Abstract
In modern medicine, the problem of detecting and classifying life-threatening arrhythmias based on ECG data remains relevant and critically important for continuous patient monitoring. This study is dedicated to developing a method for the automatic classification of six classes of dangerous arrhythmias using short ECG segments of 2 seconds duration. Existing methods for detecting dangerous arrhythmias require additional improvements to ensure high accuracy and efficiency. The goal of this research is to develop an effective method for the classification of dangerous arrhythmias to facilitate timely medical intervention. A unique method is proposed, based on transforming ECG signals into scalograms using continuous wavelet transformation. For arrhythmia classification, the AlexNet neural network is employed. The study utilizes data from the PhysioNet database and synthesized ECG data using the SMOTE method. Experimental investigations demonstrated a high accuracy of the proposed method, with an average accuracy of 98.7% for all arrhythmia classes, surpassing previously achieved maximum estimates by other researchers (93.18%). The study has been successfully completed, showcasing scientific novelty and practical significance of the results. The proposed method not only improved existing accuracy estimates but also emphasized the potential of using scalograms and neural networks for recognizing dangerous arrhythmias from ECG data. This opens new horizons for continuous monitoring and timely medical intervention, enhancing the quality of patient care.