In last decades, several advances in the neuro-intensive management have lead to decrease mortality in Intensive Care Units. A significant morbidity remains as patients survive after a traumatic coma with uncertain quality of awakening and a high risk of functional disability. Predicting awareness recovery and functional disability of those who will awake constitutes a major challenge to inform patients' relatives, to give the best chances in terms of rehabilitation resources or to adapt intensive cares to a reasonable level. Tools currently available are not sufficient neither to predict bad awakening outcome nor to predict good functional outcome. In many countries, life's support cessation is a constant call for robust evaluation as soon as possible in ICU but it is mandatory to reach a positive predictive value of non-awaking close to 100%. Many clinical, electro-physiological, biological, radiological and functional parameters have been conducted with comatose patients assuming the purpose to predict outcome. Regarding unfavourable outcome, the gold standard is the abolition of the N20 component of somatosensory evoked potentials but the specificity is high enough only for patients with anoxic coma. Several neurophysiological markers such as MMN, P300 are correlated to a favourable outcome but the sensitivity and specificity remains low for patients who suffered a severe traumatic brain injury. New Diffusion Tensor imaging sequences provide complementary information to detect small structural lesions (diffuse axonal lesions). Recently, functional MRI analyzing Resting State has also been proposed as a prognostic marker during coma. PET using Fluoro-Desoxy-Glucose is able to assess the metabolism in key regions of the awakening network in either anaesthesia or sleep. Recent studies have reported interesting results at the chronic stage but to knowledge, these tools have only been used to address pathophysiology's issues and never to improve coma prognosis at the initial stage. The investigators hypothesize that the heterogeneity of the population requires a global and accurate assessment of the central nervous system, combining structural, metabolic and functional information in order to refine the prognosis.

The protocol integrates in one-sequence most radiological markers of brain injury within a unique PET-MRI in Lyon. The most relevant originality of the study consists in confronting FDG-PET and MRI sequences to a large clinical, electrophysiological and biological battery. The added clinical value would be to question the synergistic effect of each parameter and to find out which ones are the most useful for awakening prediction, as they have not been compared in a multi-parametric database.

PET-MRI, as a new device combining physiological and prognostic questioning, allows us:

• to implement a more integrative physio-pathological analysis
• to avoid the cofounding effect of awareness' fluctuations in recording simultaneously multiple functional imaging techniques.

The RS will be analyzed at 2 epochs in order to assess the stability of brain connectivity, related to neuronal activity (glucose metabolism) and brain perfusion.

Inclusion Criteria:

1. Patients presenting either vascular or traumatic lesions
2. Patients with no response to simple command 48h after sedation's cessation
3. Patients included between 7 days (to reduce the effect of early oedema) and 30 days after the coma (to control the homogeneity of our cohort)
4. Patients included between 18 and 75 years old, to limit the risk of care withdrawal for poor previous medical condition
5. Patients evaluated by a global electrophysiological assessment at beside in ICU in a short delay before or after PET-MRI

Exclusion Criteria:

1. Patient with an associated anoxic encephalopathy
2. Patients with contra-indication to MRI
3. Patients with hypersensibility to the active molecules (FDG) or to one of this excipient
4. Pregnant women
5. Minor patients
6. Patients under legal protection
7. Patients not affiliated to French health care system
8. Patients in poor medical condition (hemodynamic, respiratory instability)
9. Patients moribund or with previous decision of care withdrawal
10. Absence of relatives to give written consent