Applying Deep Learning Techniques to Extract Diagnostic Information from ECG Images

This work is devoted to the implementation of an algorithm for extracting diagnostic information from ECG images using deep learning methods. The U-Net neural network architecture was chosen to search and segment the ECG signal area in the image. The training was conducted based on the developed data set. The values of the DICE coefficient were obtained to assess the reliability of the neural network architecture, which indicates the high accuracy of the proposed method for solving the problem.