Unified 3D MRI Representations via Sequence-Invariant Contrastive Learning

2025-01-21

AI Summary

To address the scarcity of 3D MRI data, the inability of 2D models to capture volumetric anatomical structure, and the poor generalizability of conventional self-supervised methods due to sensitivity to acquisition protocols (e.g., contrast weighting), this paper proposes the first sequence-agnostic self-supervised learning framework for quantitative MRI (qMRI). Leveraging only a single qMRI scan, our method synthesizes multi-contrast images and enforces feature consistency to disentangle anatomical structure from protocol-specific imaging factors, thereby guiding 3D CNNs to learn robust, anatomy-driven representations. Crucially, we embed the qMRI physical forward model into a contrastive learning paradigm to enable unsupervised representation regularization. Evaluated on IXI brain segmentation, ARC stroke lesion segmentation, and MRI denoising, our approach achieves +8.3% Dice and +4.2 dB PSNR gains under low-data regimes and demonstrates significantly improved cross-site generalization. Code and models are fully open-sourced.

Technology Category

• Machine Learning: Unsupervised & Self-Supervised Learning
• Computer Vision: Segmentation
• Knowledge Representation and Reasoning: Qualitative Reasoning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Web Mining and Content Analysis: Large pretrained models with web data

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. This yields a robust 3D encoder that performs strongly across varied 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). Our model also generalises effectively to unseen sites, demonstrating potential for more scalable and clinically reliable volumetric analysis. All code and trained models are publicly available.

Problem

• Self-supervised Learning
• 3D MRI
• Deep Learning

Innovation

• Quantitative MRI
• Self-supervised Learning
• Consistent Feature Representation