Skip to main content
QUICK REVIEW

[论文解读] Development of Interpretable Machine Learning Models to Detect Arrhythmia based on ECG Data

Shourya Verma|arXiv (Cornell University)|May 5, 2022
ECG Monitoring and Analysis被引用 6
一句话总结

本论文基于MIT-BIH心律失常数据集中的ECG数据,开发了可解释的深度学习模型——CNN与LSTM,以高精度检测心脏心律失常。研究评估了多种可解释性技术,发现梯度加权类激活映射(Grad-CAM)在局部可解释性方面最为有效,能一致地突出QRS波作为分类决策的关键特征。

ABSTRACT

The analysis of electrocardiogram (ECG) signals can be time consuming as it is performed manually by cardiologists. Therefore, automation through machine learning (ML) classification is being increasingly proposed which would allow ML models to learn the features of a heartbeat and detect abnormalities. The lack of interpretability hinders the application of Deep Learning in healthcare. Through interpretability of these models, we would understand how a machine learning algorithm makes its decisions and what patterns are being followed for classification. This thesis builds Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) classifiers based on state-of-the-art models and compares their performance and interpretability to shallow classifiers. Here, both global and local interpretability methods are exploited to understand the interaction between dependent and independent variables across the entire dataset and to examine model decisions in each sample, respectively. Partial Dependence Plots, Shapley Additive Explanations, Permutation Feature Importance, and Gradient Weighted Class Activation Maps (Grad-Cam) are the four interpretability techniques implemented on time-series ML models classifying ECG rhythms. In particular, we exploit Grad-Cam, which is a local interpretability technique and examine whether its interpretability varies between correctly and incorrectly classified ECG beats within each class. Furthermore, the classifiers are evaluated using K-Fold cross-validation and Leave Groups Out techniques, and we use non-parametric statistical testing to examine whether differences are significant. It was found that Grad-CAM was the most effective interpretability technique at explaining predictions of proposed CNN and LSTM models. We concluded that all high performing classifiers looked at the QRS complex of the ECG rhythm when making predictions.

研究动机与目标

  • 开发高精度、可解释的机器学习模型,用于基于ECG数据的心律失常检测。
  • 通过整合全局与局部可解释性技术,解决深度学习模型‘黑箱’特性带来的临床局限性。
  • 评估并比较可解释性方法(PDP、SHAP、PFI与Grad-CAM)在时间序列ECG数据上的有效性。
  • 使用K折交叉验证与留组交叉验证,并结合非参数统计检验,验证模型性能。
  • 通过验证模型决策与ECG形态学既定医学知识的一致性,确保临床相关性。

提出的方法

  • 在MIT-BIH心律失常数据集的单导联ECG搏动上训练卷积神经网络(CNN)与长短期记忆网络(LSTM)模型。
  • 通过将ECG搏动分割为11个时间窗口以降低维度,并提升与可解释性工具的兼容性。
  • 应用四种可解释性技术:部分依赖图(PDP)、加性可解释性Shapley值(SHAP)、置换特征重要性(PFI)与梯度加权类激活映射(Grad-CAM)。
  • 采用K折交叉验证与留组交叉验证评估模型在不同受试者间的泛化能力。
  • 执行非参数统计检验(如Wilcoxon符号秩检验)以验证不同模型间性能差异的显著性。
  • 可视化Grad-CAM热力图,分析模型对每个ECG搏动的关注区域,并对比正确与错误分类预测的差异。

实验结果

研究问题

  • RQ1在MIT-BIH数据集的ECG搏动分类中,哪种机器学习模型——CNN、LSTM或浅层分类器——能达到最高准确率?
  • RQ2全局可解释性方法(PDP、SHAP、PFI)与局部方法(Grad-CAM)在解释时间序列ECG数据的模型预测方面,哪种更有效?
  • RQ3在每类心律失常中,Grad-CAM对正确分类与错误分类ECG搏动的定位是否存在差异?
  • RQ4根据医学知识,模型在分类决策中在多大程度上依赖QRS波?
  • RQ5Grad-CAM与PFI等可解释性技术是否能在不同模型与数据划分中揭示一致且具有临床意义的模式?

主要发现

  • CNN与LSTM模型表现最佳,K折交叉验证准确率分别为94.1%与94.0%,留组交叉验证准确率分别为98.7%与97.0%。
  • Grad-CAM是最有效的可解释性技术,能生成清晰、局部化的注意力热力图,一致地突出QRS波作为主要决策特征。
  • 所有高性能模型均将注意力集中于QRS波,证实其与心律失常诊断的临床认知一致。
  • PFI提供了全局、与模型无关的特征重要性洞察,但缺乏样本级或类别级的可解释性,限制了其在单个搏动分析中的实用性。
  • SHAP与PDP效果较差,SHAP因计算成本过高,PDP则因信息量不足,尤其在高时间分辨率的时间序列数据中表现不佳。
  • 本研究证实,可解释性技术能有效揭示深度学习模型在ECG分类中的运作机制,提升信任度并增强临床适用性。

更好的研究,从现在开始

从阅读论文到最终审阅,大幅缩短您的研究时间。

无需绑定信用卡

本解读由 AI 生成,并经人工编辑审核。