标题： 单次屏气磁共振胆管胰管造影（MRCP）重建的零样本自监督学习 显示英文标题

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

摘要： 目的：研究将零样本自监督学习重建应用于减少磁共振胆管胰管成像（MRCP）的屏气时间的可行性。 方法：使用非相干k空间采样模式在3T扫描仪上从11名健康志愿者中获取屏气MRCP，导致屏气持续时间为14秒。 我们评估了零样本重建的屏气MRCP与在平均338秒内获取的呼吸触发MRCP的并行成像以及屏气MRCP的压缩感知重建。 为了解决零样本训练的长时间计算问题，我们使用了一种利用预训练网络来减少训练期间反向传播深度的训练方法。 结果：与压缩感知重建相比，零样本学习重建显著提高了视觉图像质量，特别是在信噪比和胆管分界方面，并达到了与成功呼吸触发采集的常规呼吸模式相当的质量水平。 浅层训练提供了几乎等效的重建性能，训练时间为11分钟，相比之下传统零样本训练需要271分钟。 结论：零样本学习在减少屏气时间的情况下实现了高保真的MRCP重建，而浅层训练为转化为时间受限的临床工作流程提供了实用的解决方案。 显示英文摘要

摘要： 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.