Research Article
A Novel Method for Extracting High-Quality RR Intervals from Noisy Single-Lead ECG Signals
@INPROCEEDINGS{10.1007/978-3-030-57115-3_6, author={Shan Xue and Leirong Tian and Zhilin Gao and Xingran Cui}, title={A Novel Method for Extracting High-Quality RR Intervals from Noisy Single-Lead ECG Signals}, proceedings={Bio-inspired Information and Communication Technologies. 12th EAI International Conference, BICT 2020, Shanghai, China, July 7-8, 2020, Proceedings}, proceedings_a={BICT}, year={2020}, month={8}, keywords={Electrocardiogram (ECG) R-peak detection High-quality RR intervals Noisy ECG Combining methods}, doi={10.1007/978-3-030-57115-3_6} }- Shan Xue
Leirong Tian
Zhilin Gao
Xingran Cui
Year: 2020
A Novel Method for Extracting High-Quality RR Intervals from Noisy Single-Lead ECG Signals
BICT
Springer
DOI: 10.1007/978-3-030-57115-3_6
Abstract
In previous studies, plenty of high-accuracy R-peak detection methods were performed on electrocardiogram (ECG) signal analysis. However, these excellent results were usually obtained from some standard and common databases. When applying these detectors on ECG signals collected in daily life and ordinary experiments, or acquired from wearable single-lead ECG devices, the R peak detection accuracies were usually unsatisfying. Due to the influence of data-acquiring environment and devices, the collected ECG signals were often noisy. Each R-peak detection method has its own advantages and may be superior in a certain kind of ECGs. In this study, we proposed a method combining seven R-peak detection methods to get high-quality RR Intervals (RRIs) from noisy ECGs. This new method included two steps, 1) obtain preliminary R-peak annotations through combining seven R-peak detection methods, and 2) calculate the quality score of each R-peak annotations detected in 1) according to the ECG waveform features including kurtosis, skewness and the frequency band power ratio, then exclude the wrong annotations based on the quality scores. The proposed method was evaluated on two databases: MIT-BIH Arrhythmia database and the CPSC2019 training set. The R peak detection average accuracies on these two databases were 98.89% and 55.47% respectively. The results showed that the method proposed in this paper performed better than the seven common R-peak detection methods, especially in noisy ECG signals.
