A Lightweight QRS Detector for Single Lead ECG Signals Using a Max-Min Difference Algorithm
Diptangshu Pandit,Li Zhang,Chengyu Liu,Samiran Chattopadhyay,Nauman Aslam,Chee Peng Lim
DOI: https://doi.org/10.1016/j.cmpb.2017.02.028
IF: 6.1
2017-01-01
Computer Methods and Programs in Biomedicine
Abstract:Background and objectivesDetection of the R-peak pertaining to the QRS complex of an ECG signal plays an important role for the diagnosis of a patient's heart condition. To accurately identify the QRS locations from the acquired raw ECG signals, we need to handle a number of challenges, which include noise, baseline wander, varying peak amplitudes, and signal abnormality. This research aims to address these challenges by developing an efficient lightweight algorithm for QRS (i.e., R-peak) detection from raw ECG signals.MethodsA lightweight real-time sliding window-based Max-Min Difference (MMD) algorithm for QRS detection from Lead II ECG signals is proposed. Targeting to achieve the best trade-off between computational efficiency and detection accuracy, the proposed algorithm consists of five key steps for QRS detection, namely, baseline correction, MMD curve generation, dynamic threshold computation, R-peak detection, and error correction. Five annotated databases from Physionet are used for evaluating the proposed algorithm in R-peak detection. Integrated with a feature extraction technique and a neural network classifier, the proposed ORS detection algorithm has also been extended to undertake normal and abnormal heartbeat detection from ECG signals.ResultsThe proposed algorithm exhibits a high degree of robustness in QRS detection and achieves an average sensitivity of 99.62% and an average positive predictivity of 99.67%. Its performance compares favorably with those from the existing state-of-the-art models reported in the literature. In regards to normal and abnormal heartbeat detection, the proposed QRS detection algorithm in combination with the feature extraction technique and neural network classifier achieves an overall accuracy rate of 93.44% based on an empirical evaluation using the MIT-BIH Arrhythmia data set with 10-fold cross validation.ConclusionsIn comparison with other related studies, the proposed algorithm offers a lightweight adaptive alternative for R-peak detection with good computational efficiency. The empirical results indicate that it not only yields a high accuracy rate in QRS detection, but also exhibits efficient computational complexity at the order of O(n), where n is the length of an ECG signal.