Heart Bio-mechanical Function Monitoring via Analysis of Higher Order Statistical Moments of Electrocardiogram and Blood Pressure Signals

In this study, a new heart electro-mechanical monitoring method based on the analysis of higher order moment features extracted from Discrete Wavelet Transform (DWT) of electrocardiogram (ECG) and arterial blood pressure (ABP) signals is described. In summary, appropriate digital pre-processing to remove noise and motion artifacts, application of a trous DWT to signals, extraction of first order (mean), second order (variance), third (skewness), forth (kurtosis) moments from suitable DWT scales and finally post-processing of the resulted decision statistic to determine edges and occurrence locations of each ECG and ABP events, can be stated as steps of the presented study. The presented algorithm was applied to all records of MIT-BIH Arrhythmia and MIT-BIH Polysomnographic databases and also to DAY hospital high-resolution holter database (totally more than 1,500,000 beats) and the average values of Se=99.97%, P+=99.95%, LE = 13.4 msec, CPS = 150,000 samples/sec were obtained for sensitivity, positive predictivity, maximum delineation error and computational processing speed respectively, showing marginal improvement of ECG and ABP waveforms parallel detection-delineation performance and accuracy.

Keywords:

ECG Detection-Delineation , Arterial BloodPressure , Discrete Wavelet Transform , Multi Lead Analysis , mean value , variance , skewness , kurtosis