Pattern Analysis Framework for Detection of Cardiac Disorders using VCG Signal Analysis

Bundle Branch Block (BBB), Cardiomyopathy (CMP), and Dysrhythmia (DYS) are interrelated cardiac conditions. They can make a significant impact on the electrical conduction and overall functioning of the cardiac activity. BBB is categorized by delayed or blocked electrical signals in bundle branches, while CMP refers to structural abnormalities or damage to heart muscles. DYS encompasses various irregular heart rhythms. This paper proposes a pattern recognition framework for precise detection of BBB, CMP and DYS using Vectorcar-diography (VCG). The dataset is obtained from Physikalisch Technische Bundesanstalt (PTB) Diagnostic Electrocardiogram (ECG). VCG signals are extracted from ECG and preprocessed using Empirical Mode Decomposition (EMD). Time, spectral and cepstral features are extracted from the preprocessed signals, followed by the application of the Back Elimination Method through which features with the best contribution to the accurate detection are selected. Classification is done using 10-fold cross-validation. The highest accuracy of 97.9% is achieved using KNN (Fine) classifier.