基于模式识别和集成CNN-LSTM的阵发性房颤预测模型

Prediction model of paroxysmal atrial fibrillation based on pattern recognition and ensemble CNN-LSTM

为了适用于长期心电监护和ICU等实时性、数据密集型应用场合,提出可在8 Hz采样频率的1 min心电图(ECG)片段上提前45 min预测阵发性房颤(PAF)发作的实时预测模型.采用概率符号化模式识别方法,在降采样后的ECG序列上提取出1 min窗口内的模式转移特征,降低模型的计算量和对存储空间的需求,确保实时预测的效果.提出卷积神经网络(CNN)和长短-期记忆网络(LSTM)的混合模型(CNN-LSTM),用于提取模式转移特征内隐含的局部空间特征和时间依赖特征.为了提升模型泛化能力,构建基于CNN-LSTM的集成分类器.采用Spark Streaming技术完成对ECG流式数据的读、写和计算,实现数据和模型之间的低延迟通信.所提模型在公开数据集上的准确率、灵敏度和特异度分别为91.26%、82.21%、95.79%.模型处理总延迟平均为2 s,满足实时PAF预测需求.

The real-time prediction model that can predict the onset of paroxysmal atrial fibrillation(PAF)45 min in advance on the one minute electrocardiogram(ECG)segment with 8 Hz sampling frequency was proposed,for realtime and data-intensive application scenarios such as long-term ECG monitoring and intensive care units(ICU).The probabilistic symbolic pattern recognition method was used to extract the pattern transition features within one minute window of down sampled ECG sequence,reducing the calculation complexity of the model and the demand for storage space,so as to ensure the effect of real-time prediction.A hybrid model(CNN-LSTM)of the convolutional neural network(CNN)and the long short-term memory(LSTM)was proposed to extract local spatial features and time-dependent features implied in pattern transition features.An ensemble classifier based on CNNLSTM was constructed to improve the generalization ability of the model.Spark Streaming technology was used to read,write and calculate ECG streaming data,and low latency communication between data and model was realized.The accuracy,sensitivity,and specificity of the proposed model were 91.26%,82.21%,and 95.79%respectively.The average delay of model processing was 2 s,which can meet the real-time PAF prediction demand.