An improved QRS complex detection method having low computational load


Yakut Ö., Bolat E.

BIOMEDICAL SIGNAL PROCESSING AND CONTROL, cilt.42, ss.230-241, 2018 (SCI-Expanded) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 42
  • Basım Tarihi: 2018
  • Doi Numarası: 10.1016/j.bspc.2018.02.004
  • Dergi Adı: BIOMEDICAL SIGNAL PROCESSING AND CONTROL
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus
  • Sayfa Sayıları: ss.230-241
  • Anahtar Kelimeler: Electrocardiogram (ECG), QRS complex detection, Digital FIR filter, Biomedical signal processing, Dynamic threshold, R-PEAK DETECTION, DETECTION ALGORITHM, ARRHYTHMIA DATABASE, AUTOMATIC DETECTION, FIDUCIAL POINTS, ECG, TRANSFORM, STANDARD, SYSTEM, FILTER
  • Kocaeli Üniversitesi Adresli: Evet

Özet

In this study, an improved QRS complex detection method having low complexity is proposed. This method includes two stages as preprocessing and decision making. The preprocessing stage consists of a band-pass digital FIR filter, squaring operation, moving average, normalization steps whereas the decision stage includes only one phase realizing the QRS complex detection. In the preprocessing stage, the unwanted frequency components of the Electrocardiogram (ECG) signal were reduced by using the digital FIR filter. The filtered signal was enhanced with the squaring operation and finally integrated and smoothed by moving average step. In the decision making stage, R peaks were detected employing a dynamic threshold process incorporating with the preprocessing stage for determining the QRS complex components. The R peaks were detected by comparing the time intervals between two successive R peaks with the calculated time range. For assessing the performance of the proposed method, it was tested using the ECG recordings (about 1.3 million beats) taken from the five standard databases as MIT-BIH Arrhythmia, Fantasia, MIT-BTH Noise Stress Test, QT and European ST-T. In this study, 1296137 beats of 272 cases were tested for QRS detection and the average sensitivity, Se was obtained as 99,60%, while the average positive predictivity, +P was provided as 99,77%. The contribution of the proposed method is that the training, selection, setting and prediction processes are not required while determining the necessary parameters. Therefore, this contribution reduces the complexity of the method resulting in decreasing the computational load as well without compromising on high performance indices (Se and +P). Since the proposed method provides low complexity and computational load together with high accuracy, it can be implemented as a QRS detector for applications of telemedicine and embedded systems easily in contrast to the algorithms having higher complexity. (C) 2018 Elsevier Ltd. All rights reserved.