Department of Mechatronic Engineering, Islamic Azad University-South Tehran Branch, Tehran, Iran
Abstract
This paper describes how one can design a computer program and implement it on an ordinary PC or PDA for detecting very long duration electrocardiogram (ECG) PQRST events with processing capability (time complexity) less than 1300 clock/samples along with the accuracy more than 99.86%. For detecting and delineating QRS complexes, a noise-robust instantaneous concavity analysis was applied. For detecting and delineating P- and T- waves, the analysis of local extremums of the QRS-eliminated signal was used. The proposed method was applied to several databases (more than 1,000,000 beats which were normal or abnormal) with different sampling frequencies and bit-rates. After application of the algorithm, the average detection sensitivity Se=99.96% and positive predictivity P+=99.94%, were obtained for QRS complex. The average delineation error were about -3.0 msec, 2.5 msec, and 2.8 msec, for P-QRS-T events, respectively. By implementing the proposed algorithm computer program to selected databases, the required variation for the core parameters set of the program was about 0.0% for all sampling frequencies and bit rates. The maximum computational complexity required during application of the method to databases was estimated to be lower than 1300 clocks/sample. These merits make the algorithm eligible to be implemented by a mobile-phone or PDA.
)