Episcleral plaques containing I-125 or various other isotopes are often used in the treatment of ophthalmic tumors (1-10). Plaque therapy is more accessible and less expensive than heavy charged particle teletherapy (11-13) and has dosimetric advantages compared to x-ray teletherapy. A randomized national study comparing brachytherapy with enucleation, the Collaborative Ocular Melanoma Study (COMS) is currently in progress (2).

Tumor control rates are excellent for both brachytherapy and external beam techniques, particularly for tumors up to 8-mm height. Within four years post-treatment, however, the majority of patients experience some visual loss due to radiation retinopathy. Complications include scleral necrosis, macular edema, cataract, neovascular glaucoma and vasculopathy of the retina and optic nerve (2,8,9,14-16). The high incidence of late complications suggests that it is worthwhile exploring methods of treatment optimization. The goal of this optimization would be to reduce the incidence and severity of vision-limiting complications without compromising tumor control.

Ophthalmic plaques fall into two general categories; (i) those that are supplied as ophthalmic applicators (1,5,10,14,17) with relatively long lived isotopes (Co-60 and Ru-106) and (ii) those into which sealed radioisotope sources (often referred to as "seeds") of I-125 or Ir-192 are temporarily inserted (2,4,6,9,18,22). An advantage of I-125 compared to other isotopes is that the dose distribution can be readily modified (7,18). There are four physical parameters which influence the dose distribution of an I-125 plaque; distance, source strength, anisotropy and shielding. Distance is an important factor since dose falls (approximately) as the inverse square of distance from a small radiation source in vacuum. In tissue, dose decline for I-125 radiation is somewhat more rapid due to absorption of the low-energy photon emissions. The second parameter is the distribution of activity amongst the sources. This might involve using sources of differing activities in combination with one another, and placing the highest activity sources in the portion of the plaque most distal to the region for which dose reduction is sought. The third parameter is source anisotropy (19,20). The I-125 in a model 6711 seed (3M Corp., Medical Products Div., St. Paul, Minnesota, USA) is deposited on the surface of a silver wire sealed in a thin, cylindrical, titanium shell. Photons leaving the seed perpendicularly to the axis of the silver wire (the longitudinal axis) pass through the minimum thickness of metal. Since the longitudinal axes of the sources are nominally tangent to the scleral surface, most of the direct tumor dose is delivered by these photons. As the longitudinal angle of exit deviates from perpendicular, self-absorption and the thickness of metal which the photons must cross increase. For the low-energy radiation of I-125, virtually all interaction in the metal is photoelectric absorption, which produces no scatter. The characteristic radiation emitted from the titanium is too weak to penetrate even 1 mm into tissue (21). This inherent longitudinal "self-shielding" can be exploited by orienting sources with their axes directed towards the region for which dose reduction is desired. The fourth parameter of plaque design, which can reduce dose to a particular region, is shielding by the plaque shell. A 0.5 mm thick gold shell reduces the dose from direct I-125 irradiation by more than 99.9%.

We have developed a three-dimensional computer model (18) for ophthalmic plaque dosimetry. Patient-specific ocular anatomy is derived from computed tomography (CT) and/or magnetic resonance (MR) images, fundus photography and ultrasound. Plaque orientation, source location, activity, decay, anisotropy, and collimation of the primary photon flux by the gold shell are all accounted for. By optimizing these parameters it may be possible to reduce the dose to nearby critical structures without compromising dose to the tumor. This would be expected to reduce the severity of late vision-limiting complications while maintaining tumor control. For a 5 mm tall tumor, the COMS protocol specifies a dose of 100 Gy to the tumor apex at 50-125 cGy/h. Whereas the sclera can tolerate doses of over 400 Gy without complications (5), retinal structures appear to be limited to doses below 60 Gy (5). This suggests that critical retinal structures such as the macula should be considered as primary optimization sites. The macula is a small avascular region of the retina near the posterior pole which encloses the fovea, the site of most acute vision. Although there is no general agreement on specific criteria for optimization at present, our initial figure-of-merit is the ratio of dose to the tumor apex (T) compared to dose to the center of the macula (M). The initial goal of optimization is to maximize this ratio. In this work we present computations which indicate that manipulation of the optimization parameters can substantially influence the tumor to macula (T:M) dose ratio for one of the standard COMS plaques and a posteriorly located tumor.

The model and computer system have been described previously (18) and compared to thermoluminescent dosimetry (TLD) measurements in an acrylic phantom (22). The model initializes to a "standard" eye based on the size and descriptions of Newel (23), Last (24), and the COMS group, and is subsequently adjusted to conform to transecting CT images and direct or ultrasonic measurements. The outline of posterior tumors and blood vessels are digitized from fundus photographs. Tumor height is obtained from ultrasound.

A spherical eye with an equatorial diameter of 24.6 mm was selected for this study in order to conform exactly with the design of the COMS plaques. (Trachsel Dental Studio, Inc., Rochester, MN 55901). Our model presently considers the tumor to be an irregular cone with a semispherical base, bounded by its apex and the intersection of its perimeter with the retina. The retina was modeled as a semispherical surface inset 1 mm from the external sclera. For the "optimization" test cases, the tumor perimeter was digitized from fundus photography of a posterior tumor taken from a clinical case.

The tumor height was reported to be 5.4 mm from ultrasonic examination. The apex of the tumor (in the model) was designated to be a point 6.4 mm inset from the plaque-scleral interface (5.4-mm height + 1 mm scleral thickness), along a line passing through the center of the tumor and the center of the major sphere of the eye. The plaque was positioned so that this line intersected the center of the plaque.

Dose matrices of 160X160 points were calculated for planar and spherical surfaces. Bilinear interpolation was used to estimate the dose distribution between the calculated points. The dose distribution on the retinal surface is particularly useful for studying margins around the tumor, and may eventually prove useful for correlating late visual field defects with radiation dose. The program also makes available a 3D dose volume histogram. Dose volume is estimated by dividing the eye volume into 1 mm³ voxels, and calculating the dose at the geometrical center of each voxel.

The dose constant at 1 cm was taken to be 1.32 cGy cm²/mCi h as currently recommended by the COMS. Recent calculations by Williamson (21) and measurements by Luxton et al. (25) and Weaver et al.(26), however, have indicated a 10%-20% lower exposure rate constant may be more correct. All seed activities refer to the activity at the beginning of the calculation period. The COMS dosimetry protocol presently requires that seeds be considered as isotropic point sources. For these calculations, however, the seeds were represented as anisotropic line sources (18,19,22).

For the calculations of optimization test cases 1-3, a standard 14 mm diameter COMS plaque was "loaded" with nine 4.0 mCi I-125 seeds. The particular locations in the plaque and orientation are illustrated in Figs. 1-4. In the first test case, shielding by the gold shell was ignored. The second case is identical to the first case, with the exception that shielding by the gold shell was accounted for. In the third test case, the same nine seeds were used, but the location and orientation of the seeds and the silicon seed carrier were adjusted to improve the T:M ratio. In the last test case four higher intensity seeds were used. Three 6.5 mCi sources were placed in the portion of the plaque most distal to the macula, and one 9.0 mCi seed was placed in the location closest to the macula, but shielded by the lip. In each case treatment time was adjusted to deliver 100 Gy to the apex of the tumor.