Removal of electrocardiographic interference and artifacts from diaphragm electromyography

Abstract

Diaphragmatic electromyography (dEMG) holds the potential to monitor respiration. However, its acquisition is affected by electrocardiographic (ECG) interference and motion artifacts, making its investigation and use in clinical practice still challenging. Singular value decomposition (SVD) methods have been reported in the literature denoising the dEMG. Here a new ratio index criterion is combined with a SVD based algorithm to aid this challenge. The advantage of the proposed approach is in use of the frequency spectrum as a reference to remove unwanted components from the signal. Two synthetic datasets combining EMG with ECG only and with ECG and motion artifacts were tested using a signal-to-noise ratio ranging from −20 to 0dB to assess the performance of the method. The performance was compared with an earlier developed SVD-based algorithm that employed a calibration curve for the selection of unwanted components. Our results show that our new proposed method reached significantly better performance in both time and frequency domains for the majority of presented SNRs in the dataset containing artifacts. Additionally for the same dataset, the method obtained the average median improvement in the SNR of 12 dB and, the average median percentage improvement of 157% in the reconstructed EMG signal. The solution does not need an additional reference for the ECG. Its performance was proven on the data containing not only electrocardiographic disturbance but also motion artifacts, showing promise for further use on the real data.