Porovnání elektrokardiografických a vektorkardiografických svodových systémů pro automatizovanou detekci infarktu myokardu

Abstract

Electrocardiography (ECG) is currently the most widely used method for diagnosing the electrical activity of the heart in clinical practice. Another method is vectorcardiography (VCG), which, although rarely used in clinical practice, achieves higher sensitivity in detecting pathological recordings compared to ECG in research and scientific publications. This diploma thesis is divided into two parts. The first, theoretical part deals with the cardiovascular system and the cardiac conduction system— particularly the action potentials of cardiac cells. It also discusses the issues and differences between ECG and VCG lead systems, especially in terms of electrode placement. In the case of VCG, the historical development of lead systems is reviewed, including the Frank lead system, which is used today. A key component of the theoretical part is a literature review of extracted features that can be obtained from both ECG and VCG recordings. The outcome of this review serves as the main input for the practical part. The practical part describes the input data, including the dataset used and its structure. The data were then filtered using filters selected based on the conducted literature review. A crucial step was the development of algorithms for computing a representative beat (median cycle) and for detecting the onset and offset of significant waves and complexes in ECG and VCG signals. The main focus was on extracting relevant features for the automated detection of myocardial infarction (MI). These features were subsequently subjected to relevance analysis using the mRMR method, which resulted in the reduction of certain features. For the classification analysis, a total of five classification methods were selected, and their performance, reliability, and efficiency were evaluated using metrics such as sensitivity, specificity, accuracy, positive predictive value, negative predictive value, and F1 score. In the conclusion of this thesis, comparisons were made between individual leads, lead systems, measurement methods, and different dimensions across ECG and VCG, presented in the form of heatmaps and ROC curves along with the calculated AUC parameter.