Comparison of Acute Diffusion Tensor Imaging and Conventional Magnetic Resonance Parameters in Predicting Long-Term Outcome after Blunt Cervical Spinal Cord Injury

Abstract

This prospective longitudinal study compares the ability of conventional and diffusion tensor imaging (DTI) parameters made at the cervical spinal cord injury (CSCI) site to predict long-term neurological and functional outcomes. Twenty patients with CSCI, with follow-up at 6 or 12 months, and 15 control volunteers were included. Conventional magnetic resonance imaging (MRI) and DTI parameters were measured on admission and follow-up studies. Stepwise regression analysis was performed to find relevant parameters (normalized DTI values, conventional MRI measurements, hemorrhagic contusion [HC] or non-HC [NHC]) that correlated with three primary outcome measures: patient International Standards for Neurological Classification of Spinal Cord Injury total motor score (ISNCSCI-TMS), ability to walk, and expected recovery of upper limb motor scores (ER-ULMS) at 6 or 12 months.

Univariate analysis showed HC (p < 0.0001 to 0.0098), lesion length on follow-up MRI (p < 0.0001 to 0.019), mean diffusivity (p = 0.01 to 0.045), and axial diffusivity (p = 0.004 to 0.023) predicted all three primary outcomes. Conspicuity of HC was significantly better on axial susceptibility-weighted imaging (SWI) compared with T2* images (p = 0.0009). A negative correlation existed between HC volumes on sagittal SWI images and follow-up ISNCSCI-TMS ( p = 0.02). The regression model identified NHC as the best predictor of the ability to walk (sensitivity = 88.9%; specificity = 100%; positive predictive value = 100%; negative predictive value = 91%; p < 0.0001) and lesion length on follow-up MRI as the best predictor of ER-ULMS (β coefficient = 0.12, standard error [SE] = 0.07; R² = 0.64; p = 0.0002). Finally, NHC (β coefficient = 24.2, SE = 3.7; p < 0.0001) and lesion length on initial MRI (β coefficient = 0.78, SE = 0.2; p = 0.002) were the best predictors of ISNCSCI-TMS (R² = 0.83; p < 0.0001). Our study demonstrates HC and follow-up lesion length are potential neuroimaging biomarkers in predicting long-term neurological and functional outcome following blunt CSCI.