Zeszyty Naukowe Politechniki Białostockiej. Informatyka Qrs Complex Detection in Noisy Holter Ecg Based on Wavelet Singularity Analysis

In this paper, we propose a QRS complex detector based on the Mallat and Hwang singularity analysis algorithm which uses dyadic wavelet transform. We design a spline wavelet that is suitable for QRS detection. The scales of this decomposition are chosen based on the spectral characteristics of electrocardiogram records. By proceeding with the multiscale analysis we can find the location of a rapid change of a signal, and hence the location of the QRS complex. The performance of the algorithm was tested using the records of the MIT-BIH Arrhythmia Database. The method is less sensitive to timevarying QRS complex morphology, minimizes the problems associated with baseline drift, motion artifacts and muscular noise, and allows R waves to be differentiated from large T and P waves. We propose an original, new approach to adaptive threshold algorithm that exploits statistical properties of the observed signal and additional heuristic. The threshold is independent for each successive ECG signal window and the algorithm uses the properties of a series of distribution with a compartments class. The noise sensitivity of the new proposed adaptive thresholding QRS detector was also tested using clinical Holter ECG records from the Medical University of Bialystok. We illustrate the performance of the wavelet-based QRS detector by considering problematic ECG signals from a Holter device. We have compared this algorithm with the commercial Holter system Del Mar’s Reynolds Pathfinder on the special episodes selected by cardiologist.