R-peak Detector Benchmarking using FieldWiz Device and Physionet Databases

Tiago Rodrigues, Hugo Silva, Ana Fred


The R-peak detection in an Electrocardiography (ECG) signal is of great importance for Heart Rate Variability (HRV) studies and feature extraction based on fiducial points. In this paper, a real-time and low-complexity algorithm for R-peak detection is evaluated on single lead ECG signals. The method is divided in a pre-processing and a detection stage. First, the pre-processing is based on a double differentiating step, squaring and moving window integration for QRS complex enhancement. Secondly, the detection stage is based on a finite state machine (FSM) with an adaptive thresholding for R-peak detection. The tested approach was benchmarked in a private FieldWiz Database with other commonly used QRS detectors, and later evaluated in the Physionet Databases (mitdb, nstdb, ltstdb and CinC Challenge 2014). The proposed approach resulted in a Sensivitity (Se) of 99.77% and Positive Predictive Value (PPV) of 99.18% in the FieldWiz database, comparable with the evaluated state of the art QRS detectors. In the Physionet Databases, the results showed to be highly influenced by the QRS waveform, for MIT-BIH (MITDB) achieving a median PPV of 99.79% and a median Se of 99.52%, with overall PPV of 98.35% and Se of 97.62%. The evaluated method can be implemented in wearable systems for cardiovascular tracking devices in dynamic use cases with good quality ECG signals, achieving comparable results to state of the art detectors.


Paper Citation