Abstract
The early prediction of consciousness recovery from traumatic brain injury (TBI) is crucial to make decisions about the appropriate use of prolonged intensive care. Diffusion tensor imaging (DTI) has been proposed as a biomarker of white matter injury that could be used in a classification purpose. Instead of region-of-interest-based approach, we applied voxel-based approaches (voxel-based DTI and tract-based spatial statistics) on 30 patients with TBI to identify, without any prior, the brain regions that were specifically damaged in unfavorable 1-year outcome group compared to the favorable one. DTI were acquired at mean 23 days (5-53 days) and two DTI-derived indices, fractional anisotropy (FA) and apparent diffusion coefficient (ADC), were tested. Our results showed that (1) ADC is not a relevant biomarker for early 1-year outcome prognosis; (2) FA measured in inferior longitudinal fasciculus, in cerebral peduncle, in posterior limb of the internal capsule, and in posterior corpus callosum is specifically decreased in unfavorable outcome group compare to the favorable one; (3) a linear discriminant analysis using the FA measured in these four regions showed good classification performance (sensitivity = 86% and specificity = 86%). These findings confirm the relevance of the use of DTI as biomarkers for consciousness recovery after TBI and support the possible use of these biomarkers for early classification of patients.