A Concealed Accessory Pathway(CAP)is a heart condition characterized by rapid heartbeat,palpitations,and shortness of breath.However,clinicians cannot currently diagnose CAPs via sinus rhythm Electrocardiogram(ECG).Based upon clinical cases,the present study establishes a dataset containing the preoperative sinus rhythm ECG data of healthy subjects and patients with the CAP,and proposes a Convolutional Neural Network(CNN)based on ResNet26,referred to as CAPNet,to automatically identify and predict the CAP using via the sinus rhythm ECG.An Initialization Block(IB)is established to improve the nonlinear expression of the model.An asymmetric convolution block is introduced into the bottleneck residual block to better capture the horizontal and vertical directional information of the ECG features,allowing the module to enrich the feature space.Furthermore,an attention mechanism is used to enhance the attention of the model to the key band region in the ECG.The results demonstrate that CAPNet model outperforms CNN models in predicting the CAP.The common indicators of CAPNet model including the F1 score,accuracy,sensitivity,and precision increase by 2.41,0.89,4.34,and 0.47 percentage points,respectively.These experimental results validate the effectiveness and superiority of the CAPNet model.
维普
万方数据