This program can potentially be applied to optimize plaque design, seed placement, and treatment planning. For example, in the geometry depicted in Figure 10, if the dose at the 6 mm normalization point were 100 Gy, then the estimated dose to the center of the optic disc would be about 80 Gy without accounting for lip collimation, and about half that when lip collimation is included. If a 40 Gy variation in dose to a critical structure were associated with a significant difference in complication rate, the ability to account for lip collimation will be important. In addition to plaque orientation, the effects of using mixed isotopes, and/or seeds of unequal activity can be modelled. This may be of value in reusing sources and tailoring the dose distribution to an irregular tumor volume. The ability to immediately display point dose values without necessarily calculating a complete dose matrix provides rapid feedback for plaque design, and permits detailed study of dose gradients within structures such as the lens, optic disc or macula.
Absolute dosimetry for plaques containing Ir-192 seeds, and relative dosimetry for I-125 seeds, were compared to TLD measurements in an acrylic phantom for the calculational model used in this program by Luxton et al. (17). The dosimetry for I-125 seeds, however, should be considered only approximate at this time. Recent Monte Carlo evaluations by Williamson (24) suggest that characteristic X rays from the titanium seed capsule may lead to a 7% overestimate of the specific dose constant when calibration is performed in air. These low energy photons are rapidly absorbed in water-like media and therefore have negligible contribution to tissue dose at distances greater than 0.5 mm. Williamson (24) suggests a specific dose constant (the ratio of absolute dose rate at 1 cm on the transverse bisector of a seed in a specific medium to the source strength) of 0.909 for the model 6711 seed. This is about 14% lower than the 1.035 value recommended by Ling et al. (15) for the same model seed. In light of the present uncertainties concerning the absolute dosimetry of I-125 seeds, all dose distributions reported here were normalized to a point on the central axis of the plaque, 6.0 mm from the plaque surface, and near the apex of a hypothetical 5 mm tumor. This represents a typical dose specification point at this institution and by the COMS.
The multi-window, 3-dimensional graphics interface provides an intuitive environment for both the physicist and physician to work within. Goitein and Miller (9) used a similar approach for planning proton therapy of the eye. This earlier work was implemented in FORTRAN on a computer (VAX 11/780) and required about 1.3 full-time-equivalent (FTE) man-years to implement. The program described here was developed in about 0.4 FTE man-years with the aid of the ROM "toolbox" functions built into the Macintosh computer. Since the program adheres to the recommended user interface guidelines, it is simple to learn and use, and is an example of how this new hardware and software technology can be rapidly and successfully applied to clinical dosimetry tasks.
There are several aspects of the current dosimetry system which require refinement prior to clinical implementation. The generalized approximation of intraocular anatomy within a spherical globe may be too imprecise for many variants of human ocular anatomy. Digitization of the tumor perimeter from fundus photography, estimation of the 3-dimensional tumor volume from sequential CT or MRI images, and display of the 3-dimensional tumor volume within a translucent eye would be desirable additions. These would provide visual aids during plaque positioning and permit correlation of isodose surfaces with actual tumor and normal tissue contours. Surgical reproduction of the preplanned plaque orientation and location could be difficult to achieve. If the source locations are fixed relative to the suture eyelets on the plaque, however, and the spatial location of the suture eyelets relative to real anatomical landmarks, such as the limbus, can be specified, precise surgical placement may be facilitated. Three-dimensional isodose surface displays and loading optimization based on available isotope resources would also be desirable. The lip collimation algorithm currently assumes a circular lip. This could be refined to handle lips of arbitrarily shaped perimeters. The COMS plaques use a silicone rubber carrier to offset the sources by 1 mm from the sclera. Our model presently considers this intervening medium to be tissue equivalent. For the low energy photons from I-125 seeds, photoelectric attenuation and scatter characteristics in silicone rubber could vary significantly from tissue and probably need to be accounted for. Dose resulting from scatter into the penumbral region close to the plaque perimeter is not modelled and may be of significance as well. We are presently adding these capabilities to the dosimetry system reported here.
REFERENCESAcknowledgements - Portions of the original computer system used to help develop this program were provided through an education, research and development grant from IBM Corporation (University of Southern California project Socrates), and an American Cancer Society Institutional Research Grant ACS-IRG IN-21-W through the USC Comprehensive Cancer Center. The Macintosh II implementation was assisted in part by a gift from BSD Medical Corporation.