Strategies for Robust Deep Learning Based Deformable Registration

TL;DR

This paper proposes a simple MIND feature transformation and ensembling strategy to enhance the robustness of deep learning-based deformable registration across different imaging contrasts and modalities.

Contribution

It introduces a novel robustness enhancement method using MIND features and ensembling, improving generalization in deformable registration tasks.

Findings

Improved registration robustness across diverse contrasts and modalities.

Ensembling strategy yields consistent performance gains.

Method outperforms baseline models in the LUMIR challenge.

Abstract

Deep learning based deformable registration methods have become popular in recent years. However, their ability to generalize beyond training data distribution can be poor, significantly hindering their usability. LUMIR brain registration challenge for Learn2Reg 2025 aims to advance the field by evaluating the performance of the registration on contrasts and modalities different from those included in the training set. Here we describe our submission to the challenge, which proposes a very simple idea for significantly improving robustness by transforming the images into MIND feature space before feeding them into the model. In addition, a special ensembling strategy is proposed that shows a small but consistent improvement.