Diagnosis of sinus rhythm and atrial fibrillation using artificial intelligence

The electrocardiogram (ECG) is the most used biological signal recording in clinical medicine. The ECG signal is a graph of the electrical activity of the heart obtained from the surface of the body, most often non-invasively, using electrodes. In the early days of electrocardiography, the doctor had to look at a graph written on a piece of paper, recognizing possible pathologies with his eyes, which often led to errors in the diagnosis. Today, there are many decision support systems based on complex algorithms that help the doctor in the search for artifacts that establish both the type of pathology and the localization of its markers in the signal. However, there are a large number of diagnoses, the detection of which by the developed algorithms is not effective. Moreover, such algorithms, rarely, but make a mistake. Experts see a promising solution for eliminating existing shortcomings in expert systems in the application of artificial intelligence methods that have shown their effectiveness in a variety of applied tasks. Within the framework of this article, the use of neural networks for solving diagnostic problems is considered UNct, adapted for processing a one-dimensional ECG signal, was chosen as the basic architecture of the neural network. Among a wide range of conditions of the human cardiovascular system, the main attention was focused on the detection of long-duration signal sections classified by specialists as complexes with the prevalence of sinus rhythm and atrial fibrillation (atrial fibrillation). It should be noted that the neural network considered in the framework of the work, after the necessary improvements, will be integrated into the existing diagnostic complex „CardioLighthouse", developed on the basis of Lobachevsky University.