Proton MR spectroscopic imaging depicts diffuse axonal injury in children with traumatic brain injury

Abstract

Background and purpose: Diffuse axonal injury (DAI) after traumatic brain injury (TBI) is important in patient assessment and prognosis, yet they are underestimated with conventional imaging techniques. We used MR spectroscopic imaging (MRSI) to detect DAI and determine whether metabolite ratios are accurate in predicting long-term outcomes and to examine regional differences in injury between children with TBI and control subjects.

Methods: Forty children with TBI underwent transverse proton MRSI through the level of the corpus callosum within 1-16 days after injury. T2-weighted, fluid-attenuated inversion recovery, and susceptibility-weighted MR imaging was used to identify voxels as normal-appearing or as nonhemorrhagic or hemorrhagic injury. Neurologic outcome was evaluated at 6-12 months after injury. Metabolite ratios for total (all voxels), normal-appearing, and hemorrhagic brain were compared and used in a logistic regression model to predict long-term outcome. Total and regional metabolite ratios were compared with control data.

Results: A significant decrease in N-acetylaspartate (NAA)/creatine (Cr) and increase in choline (Cho)/Cr (evidence of DAI) was observed in normal-appearing (P < .05) and visibly injured (hemorrhagic) brain (P < .001) compared with controls. In normal-appearing brain NAA/Cr decreased more in patients with poor outcomes (1.32 +/- 0.54) than in those with good outcomes (1.61 +/- 0.50, P = .01) or control subjects (1.86 +/- 0.1, P = .00). In visibly injured brains, ratios were similarly altered in all patients. In predicting outcomes, ratios from normal-appearing and visibly-injured brain were 85% and 67% accurate, respectively.

Conclusion: MRSI can depict injury in brain that appears normal on imaging and is useful for predicting long-term outcomes.

Fig 1.

Mean NAA/Cr and Cho/Cr ratios from all voxels plotted as a function of the neurologic outcome assigned by a pediatric neurologist based on the PCPC score 6–12 months after injury. CTL indicates control; MOD, moderate disability; NL, normal; SEV, severe disability; and VS, vegetative state.

Fig 2.

Total mean metabolite ratios plotted by neurologic outcomes or control (CNTL). Ratios were calculated from all voxels including those containing hemorrhagic and nonhemorrhagic DAI lesions for patients with TBI. Asterisk indicates P = .01; double asterisk, P = .000.

Fig 3.

Regional mean metabolite ratios from MRSI data collected in a transverse plane through the level of the CC plotted by neurologic outcomes compared with control (CNTL); * indicates P = .00–.05; +, P < .02.

Fig 4.

15-year-old female adolescent ejected from a car. Patient had a good outcome (GOS score = 1, normal) at 12 months after injury. Total mean metabolite ratios are 1.63 (normal) for NAA/Cr, and 1.67 for NAA/Cho, and 1.06 (increased) for Cho/Cr. A and B, T2-weighted MR image (A) corresponding SWI (B) show hemorrhagic lesions in the body of the CC and bifrontal extra-axial collections. A with grid overlay shows the 160-mm FOV and 54 (6 × 9)-voxel volume of interest (rectangle). Boxes indicate corresponding voxels in C. C, Spectral map from 54 voxels in the volume of interest shows a (1) spectrum from normal-appearing brain in the anterior CC with decreased NAA (2.0 ppm), (2) a spectrum with reduced metabolite signal intensity due to a small hemorrhagic lesion in the mid CC, and (3) a spectrum from parietal white matter with normal metabolite ratios.

Fig 5.

12-year-old boy hit by a car. Patient had a poor outcome (GOS score = 4, severe disabilities) at 12-month follow-up. Total mean metabolite ratios from MRSI were 1.14 (decreased) for NAA/Cr, 1.13 for NAA/Cho, and 1.11 (increased) Cho/Cr. A and B, Contiguous SWI images show a moderately large hemorrhagic lesion in the deep right frontal lobe. C, Corresponding T2-weighted MR image shows the lesion (circle) with an overlay of the volume of interest (rectangle). Boxes indicate corresponding voxels in D. D, Spectral map shows loss of metabolite signal intensity in area of the lesion (circle) and (2) a spectrum from normal-appearing left FWM with markedly decreased NAA/Cr and increased Cho/Cr, as compared with (3) a spectrum from normal-appearing parietal white matter with decreased NAA/Cr and normal Cho/Cr.

Fig 6.

Mean metabolite ratios from visibly injured (hemorrhagic, Hem) brain and normal-appearing (Nrm) brain were significantly different from those in control subjects (CNTL). Asterisk indicates P < .02. NAA/Cr from normal-appearing brain was the only ratio that differentiates between good- and poor- outcome groups; + indicates P = .01.