Smart computing mechanism for noise detection and elimination in ECG signal

The cardiovascular disease is one of the most common causes of death around the world. The analysis of electrocardiograms (ECGs) is an important tool in early diagnosis of arrhythmias. However, sometime the measurement data would be corrupted by noises which may cause by the wrong equipment operation, poor contact of the electrode, or even the breath of the users. These noises would make cardiologists or automatic detection system hard to make a correct diagnosis. Therefore, the noise detection and elimination of ECG data become an important issue. In this study, we proposed a detection and elimination mechanism for the five types of noise. Besides, if the segment does not have any important information and cannot be repaired, we will eliminate it and combine the remaining usable segments into a pure signal for ECG enhancement. The experimental results showed that the noise recognition classifier yielded 0.956 area under the ROC curve, 84.4% accuracy, 97.8% sensitivity, and 80.5% specificity, respectively. The accuracy of disease detection system also could be improved by using the combination of usable segments. Hence, we believed this smart computing mechanism could address ECG enhancement and interpret a contaminated ECG signal more accurately.