Diagnosis of Diffuse Axonal Injury with Diffusion Tensor Imaging

Presentation Type:

Scientific Paper

General Subject Classification:


Time / Location:

Mon, 6/13, 4:00 PM
International West


  • Sylvain Bouix, PhD
    Psychiatry Neuroimaging Laboratory, Brigham and Women's Hospital, Harvard Medical School
  • Ofer Pasternak, PhD
    Psychiatry Neuroimaging Laboratory, Brigham and Women's Hospital, Harvard Medical School


  • We wish to demonstrate that DTI is a modality that has a specific sensitivity to TBI and in combination with state-of-the-art neuroimaging techniques can improve the diagnosis of mTBI by showing specific location of brain injury.


The most common form of TBI is Diffuse Axonal Injury (DAI). This injury results from the unequal rotation and/or acceleration forces applied to the parenchyma, which stretch and injure the axons. Currently, conventional imaging techniques are not likely to diagnose DAI, because these methods are not sensitive to subtle white matter pathology. However, recent studies have shown that Diffusion Tensor imaging (DTI) is able to quantify DAI compared with other imaging modalities. In our work, we show how to use DTI processing techniques to better diagnose DAI.

Our approach is to build a normative white matter (WM) atlas based on 46 high resolution DTI scans from our Normal Control (NC) database. The dataset contains 8 females and 38 males (age is 34.63+/-12.13 years). This atlas includes both quantitative (Fractional anisotropy FA) and qualitative information about the location and diffusion properties of WM in a normal population.

We also acquired DTI scans with the same protocol and the same scanner on patients diagnosed with mild TBI (9 patients, GCS=13-15, remained symptomatic at least 2-months after injury). For a TBI subject, we registered its FA map to the atlas and, for each voxel, computed the z score of the FA of the patient against the distribution of FA in the normal population. We found that our tool located many areas where the z score was greater than 2, indicating a significant deviation from the NC mean FA. In contrast, when we computed the z score map of a NC against the normal population, we found a few very small regions with a z score greater than 2.

Our results suggest that comparing the FA map of an individual diagnosed with mTBI to a normative atlas can reveal locations where the brain is affected, increasing the specificity of the diagnosis.