Alternative Approach to Investigation of Cardiovascular Diseases: Analysis of ECG Signal with Robust Local Mean Decomposition

According to the World Health Organization, cardiovascular diseases (CVDs) are the group of diseases that cause the most deaths worldwide. Among the types of CVD, myocardial infarction (MI) stands out as the most researched and difficult to diagnose by clinicians. MI is divided into two subgroups ST-elevation-MI (STEMI) and Non-ST-elevation (NSTEMI) depending on the ST segment in the Electrocardiogram (ECG) signals. Early diagnosis of these diseases importance in terms of reducing the risk of death. This study aims to develop a system that automatically analyzes ECG signals belonging to CVDs, which can be difficult for clinicians to analyze manually. In this study, signal processing was performed using the Robust Local Mean Decomposition (RLMD) method, and feature selection was performed using the LASSO and Chi-square methods. Random Forest (RF) and Support Vector Machines (SVM) algorithms were used for classification. The classification process was performed among CVDs (three groups) and by including healthy control data (four groups). The results were evaluated with accuracy, Area Under the Curve (AUC), and negative predictive value (NPV) criterion rates. The findings obtained showed that the RF algorithm was slightly superior to SVM in classification performance. Similarly, the LASSO method achieved more successful results in feature selection. It was observed that the AUC and NPV rates were over 84% for the groups examined in two different classifiers and feature selection methods. This situation proves that the proposed method is a stable and reliable analysis tool. This research draws attention as one of the limited number of studies using the RLMD method in the analysis of ECG signals and the diagnosis of CVD. The presented study can contribute to the development of systems that can overcome the difficulties experienced in manual analysis, especially in the diagnosis of CVDs and to support physicians in clinical processes.