In this talk I will present ongoing work on EEG signal analysis for critical care applications, carried out as part of my PhD thesis with CEA Leti and CHU Grenoble. EEG has been shown to be interesting for prognostication and complication prediction for certain types of comatose patients. But its potential remains partially untapped because of the high cost and low convenience of having records analysed by human experts. First I will introduce the need for automated EEG analysis in the intensive care unit. Second I will present how we applied supervised deep learning approaches to the slightly different task of EEG sleep scoring as a proof of concept of the use of artificial neural network methods on raw EEG. Finally, I will mention ongoing research directions for unsupervised analysis of critical care EEG using deep neural networks.