Electronic Portal Imaging Devices Using Artificial Neural Networks

Abstract

The aim of this work was to use the Artificial Neural Network (ANN) in External Beam Radiation Therapy (EBRT), especially for pre-treatment patient-specific quality assurance of Conformational Radiation Therapy (CRT) and Intensity-Modulated Radiation Therapy (IMRT) using Electronic Portal Imaging Device (EPID). The EPIDs need frequent calibration and complex setting in order to be used with dedicated dosimetry software. The idea was to create a model with ANN algorithms allowing the reconstruction of the 2D dose distribution comparable with a corresponding Treatment Planning System (TPS) solution. The supervised ANN algorithms work with two phases—learning and recognition. Learning was performed using data sets regarding CRT and IMRT composed of 8 and 11 input/output respectively. To compare ANN predicted and planned results the global gamma index was used, obtaining a \( \gamma_{{(2\% ,2\,\text{mm})}} = 99.78\% \) and \( \gamma_{{(2\% ,2\,\text{mm})}} = 99.7\% \), respectively. This first work showed the capability of ANN to reconstruct the absorbed dose distribution based on EPID signals.