Computer-Aided Detection of Atrial Fibrillation Episodes from Electrocardiogram Signals Using Variational Mode Decomposition

Atrial Fibrillation (AF) is defined as a heart beating irregularly. It may increase the risk of stroke and sometimes results in heart failure. Lately, the number of patients with AF episodes has increased exponentially worldwide. Early detection is the only way to reduce and control its effect. However, it is crucial to detect AF in the early stages in most patients who do not show any significant symptoms in the early stage. Many recent studies have shown ways to identify early symptoms based on the signature present in the recording of the electrocardiogram. Usually, the signature is present in terms of either frequency domain or time-frequency domain characteristics. None of the studies have shown classification using features from each domain and described the specific behavior of each feature relating to the characteristic of AF. In this work, a rigorous collection of features from each domain is calculated and analyzed for the automatic detection of AF patterns using variational mode decomposition. High-frequency decomposed components have been used for further processing. The dataset from the MIT-BIH repository has been utilized to measure the robustness of the introduced approach. We obtained the highest accuracy of 98.67% on the MIT-BIH database. Obtained indicators show that the proposed method outperformed latest approaches.