ECG fiducial points extraction using QRS morphology and adaptive windowing for real-time ECG signal analysis

The proper performance of computerized ECG analysis is highly dependent on the accuracy of extracted fiducial points. The aim of this study is to develop an ECG analysis algorithm which analyzes 12 lead ECG signals simultaneously and determines global duration and intervals including P duration, PQ interval, QRS duration, and QT interval. The basic concept of this algorithm is based on the method previously used for automatic detection of wave boundaries using slope feature and adaptive thresholding. The method of identifying the type of QRS complex, QRS classification, and threshold determination were significantly modified considering the morphological features of each signal. These modifications led to less computational complexity for implementing on electrocardiography system and high accuracy to be adaptable with medical standard requirements. Furthermore a novel algorithm for calculating the heart axis have presented; performing based on combination of limb leads, this algorithm greatly reduces error rate in heart axis determination. For performance evaluation, both ECG data collected in Rajaie cardiovascular center and CSE database have been used. Mean difference and standard deviation of measurement results in millisecond estimated by proposed method vs CSE reference values are 0.7 and 3.70 for P duration, 1.40 and 3.52 for PQ interval, -1.10 and 3.04 for QRS duration, and 4.09 and 5.61 for QT interval.