Deep Neural Networks for Automatic Classification of Anesthetic-Induced Unconsciousness
Despite the common use of anesthetics to modulate consciousness in the clinic, brain-based monitoring of consciousness is uncommon. We combined electroencephalographic measurement of brain activity with deep neural networks to automatically discriminate anesthetic states induced by propofol. Our results with leave-one-participant-out-cross-validation show that convolutional neural networks significantly outperform multilayer perceptrons in discrimination accuracy when working with raw time series. Perceptrons achieved comparable accuracy when provided with power spectral densities. These findings highlight the potential of deep convolutional networks for completely automatic extraction of useful spatio-temporo-spectral features from human EEG.
KeywordsConsciousness Anesthesia EEG Deep learning
We acknowledge funding from the UK Engineering and Physical Sciences Research Council [EP/P033199/1], the Belgian National Fund for Scientific Research, the European Commission, the Human Brain Project, the Luminous project, the French Speaking Community Concerted Research Action, the Belgian American Educational Foundation, the Wallonie-Bruxelles Federation, the European Space Agency, the University and University Hospital of Liège (Belgium). This research was undertaken with the support of the Alan Turing Institute (UK Engineering and Physical Sciences Research Council Grant EP/N510129/1).
- 6.Stober, S., Cameron, D.J., Grahn, J.A.: Using convolutional neural networks to recognize rhythm stimuli from electroencephalography recordings. Neural Inf. Process. Syst. 2014, 1–9 (2014)Google Scholar
- 9.Krizhevsky, A., Sutskever, I., Hinton, G.E.: ImageNet classification with deep convolutional neural networks. Adv. Neural Inf. Process. Syst., 1–9 (2012). http://dx.doi.org/10.1016/j.protcy.2014.09.007