Loading...

Ictal EEG signal denoising by combination of a semi-blind source separation method and multiscale PCA

Pouranbarani, E ; Sharif University of Technology

1273 Viewed
  1. Type of Document: Article
  2. DOI: 10.1109/ICBME.2016.7890961
  3. Abstract:
  4. Contamination of ictal Electroencephalogram (EEG) signals by muscle artifacts is one of the critical issues related to clinically diagnosing seizure. Over the past decade, several methods have been proposed in time, frequency and time-frequency domain to accurately isolate ictal EEG activities from artifacts. Among denoising approaches Canonical Correlation Analysis (CCA) and Independent Component Analysis (ICA) are widely used. Denoising based on Generalized EigenValue Decomposition (GEVD) is one of the Semi-Blind Source Separation (SBSS) methods which has been recently proposed. In the GEVD-based method, a couple of time-frequency covariance matrices are used. These time-frequency (TF) covariance matrices are calculated in the time-frequency domain using a special time-frequency mask. This time-frequency mask is extracted based on the time-frequency signature corresponding to the time-frequency spectrum of an ictal source obtained by CCA. In this paper, we use MultiScale Principal Component Analysis (MSPCA) in order to extract the time-frequency mask. To this end, a novel SBSS method, called CCA-MSPCA-TF-GEVD, is proposed and the efficacy of CCA-MSPCA-TF-GEVD compared with that of CCA, ICA and CCA-TF-GEVD is presented. The simulation results, using simulated data, validate the superiority of the proposed method compared with other methods. In addition, the applicability of the proposed denoising method for source localization is evaluated. © 2016 IEEE
  5. Keywords:
  6. Ictal EEG signal ; Multiscale principal component analysis (MSPCA) ; Muscle artifacts ; Biomedical engineering ; Biomedical signal processing ; Biophysics ; Blind source separation ; Covariance matrix ; Eigenvalues and eigenfunctions ; Electroencephalography ; Frequency domain analysis ; Independent component analysis ; Muscle ; Speech recognition ; Canonical correlation analysis ; EEG signals ; Electroencephalogram signals ; Generalized eigenvalue decompositions (GEVD) ; Independent component analysis(ICA) ; Multi-scale principal component analysis ; Semi-blind source separation (SBSS) ; Time-frequency spectrum ; Principal component analysis
  7. Source: 2016 23rd Iranian Conference on Biomedical Engineering and 2016 1st International Iranian Conference on Biomedical Engineering, ICBME 2016, 23 November 2016 through 25 November 2016 ; 2017 , Pages 226-231 ; 9781509034529 (ISBN)
  8. URL: https://ieeexplore.ieee.org/document/7890961