ECG Characteristic Wave Detection Based on Deep Recursive Long Short-Term Memory

Detection of characteristic waves plays a very important role in electrocardiography (ECG) analysis. Long short-term memory (LSTM) has been successfully used in sequential data analysis. Many studies focused on the detection of the obvious electrocardiographic complex, which consists of the Q, R, and S Waves (QRS complex), rather than on the detection of single P or T wave. This study aimed to propose an LSTM-based method to detect ECG characteristic waves, including the peak of P wave, the onset and offset of QRS complex, and the peak of T wave. The proposed method used two LSTM layers and two fully connected layers to delineate all waves (including P, QRS, and T). The proposal method achieved good performance in terms of characteristic point detection accuracy in the public QT ilataset. The source codes for reproducible research were released at https://github.com/shipengai/ECG-Segment-LSTM.