Computer-aided method for paroxysmal atrial fibrillation detection based on ECG signals.

Paroxysmal atrial fibrillation (PAF) is a common cardiac arrhythmia characterized by irregular contractions of the atria. In this paper, we propose a new classification method based on surface 12-lead electrocardiography (ECG) signals by using statistical analysis and deep learning models for PAF detection. First, we extract signals with a frequency range of 1-40Hz from preprocessed ECG and divide them into three frequency bands. Secondly, we calculate the phase locking value (PLV) between the twelve channels of the ECG signals and generate a correlation matrix. After bilinear interpolation and binarization, the PLV correlation matrix is obtained as a feature and plotted as a feature map. Finally, we input feature maps of different frequency bands into the Rstnet18 convolutional neural network, and achieving a promising performance.