Prognostic ability of diffusion tensor imaging parameters among severely injured traumatic brain injury patients

Abstract

Diffuse axonal injury (DAI) represents the most common primary intra-axial form of traumatic brain injury (TBI), comprising approximately half of all such injuries. Patients presenting with DAI follow a highly variable clinical course, with initial status frequently discrepant from long-term neurological outcome. Diffusion Tensor Imaging (DTI) is sensitive to disruptions in neuronal structure that may not be appreciated on CT or conventional MRI and may serve as an important prognostic imaging marker. In this study, we retrospectively evaluated the data from 84 patients to determine if the whole brain DTI parameters (axial diffusivity λ1, radial diffusivity λPERP, apparent diffusion coefficient ADC, and fractional anisotropy FA) are predictive of their clinical outcome as determined by discharge Glasgow Coma Scale (GCS). The first group consisted of 52 severely injured patients (GCS≤8) that either died (n=10), had poor outcome (n=12) or good outcome (n=27). The second group was comprised of mildly injured patients (GCS≥14 during entire hospitalization) that served as the reference group. Whole brain measurements of the DTI parameters were measured on each patient, and using non-parametric statistics, the measures from within each group were compared. Significant differences were found in ADC, λ1, and λPERP between the three outcome groups. Further, these measures were shown to be significantly related to GCS at scan. Using ordinal logistic regression models adjusted for age, gender, and admission GCS, DTI parameters were shown to significantly predict outcomes in severe TBI patients (death, poor outcome, good outcome, or mild injury). Evaluation of TBI patients may be improved using DTI measures as they correlate well with clinical measures, reflect the severity of injury, and can predict outcome. This method may provide a valuable independent tool to predict clinical outcomes in DAI. © 2010 Springer-Verlag.