基于模式识别和集成CNN-LSTM的阵发性房颤预测模型

杨萍,王丹,康子健,李童,付利华,余悦任

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

Ping YANG,Dan WANG,Zi-jian KAGN,Tong LI,Li-hua FU,Yue-ren YU

表 4 模型与相关工作在训练集和测试集随机划分模式下的比较

Tab.4 Comparison between proposed model and related studies under random partition mode of training and testing sets

文献 特征提取方法 t/min f/Hz ACC/% SEN/% SPE/% Boon等[4] HRV特征，GA 15 128 79.3 77.4 81.1 10 128 68.8 58.5 81.1 Narin等[5] HRV线性和非线性特征组合，GA 5 128 90.0 92.0 88.0 Sutton等[24] PSPR，模板匹配 1 8 82.1 100.0 73.6 Boon等[26] HRV特征，GAⅢ 5 128 87.7 86.8 88.7 Costin等[27] HRV特征, QRS复合波形态变异性 30 128 89.4 89.3 89.4 本研究方法 PSPR+CNN-LSTM+集成学习 1 8 91.3 82.2 95.8