Diagnosis of atrial fibrillation using self-complementary attentional convolutional neural network.

•This study proposes a self-complementary attention mechanism (SCNet). A self-complementary attentional convolutional neural network (SCCNN) was designed using SCNet.•SCCNN can extract not only the shallow features of near-neighborhood sampling points and far-interval sampling points in ECG signals but also the deep channel and spatial features. This information is complemented by integrating the information of both features.•In addition, to extract information using a 2D convolutional template and increase the perceptual field of convolutional computation, we propose a Z-shaped signal reconstruction method to convert a 1D ECG sequence into a 2D ECG signal matrix.