XCAT-CoMBAT training data

Abstract

The XCAT-CoMBAT data can be used to train deep neural networks in an open-source framework for AI-guided adaptive radiotherapy. This framework, called Voxelmap, seeks to learn patient-specific geometric correspondences between 3D internal organ motion and 2D images acquired during radiotherapy. By learning these correspondences, Voxelmap aids in delivering a therapeutic dose to targets while minimising exposure to the surrounding organs-at-risk; thereby reducing the toxic side-effects of radiotherapy. The XCAT data are used to train 3 separate networks designed to assist in x-ray guided treatments, while the CoMBAT data are used to train 2 networks designed to assist in MRI guided treatments. We hope that by making our code and data available to the broader academic community, this work will help catalyse advancements in next generation radiotherapy. (https://github.com/Image-X-Institute/Voxelmap)