Automatická segmentácia vybraných biologických signálov na základe strojového učenia

Abstract

The segmentation of ECG and the interpretation of its waves play a significant role in the analysis of cardiovascular diseases, which is why many automatic approaches using advanced machine learning techniques for ECG wave detection are currently being studied. The diploma thesis focuses on designing a neural network model based on deep learning for the automatic detection of individual ECG waves, emphasizing testing various hyperparameter settings of the model concerning overall performance and robustness. The theoretical part of the thesis includes basic principles of supervised learning and an overview of machine learning approaches to the segmentation of 1D biosignals. In the practical part, a segmentation model of the neural network was designed and subsequently tested for various hyperparameter settings. The practical part also includes creating a software interface to train the segmentation model on the user's data.