CoRe: Joint Optimization with Contrastive Learning for Medical Image Registration

Overview

This repository contains the code and resources for the for joint optimization with Contrastive learning for medical image Registration (CoRe). CoRe integrates contrastive learning directly into the registration pipeline, enhancing the robustness of extracted features to anatomical variations and improving the overall reliability of the framework.

We evaluate the proposed approach on abdominal and thoracic anatomical regions in both intra-patient and inter-patient registration settings, achieving superior performance compared to competitive approaches.

Installation

To install the necessary dependencies, run the following command:

pip install -r requirements.txt

Usage

To use this project, follow these steps:

1. Set desirable settings in the file main.py.
2. Run main.py.