Zero-shot self-supervised learning of single breath-hold magnetic resonance cholangiopancreatography (MRCP) reconstruction

2025-08-08

AI Summary

Long breath-hold durations and suboptimal image quality in magnetic resonance cholangiopancreatography (MRCP) hinder clinical utility. Method: We propose a zero-shot self-supervised reconstruction framework tailored for single-breath-hold MRCP, integrating compressed sensing, parallel imaging, and incoherent k-space sampling. A shallow-training strategy is introduced: deep network parameters are frozen using a pretrained model, while only shallow-layer weights are optimized—significantly reducing backpropagation depth. Contribution/Results: This approach reduces per-scan zero-shot training time from 271 to 11 minutes, enabling the first clinically feasible real-time zero-shot MRCP reconstruction. Quantitative and qualitative evaluations demonstrate that our method achieves signal-to-noise ratio (SNR) and biliary duct visibility comparable to respiratory-triggered gold-standard acquisitions, and substantially outperforms conventional compressed sensing reconstruction. The framework establishes a new paradigm for rapid, high-fidelity, registration-free clinical MRCP.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multimodal Learning
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Purpose: To investigate the feasibility of applying zero-shot self-supervised learning reconstruction to reduce breath-hold times in magnetic resonance cholangiopancreatography (MRCP). Methods: Breath-hold MRCP was acquired from 11 healthy volunteers on a 3T scanner using an incoherent k-space sampling pattern leading to a breath-hold duration of 14s. We evaluated zero-shot reconstruction of breath-hold MRCP against parallel imaging of respiratory-triggered MRCP acquired in 338s on average and compressed sensing reconstruction of breath-hold MRCP. To address the long computation times of zero-shot trainings, we used a training approach that leverages a pretrained network to reduce backpropagation depth during training. Results: Zero-shot learning reconstruction significantly improved visual image quality compared to compressed sensing reconstruction, particularly in terms of signal-to-noise ratio and ductal delineation, and reached a level of quality comparable to that of successful respiratory-triggered acquisitions with regular breathing patterns. Shallow training provided nearly equivalent reconstruction performance with a training time of 11 minutes in comparison to 271 minutes for a conventional zero-shot training. Conclusion: Zero-shot learning delivers high-fidelity MRCP reconstructions with reduced breath-hold times, and shallow training offers a practical solution for translation to time-constrained clinical workflows.

Problem

• Reduce breath-hold time in MRCP using zero-shot self-supervised learning
• Improve image quality compared to compressed sensing reconstruction
• Decrease training time for zero-shot learning in clinical workflows

Innovation

• Zero-shot self-supervised learning for MRCP
• Shallow training reduces computation time
• Pretrained network enhances reconstruction efficiency