Investigation of peak detection methodologies for ECG signals

摘要

The electrocardiogram (ECG) is a vital tool with which medical practitioners are able to assess and diagnose a patient’s heart condition; the accuracy of the signal is vital for correct diagnosis. Digital signal processing techniques are used to filter accurate information from the input signals received from the body. The implementation of peak detection algorithms enhances the accuracy of ECG wave form deflections which provide diagnostic information to the medical professionals to ascertain an individual’s cardiovascular condition. The ECG wave form of one cardiac cycle consists of five wave form deflections, each denoted as ‘P, Q, R, S, T’. In this thesis the objective is to investigate the accurate measurement of the R peak deflection within an ECG signal; this deflection possesses the highest magnitude and is used to ascertain the heartbeat rate. Testing was conducted by the implementation of nine peak detection algorithms that were derived from the 1990 paper “A Comparison of the Noise Sensitivity of the Nine QRS Detection Algorithms” , written by Friesen, Jannett, Jadallah, Yates, Quint, and Nagle. These nine algorithms were implemented using MATLAB software and ECG signals acquired from the MIT-BIH Arrhythmia Database. The results demonstrated the performance of each algorithm with regard to accuracy of R peak detection. Analysis of the algorithms was conducted using a synthetic test signal to ascertain an improvement in the peak detection results. The nine algorithms were subjected to additional signals obtained from the MIT-BIH Arrhythmia Database which provided varying results due to a large amount of interference within these ECG signals, but compared favourably to the results from the Friesen et al paper. Further testing and analysis more clearly defined the performance of the algorithms with different levels of noise and types of filters. Additional experiments were conducted with the intent of finding a new peak detection algorithm that would compare favourably or outperform results obtained from the above paper. The algorithm was compared using eight arrhythmia ECG signals. The results demonstrated that after the fine tuning of threshold settings, this new algorithm succeeded in performing well by returning similar results found by Friesen et al.