Cardiac arrhythmias predictive detection methods with wavelet-SVD analysis and support vector machines

We expand the idea to develop new bio-signal processing tools that could predict possibility of future risk of abnormalities in ECG signals. The goal is to detect an inherent defect hidden in an ECG signal using wavelet analysis and support vector machines. We apply singular value decomposition analysis of spectral energy distribution in time-frequency plane to extract features, which is essentially independent of the actual duration of each event. We then classify the life threatening cardiac arrhythmias using support vector machines. We also investigate robustness of the developed system under presence of continuous Gaussian white noise. We obtained 92% sensitivity and 75% specificity for clean data and 81% sensitivity and 62% specificity for noisy data on our database. The proposed method could assist the health care professionals by earlier prediction of a disease and hence could facilitate in patient management, i.e., to provide a proper treatment for prevention or reduction of the future risk.