Simulation Study of Dose Estimation via Compton-based Prompt Gamma Imaging during Proton Therapy: a Deconvolution Approach based on Evolutionary Algorithm

In this work, we present a deconvolution approach to predict depth dose distributions from prompt gamma images obtained by a Compton camera. Based on the assumption that the one-dimensional PG distribution can be derived from the convolution of the depth dose distribution and a filter kernel, an approximation function of Bragg curve and an evolutionary algorithm (EA) are introduced to achieve the inverse process and estimate the depth dose distribution from a measured PG distribution. Filter kernels used in the approach are obtained by the same evolutionary algorithm with predictions of the PG and depth dose distributions. The applicability of this approach is demonstrated for the monoenergetic proton irradiation of several homogeneous phantoms made of different tissue materials, using Monte-Carlo simulated data. The distal falloff of depth dose distribution is reconstructed within less than 1 mm deviation, showing its potential for real-time dose estimation in future.