A robust QRS detection using novel pre-processing techniques and kurtosis based enhanced efficiency

The QRS complex detection is very imperative step prior to detecting any other wave component, beat segments, or any other morphological parameters of electrocardiogram (ECG). The paper presents a novel technique for QRS complex detection with minimum pre-processing required. The technique uses two-stage median filter and Savitzky-Golay smoothing filter for pre-processing. The decision rule is based on root-mean-square of the signal. The technique evolves with a novel approach for isoelectric line detection and a kurtosis-based false positive correction method. The robustness of the proposed technique been substantiated on Fantasia Database (FTD), MIT-BIH Arrhythmia Database (MIT-AD), MIT-BIH Normal Sinus Rhythm Database (MIT-NSD), and BIDMC Congestive Heart Failure Database (CHFD). The detection sensitivity (Se), positive predictivity (+P), detection error rate (DER), and accuracy (Ac) of proposed technique on FTD is Se = 99.90%, +P = 99.91%, DER = 0.19%, and Ac = 99.81%. On MIT-AD the technique has Se = 99.50%, +P = 99.56%, DER = 0.93%, and Ac = 99.08%. The technique shows Se = 99.36%, +P = 99.43%, DER = 1.21%, and Ac = 98.81% on MIT-NSD. Tested on CHFD, the techniques has Se = 99.68%, +P = 99.78%, DER = 0.52%, and Ac = 99.46%. The technique is also tested on self-recorded data-set and has Se = 99.97%, +P = 99.99%, DER = 0.04%, and Ac = 99.96%. The observation of above results shows the high performance and robustness of QRS complex detection, which will lead to highly accurate cardiac health prognosis.