Coronary heart disease is one of the main causes of death in humans. Despite this, modern electrocardiography (ECG) methods are limited in obtaining complete information about the progression of pathology due to insufficient  throughput. To eliminate this limitation, a new approach known as ultra-high resolution electrocardiography (UHR ECG) is being developed to detect abnormal changes in the cardiovascular system in areas of electrocardiosignal processing (ECS), which are usually perceived as artifacts. However, the expansion of the amplitude and frequency range complicates the task of analyzing the data obtained, since many traditional ECS analysis methods are  ineffective when applied to UHR ECG. This study demonstrates the effectiveness of using the fifteen-layer convolutional neural network (CNN) AlexNet to solve the problem of classifying ECS obtained using the UHR ECS  methodology. An extensive data set of labeled ECG recordings obtained using the UHR ECG method on Vistar series experimental rats during experiments on modeling acute myocardial ischemia was used to train and evaluate the  quality of predictions of the CNN model.