cases was malpositioned, 3 (33%) were eccentric, and 2 (22%) were displaced or tilted outwards.
Our study (using 3D-US) found that changes in tumor height during radiotherapy were statistically significant (P<.001), while changes in the tumor width (P=.08) and length (P=.09) were not significant. This finding underscores the difficulty in assessing changes in plaque centration and movement. Further refinements or additional techniques may be required to study plaque movements and changes in tumor basal dimensions.
Although relatively expensive and unavailable, magnetic resonance imaging has also been used to assess plaque position and offers computer-assisted multiplanar imaging capabilities (transverse, sagital, and coronal planes). Magnetic resonance imaging also provides 3D information; however, magnetic resonance imaging has some disadvantages. It is time consuming (40 minutes) and has a relatively low resolution (slice thickness, 3 mm; resolution, 0.5 mm). In contrast with magnetic resonance imaging, 3D-US allows for rapid acquisition (7.5 seconds) of multiple 2D images around a given axis. This data can be used to reconstruct a 3D image of the area of interest (6 seconds) that can be sectioned and viewed in any direction or orientation.
Transillumination and scleral depression will continue to be indispensable methods for episcleral plaque placement. The drawbacks of ultrasonographic and transillumination techniques reside in their relative inability to define tumor margins that extend anteriorly into the zone of the ora serrata, the ciliary body, and the iris. Transillumination shadows typically merge with the highly pigmented and thickened ciliary body, and orbital bones impede low-frequency contact ultrasonographic imaging of the anterior segment. The only way to image anterior segment tumors with standard ultrasonography is with water bath techniques that allow the transducer sufficient distance from the study object. Unfortunately, construction of a water bath during surgery is impractical, and the plaque's ultrasonographic shadow can obscure the tumor.
The methods of radiation dose calculation are based on ultrasonographic measurements of tumor thickness because the prescription point is the tumor's apex. Certain assumptions are made regarding the thickness of the sclera, the positioning of the plaque as "flush" with the episclera, and the location of the seeds within the plaque. In vivo tumor measurements of the STAD suggested that the tumor and the sclera may have thickened during the course of radiotherapy (Table 4, Table5 and Table 6). Although knowledge of the actual STAD would allow for less variability in irradiation dose and a better analysis of the effects of radiotherapy (local control and complications), it may require flexible scheduling of the operating room and multiple physics calculations. It is reasonable to assume that improved plaque localization and dosimetry would make radiotherapy more effective and would improve local control.
This study demonstrates that 3D-US offers unique views of episcleral plaques beneath their intraocular tumors. Three-dimensional ultrasonography suffers from some of the same limitations that affect 2D-US, but its ability to store and replay the examination from any orientation offers the potential to improve plaque placement and irradiation dosimetry.
Accepted for publication November 26, 1997.
Presented in part at the annual meeting of the Association for Research in Vision and Ophthalmology, Ft Lauderdale, Fla, May 14, 1997.
Reprints: Paul T. Finger, MD, the Ocular Tumor Service, the New York Eye and Ear Infirmary, 310 E 14th St, New York, NY 10003 (e-mail: firstname.lastname@example.org).