Using percolation networks to incorporate spatial-dose information for assessment of complication probability in radiotherapy.

Abstract

This study investigates an extension of recent cluster based methods of assessing the probability of complication in normal organs following radiotherapy treatment which delivers a spatially non-uniform radiation dose distribution. Current methods of assessing this complication probability are spatially degenerate and do not adequately assess the contiguity of damage done to tissue. Therefore, new measures of assessing complication after radiation exposure have been proposed for parallel and serial type organs. In parallel organs an interaction between cells within a functional subunit is stipulated and complication is regarded as a weighted sum of all clusters in the organ. This allows the assessment to account for all damage to the tissue whilst emphasising the importance of damage that accumulates into large and connected spatial regions addressing a deficiency in the current method of calculating complication probabilities. Several spatially-varying doses were analysed and simulated in silico. The simulations produce complication risk estimates for homogeneous dose distributions that are comparable to empirical results but which deviate with any dose inhomogeneity. The simulations also show that the standard method of dose transformation to an effective uniform dose is not valid in cluster based models.