CBCT-guided adaptive radiotherapy using self-supervised sequential domain adaptation with uncertainty estimation

Nima Ebadi, Ruiqi Li, Arun Das, Arkajyoti Roy, Papanikolaou Nikos, Peyman Najafirad

Abstract: Highlights•A sequence transduction DL model to learn a predictive segmentation of tumors given the weekly CBCT images of a cancer patient under radiotherapy treatment.•A self-supervised domain adaptation (SDA) technique to align the representations from publicly available CT data with the in-house CT pre-treatment images and CBCT mid-treatment images.•A spatiotemporal uncertainty estimation technique to improve reliability and the quality of detection of erroneous regions.•An end-to-end computer-vision based solution for reliable adaptive radiotherapy treatment planning workflow.