Deep Learning with Electrocardiograms

Electrocardiograms (ECGs) are valuable medical tests, due in part to the fact that they are noninvasive and able to detect a wide variety of cardiovascular conditions, including sleep apnea. We show how deep learning can be used to denoise and diagnose ECGs. In particular, we use convolutional block attention to create two models. The first is a denoising autoencoder that can effectively remove noise from ECGs, and the second is a convolutional model that is capable of diagnosing whether or not a patient has COVID-19 with 99.00% accuracy. In introducing the models, we also give a comprehensive overview of convolutional layers, convolutional block attention, and denoising autoencoders.