A convolution algorithm for brachytherapy dose computations in heterogeneous geometries

Currently-available brachytherapy dose computation algorithms ignore heterogeneities such as tissue-air interfaces, shielded gynecological colpostats, and tissue-composition variations in 125I implants despite dose computation errors as large as 40%. To calculate dose in the presence of tissue and applicator heterogeneities, a computer code has been developed that describes scatter dose as a 3-D spatial integral which convolves primary photon fluence with a dose-spread array. The dose-spread array describes the distribution of dose due to multiple scattering about a single primary interaction site and is precomputed by the Monte Carlo method. To correct for heterogeneities traversed by the primary photons, the dose-spread array is renormalized to reflect the density and composition of the element, and the distance to the point of interest is scaled by the path-length of the intervening medium. Convolution calculations for 125I and 137Cs point sources in the presence of finite phantoms, air voids and high-density shields have been compared to the corresponding Monte Carlo calculations. The convolution code absolute and relative dose rate predictions are shown to agree with Monte Carlo calculations within 3%. Direct evaluation of the 3-D spatial convolution integral using 1-D adaptive integration reveals efficiency gains of 20-50 relative to Monte Carlo photon-transport calculations.