Fast deformable image registration uncertainty estimation for contour propagation in daily adaptive proton therapy

Andreas Smolders, Florian Amstutz, Ye Zhang, Damien Charles Weber, Tony Lomax, Francesca Albertini

Keywords: Deformable Image Registration, Uncertainty, Proton Therapy, Adaptive Planning, Deep Learning

TL;DR: Using deep learning to speed up uncertainty quantification of deformable image registration

Abstract: In daily adaptive proton therapy, deformable image registration (DIR) can be used to propagate manually delineated contours from a reference CT to the daily CT for plan reoptimization. However, the ill-posedness of DIR implies uncertainty on the DIR hyperparameters, which results in uncertainty in the displacement field. In this work, a fast deep learning method is developed to predict the uncertainty associated with a DIR result without the need for Monte-Carlo (MC) sampling. It is shown that this results in a significant time reduction compared to MC whilst leading to similar probabilistic contours.

Registration: I acknowledge that acceptance of this work at MIDL requires at least one of the authors to register and present the work during the conference.

Authorship: I confirm that I am the author of this work and that it has not been submitted to another publication before.

Paper Type: novel methodological ideas without extensive validation

Primary Subject Area: Image Registration

Secondary Subject Area: Uncertainty Estimation

Confidentiality And Author Instructions: I read the call for papers and author instructions. I acknowledge that exceeding the page limit and/or altering the latex template can result in desk rejection.