Time plane ECG feature extraction using Hilbert transform, variable threshold and slope reversal approach

The electrocardiogram (ECG) representing the electrical activity of the heart is the key bio-signal for aiding the clinical staff in disease diagnosis. Generally the various characteristic features of ECG are extracted and used for decision-making purposes. In the present paper, an accurate detection technique of various points of QRS complex, QRS duration, R peak height, T peak, T onset and T offset points, T peak height, ST and QT segment duration is proposed. The algorithm is tested on various ECG data (Normal, Myocardial Infraction etc.) of all the 12 leads, taken from PTB Diagnostic ECG Database (PTB-DB). In this proposed method, a single lead ECG signal is smoothed and its first derivative is computed at first. Hilbert Transform of the first derivative of the ECG signal makes easier to locate the regions of R peaks. On the both sides of the R peak, a slope reversal will be noted as Q peak (left side) and S peak (right side). On the left of Q peak and on the right of S peak, any zero slope in the differentiated ECG signal will be identified as QRS onset and QRS offset points respectively. T peaks are detected using the same approach used for detecting R peaks and T onset, T offset points are detected using the approach used for QRS onset and offset detection. After extracting all these time domain ECG features, ECG baseline modulation correction is done for all cardiac data set. R and T peak height, QRS, QT and ST segment duration of each cardiac cycle is measured and also the average error of measurement is calculated. A good scale of accuracy is achieved in Sensitivity (Se=99.81%) and Positive Predictivity (PP=99.93%) of R peak detection. The algorithm is implemented on MATLAB 7.1 environment.