Signal Quality Index-Based Two-Step Method for Heart Rate Estimation by Combining Electrocardiogram and Arterial Blood Pressure Signals
Yulin Liu,Shoushui Wei,Shuai Zhang,Feifei Liu,Jianqing Li,Chengyu Liu
DOI: https://doi.org/10.1166/jmihi.2018.2462
2018-01-01
Journal of Medical Imaging and Health Informatics
Abstract:Accurately monitoring heart rate (HR) is important in intensive care unit (ICU) application, which is usually performed by electrocardiogram (ECG) analysis. However, the contamination of ECG is common and this may lead to the erroneous estimation of HR. In this study, an improved HR detector, named as signal quality index (SQI)-based two-step method, was proposed. This method consists of two steps: (1) HR estimation by SQI-based Kalman filter for single channel signal, (2) Sal-based HR estimation by fusing the results from ECG and ABP signals. First, two ORS detectors, GQRS and SQRS, were used to obtain SQI and HR from ECG signal, and the wabp algorithm and a fuzzy logic analysis method were used to obtain SQI and HR values from ABP signal. Then, HR values derived from ECG and ABP were separately processed by Kalman filter. HR after Kalman filter were fused based on the signal SQIs, to obtain the fused HR. The method was evaluated on the set-p (with good signal quality) and set-p2 (with poor signal quality) databases from the PhysioNet/CinC Challenge 2014. For set-p, mean absolute errors from the fusion of ECG and ABP signals were 0.233, 0.289, 0.417, 0.456. 0.440 and 7.865 beat/min for the six SQI levels (0.9-1.0, 0.8-0.9, 0.7-0.8, 0.6-0.7, 0.5-0.6, and 0-0.5) respectively. As comparison, for set-p2, the signal quality was relative low, mean absolute errors from the fusion of ECG and ABP signals were 0.191, 0.391, 0.688, 1.135, 0.921 and 13.32 beat/min for the six SQI levels respectively. As expected, the fusion method performed better than the single channel signal (ECG or ABP).