A novel Mean Teacher framework for domain adaptive lung registration

Alexander Bigalke, Lasse Hansen, Mattias P Heinrich

Keywords: Registration, Domain adaptation, Mean Teacher

Paper Status: based on accepted/submitted conference paper (if accepted please deanonymise PDF)

Abstract: Recent deep learning-based registration models achieve excellent results but require plenty of labeled training data and suffer from domain shifts between training and test data. As a remedy, we present a novel method for domain adaptive image registration. We propose to reduce the domain shift through self-ensembling and embed a keypoint-based registration model into the Mean Teacher paradigm. We extend the Mean Teacher to the registration problem by 1) adapting the stochastic augmentation scheme and 2) combining learned feature extraction with differentiable optimization. This enables us to guide the learning process in the unlabeled target domain by enforcing consistent predictions of the learning student and the temporally averaged teacher model. We evaluate the method for exhale-to-inhale lung CT registration under two challenging adaptation scenarios. Our method consistently improves on the baseline model by 44%/47% while even matching the accuracy of models trained on target data. Source code is available at https://anonymous.4open.science/r/reg-da-mean-teacher.