Automatic bad channel detection in intracranial electroencephalographic recordings using ensemble machine learning. Issue 3 (March 2018)

• Record Type: Journal Article
• Authors: Tuyisenge, Viateur; Trebaul, Lena; Bhattacharjee, Manik; Chanteloup-Forêt, Blandine; Saubat-Guigui, Carole; Mîndruţă, Ioana; Rheims, Sylvain; Maillard, Louis; Kahane, Philippe; Taussig, Delphine; David, Olivier
• Is Part Of: Clinical neurophysiology. Volume 129:Issue 3(2018:Mar.)
• Page Start: 548
• Page End: 554
• Publication Date: 2018-03
• Subjects: Intracranial EEG -- Stereo-EEG -- Bad channels -- Feature extraction -- Ensemble bagging -- Machine learning; DES direct electrical stimulation -- ECoG electrocorticography -- EEG electroencephalography -- iEEG intracranial electroencephalography -- SEEG stereo-electroencephalography; Neurophysiology -- Periodicals; Electroencephalography -- Periodicals; Electromyography -- Periodicals; Neurology -- Periodicals; 612.8
• DOI: 10.1016/j.clinph.2017.12.013
• Languages: English
• ISSNs: 1388-2457

Abstract

Highlights: We propose a method that detects automatically bad channels from intracranial EEG (iEEG) datasets. It computes iEEG features specific to bad channels and uses an ensemble bagging classifier. The bad channel classification accuracy was demonstrated to be excellent on a large data sample. Abstract: Objective: Intracranial electroencephalographic (iEEG) recordings contain "bad channels", which show non-neuronal signals. Here, we developed a new method that automatically detects iEEG bad channels using machine learning of seven signal features. Methods: The features quantified signals' variance, spatial–temporal correlation and nonlinear properties. Because the number of bad channels is usually much lower than the number of good channels, we implemented an ensemble bagging classifier known to be optimal in terms of stability and predictive accuracy for datasets with imbalanced class distributions. This method was applied on stereo-electroencephalographic (SEEG) signals recording during low frequency stimulations performed in 206 patients from 5 clinical centers. Results: We found that the classification accuracy was extremely good: It increased with the number of subjects used to train the classifier and reached a plateau at 99.77% for 110 subjects. The classification performance was thus not impacted by the multicentric nature of data. Conclusions: The proposed method to automatically detect bad channels demonstrated convincing results and can be envisaged to be used on larger datasets for automatic quality control of iEEG data. Significance: This is the first method proposed to classify bad channels in iEEG and should allow to improve the data selection when reviewing iEEG signals.