High-dimensional discriminant analysis of human cardiac arrhythmias

Sudden ventricular arrhythmia is a leading cause of death. It is important to be able to distinguish between different arrhythmias in order to deliver proper treatment. This study presents results of linear and quadratic discriminant analysis of normal sinus rhythm, ventricular fibrillation and ventricular tachycardia in different representation spaces, using different observation lengths. In particular, 0.5 s, 1 s, 2 s and 4 s segments of electrocardiogram waveforms are considered, along with their magnitude spectra, and lower dimensional projections of magnitude spectra onto principal components. All considered representations are of much higher dimension than in prior art. Results suggest that Fourier magnitude spectra of 2 s windows, or low dimensional projections, are sufficient for achieving best classification results. Results also suggest that additional improvements could be obtained by considering more sophisticated non-linear decision boundaries.