This article explores the problem of analyzing signals obtained ex vivo from rat hearts using a multi-electron array. Existing analytical methods have significant drawbacks, including low efficiency on noisy or non-stationary data and poor scalability in high-dimensional spaces. This article describes the application of the developed software package to electrogram segmentation to identify moments of cardiac tissue activation, enabling the assessment of coronary blood flow, myocardial metabolism, and cardiac contractile function under the influence of various drugs. Neural network architectures such as UNet, its modification UNet++, and SegNet were developed, adapted, and analyzed as basic signal segmentation methods. Training was conducted using a proprietary dataset of signals in which an expert marked the areas of suspected activation moments. The dataset itself consists of signals recorded from 64 electrodes on the surface of ten rat hearts, with durations ranging from 2 to 10 min.

The problem of studying, treating and preventing cardiovascular diseases is one of the most important for the modern world. Advanced technologies allow recording a huge amount of data, thereby increasing the completeness of the picture of biological processes occurring in the heart. The need to automate the analysis of experimental and medical data in the field of heart research is growing significantly. A promising approach is the use of artificial intelligence-based methods for analyzing data recorded during the study of the bioelectrical activity of the heart. The presented work describes the creation of a software package for analyzing epicardial electrograms of isolated rat hearts. The main part of the package is a segmenting neural network for localizing myocardial activation moments based on the UNet architecture. The choice of this architecture is due to its effectiveness in segmenting multidimensional data. Within the framework of the article, the basic architecture of the UNet network was modified to process one-dimensional signals. The purpose of the work was to create software that includes the following functionality: creating a data set for training, validation and testing, training a model, creating and editing markup. The program will automatically localize activation moments to speed up and improve the efficiency of the study. The developed complex is a promising solution for automating the analysis of multichannel electrograms of the heart and its structures using a segmenting neural network based on the UNet architecture and pre- and post-processing algorithms.

The article presents the results of applying artificial intelligence algorithms to analyze multichannel electrograms of bioelectrical activity registered by microelectrode arrays from myocardium of isolated rat hearts under the influence of L-carnitine and adrenaline. Localization of activation times on electrograms was performed using a new neural network model based on the U-Net architecture adapted for one-dimensional signal segmentation. In the organism, L-carnitine is necessary for the oxidation of fatty acids. But the effects of its supplemental intake on the cardiovascular system are still poorly understood. A decrease in heart rate, myocardial excitation conduction velocity, intensity of coronary blood flow and physiologic cardiac response to adrenaline action ex vivo caused by L-carnitine were shown. The methodological approach, applying a neural network to analyze data from microelectrode arrays, tested in this research can be translated into clinical practice for electrograms obtained by electrophysiology study of human heart. The neural network model can automate the analysis of data from microelectrode arrays to study the frequency and spatiotemporal characteristics of myocardial bioelectrical activity

The using of modern electrophysiological methods during scientific research implies processing of huge data sets which becomes a significant problem. Progress in the development of equipment for registration of multiple local field potentials from the tissue surface contributes to the study of spatio-temporal characteristics of myocardial bioelectrical activity at a more detailed level. But it also raises the need for automation of data analysis, which can be accomplished by algorithms based on machine learning. New scientific approaches can contribute to the discovery of new facts about substances widely used but with under-researched efficacy. The paper presents the results of epicardial mapping by flexible microelectrode arrays of myocardium of isolated rat heart perfused with a solution containing L-carnitine, nutritional supplement that may be recommended for people suffering from cardiovascular disease. Electrograms from microelectrode arrays were analyzed using a neural network model based on U-Net architecture adapted for segmentation of one-dimensional signal. It was shown that the presence of L-carnitine in the perfusion solution caused a decrease in heart rate, myocardial excitation conduction velocity, coronary blood flow intensity of the isolated rat heart and suppressed physiological responses of the heart to adrenaline stimulation. ### Competing Interest Statement The authors have declared no competing interest.

The problem of explainability of artificial intelligence is extremely relevant at the present time. The most acute need for this tool is felt in the areas of practical application of intelligent systems. One of these areas is medicine. Extracting knowledge and understanding the reason for the model's prediction will allow us to provide in a human-readable form the main factors used by the model to predict. The article will consider a method for highlighting areas of the electrocardiogram associated with signs of cardiovascular diseases using the transformer machine learning architecture. Visualization of the attention of the transformer to the ECG signal allows you to highlight the segments of the electrocardiogram, which may show signs of various cardiovascular diseases. For the diagnosis of inferior myocardial infarction, the algorithm for labeling the data extracted from the trained model identified Q and R waves, for anterior septal myocardial infarction, Q and T waves, as well as the ST segment, and for left ventricular hypertrophy, the ST segment. Areas of attraction of "attention" of the neural network are shown graphically and displayed in the article.

Diagnosis by electrocardiogram (ECG) is an extremely urgent and important task, the quality, timeliness and speed of which people’s lives and health depend on. To date, a large number of researchers treat neural networks as a panacea, hoping that any task can be solved quickly and without problems. Often, this approach does not lead to the best results. The article explores the use of deep learning as a universal tool for solving the problem in determining pathological ECG signals with markers of myocardial infarction, hypertrophy, conduction disturbances, and changes in ST segment morphology. During the experiments, the positive impact of using thresholding and replacements to increase the predictive ability of the network, the use of various ensembles on trained deep learning models was established. The addition of artificial models also improves the classifying ability of ensembles. Returning a random number in the absence of a single mode also makes it possible to increase the accuracy of the ensemble.

The paper presents the results of investigation of efficiency of the parallelization of a global optimum search algorithm. Two stages of solving the problem were distinguished: the search of a rough approximation to the global solution and its local refining. The novelty of the parallel algorithm considered in the present paper consists in the parallelization of the whole computational process, both at the global search stage and at the one of the local refining of the solution found. The theoretical estimates of the algorithm parallelization efficiency are presented. The computational experiments were carried out. The results of the ones confirmed the theoretical conclusions.

It is known that the most dangerous predictors of sudden cardiac death is the appearance in a person of signs of atrial and ventricular fibrillation. At the same time, this type of rhythm can rarely be detected during a quick screening examination, which greatly complicates the diagnosis and, as a result, the choice of effective treatment. Within the framework of this article, the experience of using the most popular architectures of convolutional neural networks adapted for the analysis of the electrocardiogram signal is presented to localize areas of sinus rhythm and fibrillation. As the networks under study, such architectures as ResNet, DenseNet and XceptionNet were considered. Each of their architectures showed good results, but the DenseNet network showed the best results. Due to a number of advantages described in the article, the DenseNet network was chosen. It should be noted that the chosen architecture of the neural network, after a number of necessary modifications and improvements, will be integrated into the existing diagnostic complex for cardiovascular diseases, developed on the basis of Lobachevsky University.