Deep learning based registration using spatial gradients and noisy segmentation labels

TL;DR

This paper presents a deep learning approach for medical image registration that uses a symmetric formulation and leverages diverse datasets with noisy segmentation labels, achieving competitive results in the Learn2Reg challenge.

Contribution

The work introduces a symmetric registration model and integrates multiple datasets with noisy labels to improve registration accuracy.

Findings

Achieved a mean dice score of 0.64 and 0.85 on two tasks.

Secured third place in the Learn2Reg 2020 challenge.

Code and models are publicly available.

Abstract

Image registration is one of the most challenging problems in medical image analysis. In the recent years, deep learning based approaches became quite popular, providing fast and performing registration strategies. In this short paper, we summarise our work presented on Learn2Reg challenge 2020. The main contributions of our work rely on (i) a symmetric formulation, predicting the transformations from source to target and from target to source simultaneously, enforcing the trained representations to be similar and (ii) integration of variety of publicly available datasets used both for pretraining and for augmenting segmentation labels. Our method reports a mean dice of $0.64$ for task 3 and $0.85$ for task 4 on the test sets, taking third place on the challenge. Our code and models are publicly available at https://github.com/TheoEst/abdominal_registration and \https://github.com/TheoEst/hippocampus_registration.