Utilizing long short-term memory neural network for effective prediction of electrocardiosignals and pathology identification

Currently, neural networks (NN) are becoming an increasingly common tool in the field of diagnosis of diseases of the cardiovascular system. These architectures demonstrate high accuracy in analyzing complex data such as electrocardiosignals (ECS). Among various NN types, long-term memory (LSTM) stands out as a powerful method for analyzing time series. This technology has the ability to remember long-term dependencies in data, which makes it particularly useful for processing sequential data, including ECS. However, using the standard electrocardiography (ECG) method, it is impossible to obtain complete information about the stages of development of cardiovascular pathologies. The current trend towards increasing the informative value of the ECS has led to the development of a new method of ultra-high resolution electrocardiography (UHR ECG). This method significantly improves the detection of pathologies in the cardiovascular system in various areas of ECG treatment and also allows the detection of early markers of acute myocardial ischemia. UHR ECG provides an opportunity for a more detailed analysis of cardiac signals, which can be crucial in the early diagnosis and treatment of cardiovascular diseases. In this work, a six-layer LSTM was used to predict the shape of ECS obtained using the UHR ECG method in experimental rats. The experiments were aimed at modeling acute myocardial ischemia in order to identify trends in the development of markers of this pathology. The results of the study may contribute to the creation of more effective methods for diagnosing and monitoring cardiovascular diseases.