Chen Sully F, Guo Zhicheng, Ding Cheng, Hu Xiao, Rudin Cynthia
Duke University School of Medicine, Durham, NC, USA.
Department of Electrical and Computer Engineering, Duke University, Durham, NC, USA.
Nat Mach Intell. 2024 Oct;6(10):1132-1144. doi: 10.1038/s42256-024-00898-4. Epub 2024 Sep 18.
Rapid, reliable and accurate interpretation of medical time series signals is crucial for high-stakes clinical decision-making. Deep learning methods offered unprecedented performance in medical signal processing but at a cost: they were compute intensive and lacked interpretability. We propose sparse mixture of learned kernels (SMoLK), an interpretable architecture for medical time series processing. SMoLK learns a set of lightweight flexible kernels that form a single-layer sparse neural network, providing not only interpretability but also efficiency, robustness and generalization to unseen data distributions. We introduce parameter reduction techniques to reduce the size of SMoLK networks and maintain performance. We test SMoLK on two important tasks common to many consumer wearables: photoplethysmography artefact detection and atrial fibrillation detection from single-lead electrocardiograms. We find that SMoLK matches the performance of models orders of magnitude larger. It is particularly suited for real-time applications using low-power devices, and its interpretability benefits high-stakes situations.
对医学时间序列信号进行快速、可靠且准确的解读对于高风险临床决策至关重要。深度学习方法在医学信号处理中展现出了前所未有的性能，但代价是：它们计算量大且缺乏可解释性。我们提出了学习内核的稀疏混合（SMoLK），这是一种用于医学时间序列处理的可解释架构。SMoLK学习一组轻量级的灵活内核，这些内核构成一个单层稀疏神经网络，不仅提供可解释性，还具备效率、鲁棒性以及对未见数据分布的泛化能力。我们引入参数缩减技术来减小SMoLK网络的规模并保持性能。我们在许多消费级可穿戴设备常见的两项重要任务上测试了SMoLK：光电容积脉搏波描记术伪像检测和单导联心电图房颤检测。我们发现SMoLK与规模大几个数量级的模型性能相当。它特别适用于使用低功耗设备的实时应用，其可解释性在高风险情况下具有优势。
