Reconstruction of EEG from limited channel acquisition using estimated signal correlation. (May 2016)

• Record Type: Journal Article
• Authors: Ramakrishnan, A.G.; Satyanarayana, J.V.
• Is Part Of: Biomedical signal processing and control. Volume 27(2016)
• Page Start: 164
• Page End: 173
• Publication Date: 2016-05
• Subjects: Correlated signals -- Karhunen–Loeve Transform -- Electroencephalography -- Motor-imagery tasks -- Compressed sensing -- Convex optimization -- EEG electrode placement; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2016.02.004
• Languages: English
• ISSNs: 1746-8094

Abstract

Abstract : Highlights: KLT computed using all channels of EEG captures the inter-channel correlation. Correlation learnt is used to reconstruct channels not recorded in a test session. Estimation of missing or noisy channels and 10-10 channels from 10-20 recordings. Average NMSE is less than 1.0% when only a few channels (≤ 6) are predicted. Useful in sleep studies for predicting disconnected channels and in wireless EEG. Abstract: Nearby scalp channels in multi-channel EEG data exhibit high correlation. A question that naturally arises is whether it is required to record signals from all the electrodes in a group of closely spaced electrodes in a typical measurement setup. One could save on the number of channels that are recorded, if it were possible to reconstruct the omitted channels to the accuracy needed for identifying the relevant information (say, spectral content in the signal), required to carry out a preliminary diagnosis. We address this problem from a compressed sensing perspective and propose a measurement and reconstruction scheme. Working with publicly available EEG database, we have demonstrated that up to 12 channels in the 10-10 system of electrode placement can be estimated within an average error of 2% from recordings of the remaining channels. As a limiting case, all the channels of the 10-10 system can be estimated using recordings on the sparser 10-20 system within an error of less than 20% in each of the significant bands: delta, theta, beta and alpha.