Chinmayee Dora1, Pradyut Kumar Biswal2. 1. Dept. of Electronics & Telecommunication Engineering, International Institute of Information Technology, Bhubaneswar, India. 2. Dept. of Electronics & Telecommunication Engineering, International Institute of Information Technology, Bhubaneswar, India. Electronic address: pradyut@iiit-bh.ac.in.
Abstract
BACKGROUND AND OBJECTIVE: Ambulatory based healthcare system use limited electrodes for electroencephalogram (EEG) acquisition at concerned electrode position, to minimize the instrumentation and computational complexity. But, again the possibility of contamination is inevitable depending on the electrode position on the scalp. This paper proposes an electrocardiogram (ECG) artifact correction algorithm in the absence of coherent ECG for automatic analysis/diagnosis of the acquired contaminated single channel EEG signal. METHODS: The proposed algorithm uses an enhanced and modified version of signal decomposition i.e. modified variational mode decomposition (mVMD) to obtain band limited intrinsic mode functions (BLIMFs) from EEG epoch. The mVMD is found to be useful when the signal contains properties that are correlated. Further, exploiting the correlation among the obtained mode functions the ECG artifact components are identified. An ECG reference is estimated and QRS complexes are suppressed. The effective EEG reconstruction is performed by simply adding the remaining BLIMFs to QRS complex suppressed estimated reference. This is owing to the robust reconstruction feature provided by mVMD. RESULTS: Upon the comparative evaluation for both real and semi-simulated dataset, the proposed algorithm is providing less distortion to the EEG brain activity frequency bands, and is also less computationally intensive than the existing Ensembled empirical mode decomposition (EEMD) based algorithm that requires ECG reference channel. Evaluation of semi-simulated dataset obtained an average correlation of 97% between EEG signals before contamination and after correction of ECG artifact. CONCLUSION: The proposed algorithm efficiently corrects the ECG artifact from EEG while overcoming the limitations such as, (1) requirement of a reference ECG channel, (2) requirement of R-R interval or amplitude thresholding for QRS complex identification.
BACKGROUND AND OBJECTIVE: Ambulatory based healthcare system use limited electrodes for electroencephalogram (EEG) acquisition at concerned electrode position, to minimize the instrumentation and computational complexity. But, again the possibility of contamination is inevitable depending on the electrode position on the scalp. This paper proposes an electrocardiogram (ECG) artifact correction algorithm in the absence of coherent ECG for automatic analysis/diagnosis of the acquired contaminated single channel EEG signal. METHODS: The proposed algorithm uses an enhanced and modified version of signal decomposition i.e. modified variational mode decomposition (mVMD) to obtain band limited intrinsic mode functions (BLIMFs) from EEG epoch. The mVMD is found to be useful when the signal contains properties that are correlated. Further, exploiting the correlation among the obtained mode functions the ECG artifact components are identified. An ECG reference is estimated and QRS complexes are suppressed. The effective EEG reconstruction is performed by simply adding the remaining BLIMFs to QRS complex suppressed estimated reference. This is owing to the robust reconstruction feature provided by mVMD. RESULTS: Upon the comparative evaluation for both real and semi-simulated dataset, the proposed algorithm is providing less distortion to the EEG brain activity frequency bands, and is also less computationally intensive than the existing Ensembled empirical mode decomposition (EEMD) based algorithm that requires ECG reference channel. Evaluation of semi-simulated dataset obtained an average correlation of 97% between EEG signals before contamination and after correction of ECG artifact. CONCLUSION: The proposed algorithm efficiently corrects the ECG artifact from EEG while overcoming the limitations such as, (1) requirement of a reference ECG channel, (2) requirement of R-R interval or amplitude thresholding for QRS complex identification.