Brain injury is the most frequent cause of death in the first 40 years of life.There have been numerous attempts to classify head injury.A classification based on computerized tomography (CT) findings has never been confirmed to be of high predictive value.Magnetic resonance imaging (MRI) is difficult to perform in comatose patients while they are ventilated, but it has been done lately with increasing frequency.Patients and Methods: From 1998 to 2009 in a prospective study 220 patients in coma after a head injury were examined at our institution.All patients were in coma for at least 24 hours, MRI was obtained within 8 days of the injury, median after 3.5 days after an initial CT had been performed.T1 and T2 weighted images were obtained with a 1.5 Tesla magnet.Lesions were identified by a team of neuroradiologists blinded to the clinical findings.The location of the brain, where a lesion was identified was noted, not the extent or volume of the lesion.Outcome was classified according to the Glasgow Outcome Scale.Results 6 months after the injury were calculated with statistical means including cross tables, Fishers exact test, Anova, Chi2 test.Results: The frequency of brain stem lesions was 67%, almost none of these could be identified with CT.Statistical analysis identified a significant correlation of mortality with a brain stem lesion.Within the group of brain stem lesions bilateral pontine lesions had the worst prognosis.Conclusion: Obviously the brain stem lesions are associated with increased mortality.It may be concluded that compression of the brain stem from swelling, edema or hematoma should be avoided.This appears to support the idea of a generous indication to perform craniectomy when clinical signs of brain stem compression, such as anisocoria or posturing, are noted.