Rémy LEHEMBRE

Probing consciousness is a major challenge in the evaluation of patients with disorders of consciousness (DOC). Improvement of behavioral scales have led to more accurate bedside assessments but are still bound to variability due to different factors such as the subjectivity and experience of the assessor and fluctuations of the patient state. Brain imaging techniques such as positron emission tomography (PET), functional MRI (fMRI), diffusion tensor imaging (DTI) or electroencephalography (EEG) may provide different and complementary information on metabolism, haemodynamic function, tissue organization and electrical activity of the brain. Measurements obtained from these methods could lead to objective assessments and quantification of correlates of consciousness.

In this thesis we explored electrophysiological markers of brain activity to 1) better comprehend mechanisms of loss of consciousness, 2) differentiate patients with vegetative state / unresponsive wakefulness syndrome (VS/UWS) from patients in minimally conscious state, and 3) provide means of communications for patients with DOC.

The first part of this thesis presents a state of the art on the use of EEG in the context of DOC, in particular we review the existing literature on standard clinical EEG, quantitative EEG (qEEG), brain computer interfaces, and event related potentials. Then signal processing techniques for EEG are described, namely time frequency decompositions at the scalp level, functional and effective connectivity and source reconstruction methods.

In the second part, the results of three analyses are exposed. First a descriptive approach of loss of consciousness under ketamine is provided and reveals a concomitant increase in theta and gamma frequencies at the scalp and source level in frontal and parietal cortices. Furthermore, assessment of effective connectivity with dynamical causal modelling (DCM) shows an increase in intracortical connectivity during ketamine induced unresponsiveness. Second, scalp level differences in power spectra and connectivity measurements are used to discriminate VS/UWS patients from MCS patients. We found that VS/UWS patients display slow delta oscillations while MCS patients have more power and higher connectivity in the theta range. Finally, the use of source reconstruction based classification of upcoming finger movements did not improve classification rates obtained with common spatial patterns (CSP) in the studied dataset.

Members of the jury :
Prof. Benoît MACQ, promoter (UCL)
Prof. Michel VERLEYSEN, chairperson (UCL)
Prof. Christophe DE VLEESCHOUWER (UCL)
Prof. Steven LAUREYS (ULg)
Dr Quentin NOIRHOMME (ULg)
Dr Andrea SODDU (ULg)