Unified 3D MRI Representations via Sequence-Invariant Contrastive Learning
Abstract. Self-supervised deep learning has accelerated 2D natural image analysis but remains difficult to translate into 3D MRI, where data are scarce and pre-trained 2D backbones cannot capture volumetric context. We present a sequence-invariant self-supervised framework leveraging quantitative MRI (qMRI). By simulating multiple MRI contrasts from a single 3D qMRI scan and enforcing consistent representations across these contrasts, we learn anatomy-centric rather than sequence-specific features. The result is a single 3D encoder that excels across tasks and protocols. Experiments on healthy brain segmentation (IXI), stroke lesion segmentation (ARC), and MRI denoising show significant gains over baseline SSL approaches, especially in low-data settings (up to +8.3% Dice, +4.2 dB PSNR). It also generalises to unseen sites, supporting scalable clinical use. Code and trained models are publicly available.
Year 2025Citation year 2026Kind workshop paperVenue Simulation and Synthesis in Medical ImagingSeries Lecture Notes in Computer ScienceVolume 16085Pages 63–74DOI 10.1007/978-3-032-05573-6_7
Materials
Contrast-squared implementation
Sequence-invariant pre-training and transfer experiments for 3D segmentation and denoising.
The public Hugging Face collection associated with the SASHIMI study.
Task-specific and low-data comparisons; interpret the reported gains within their respective datasets and evaluation settings.
@inproceedings{chalcroft2026unified,
author = {Chalcroft, Liam and Crinion, Jenny and Price, Cathy J. and Ashburner, John},
title = {{Unified 3D MRI Representations via Sequence-Invariant Contrastive Learning}},
booktitle = {Simulation and Synthesis in Medical Imaging},
year = {2026},
series = {Lecture Notes in Computer Science},
volume = {16085},
pages = {63--74},
publisher = {Springer},
doi = {10.1007/978-3-032-05573-6_7},
url = {https://link.springer.com/chapter/10.1007/978-3-032-05573-6_7}
}