| Resource | Link |
|---|---|
| 📄 Paper | arXiv:2501.12057 |
| 🎮 Demo | Data Augmentation Demo |
| 🤗 Models | Download weights |
This repository contains the official implementation of Unified 3D MRI Representations via Sequence-Invariant Contrastive Learning. Our method learns robust, sequence-agnostic representations from multi-site, multi-sequence MRI data using self-supervised learning, enabling improved performance on downstream tasks like segmentation and denoising.
Explore our data augmentation pipeline in the Data Demo Notebook, which visualizes:
- Standard geometric/intensity augmentations
- Sequence simulation via Bloch equations
- Paired multi-contrast views
@article{chalcroft2025unified,
title={Unified 3D MRI Representations via Sequence-Invariant Contrastive Learning},
author={Chalcroft, Liam and Crinion, Jenny and Price, Cathy J and Ashburner, John},
journal={arXiv preprint arXiv:2501.12057},
year={2025}
}