A self-collimating convolution backprojection algorithm for optimizing dose distributions of I-125 prostate implants

The algorithm presented here for optimizing brachytherapy dose distributions is based on the idea that the seed distribution can be modeled as an activity distribution determined analogously to gamma camera imaging. The peripheral dose to the tumor is converted to a set of uncollimated projection data that are then filtered and backprojected to produce an initial seed distribution. The actual doses resulting from the seed placement are used to correct the initial projection data for attenuation, scatter, and lack of collimation. The corrected projection data are backprojected a second time to yield the optimized but unconstrained seed distribution. Clinical constraints such as the number of different seed activities, the maximum seed activity, the minimum peripheral tumor dose, and the minimum percentage of the volume which receives less than a specified dose are then applied to the unconstrained solution. Through the entire process, the dose calculations are functions of source anisotropy, scatter, and attenuation. When applied to a set of elliptical contours, the algorithm produces elliptical peripheral dose isodose contours and reasonable dose volume histograms for a constrained solution. The results for actual patient prostate contours were not as good, primarily because of the difficulties encountered in dealing with the irregular geometry of the prostate. However, the algorithm shows promise for further research.