CM-UNet: A Self-Supervised Learning-Based Model for Coronary Artery Segmentation in X-Ray Angiography

2025-07-22

AI Summary

Coronary artery segmentation in X-ray angiography is clinically hindered by severe scarcity of annotated data, limiting model accuracy. To address this, we propose the first self-supervised learning framework for this task, introducing a contrastive masked autoencoding pretraining strategy integrated with a U-Net architecture. Leveraging abundant unlabeled angiograms, our method learns robust feature representations and subsequently fine-tunes the model using only 18 annotated images. Compared to a non-pretrained baseline, our approach reduces Dice score degradation under few-shot conditions from 46.5% to merely 15.2%, substantially alleviating annotation dependency. This work empirically validates the efficacy of self-supervised learning for coronary artery segmentation and establishes a transferable methodology for low-resource medical image segmentation. By enabling high-performance segmentation with minimal annotations, it holds promise for improving diagnostic efficiency and patient outcomes in clinical practice.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Unsupervised & Self-Supervised Learning
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Large pretrained models with web data
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Accurate segmentation of coronary arteries remains a significant challenge in clinical practice, hindering the ability to effectively diagnose and manage coronary artery disease. The lack of large, annotated datasets for model training exacerbates this issue, limiting the development of automated tools that could assist radiologists. To address this, we introduce CM-UNet, which leverages self-supervised pre-training on unannotated datasets and transfer learning on limited annotated data, enabling accurate disease detection while minimizing the need for extensive manual annotations. Fine-tuning CM-UNet with only 18 annotated images instead of 500 resulted in a 15.2% decrease in Dice score, compared to a 46.5% drop in baseline models without pre-training. This demonstrates that self-supervised learning can enhance segmentation performance and reduce dependence on large datasets. This is one of the first studies to highlight the importance of self-supervised learning in improving coronary artery segmentation from X-ray angiography, with potential implications for advancing diagnostic accuracy in clinical practice. By enhancing segmentation accuracy in X-ray angiography images, the proposed approach aims to improve clinical workflows, reduce radiologists' workload, and accelerate disease detection, ultimately contributing to better patient outcomes. The source code is publicly available at https://github.com/CamilleChallier/Contrastive-Masked-UNet.

Problem

• Accurate coronary artery segmentation in X-ray angiography
• Reducing reliance on large annotated datasets
• Improving disease detection and clinical workflow efficiency

Innovation

• Self-supervised pre-training on unannotated datasets
• Transfer learning with limited annotated data
• Fine-tuning with minimal annotated images