Early stage prediction of sudden cardiac death

Sudden cardiac death is one among the category of natural deaths due to cardiac causes. The symptoms of death arise almost one hour or several minutes prior to the incidence. The onset of symptoms is called sudden cardiac arrest. It may appear due to prior unhealthy state of heart or sometimes without any known cardiac cause. Worldwide, the researchers and physiologists are facing challenge and publishing all possible solutions to predict the sudden cardiac death at an early stage. Recently the heart rate variability (HRV) and all its possible parameters classified with different classifiers like k-nearest neighbor (k-NN), support vector machine (SVM), multi-layer perceptron (MLP) etc. are used extensively for the prediction of SCD. But clinical applications of such methods are still questionable due to false detection of QRS peaks in ECG signals for SCD patients. An attempt has been made in this paper to predict sudden cardiac death at an early stage i.e. one hour prior to its occurrence using heart rate variability analysis. The ECG signals are taken from online database for normal sinus rhythm (healthy subject) and sudden cardiac death (SCD subject). Different derived measures of HRV classified with k-NN classifier strongly confirm the prediction of SCD at such an advanced stage. For clinical applications of such methods, an effect of incorrect detection of QRS peaks on heart rate (beats per minute, BPM) is significantly considered and presented here.