Purpose: Traumatic brain injury (TBI) is a major cause of death and disability worldwide. TBI is often characterized by white matter damage producing alterations in brain connectivity. These alterations are commonly suspected to disrupt the function of large-scale networks that support cognition. One analytic approach to examine these alternations is graph theory, which examines the brain from a network perspective. The present study aimed to examine whether injury severity modulates the association between white matter connectivity and cognition in TBI using a diffusion MRI based network analysis.
Materials & Methods: Graph theoretical network analysis was applied in order to assess differences in structural connectivity between TBI and healthy control subjects. We performed DTI scans of 22 patients with chronic TBI in different injury severities (mild and moderate-severe TBI) defined by Glasgow Coma Scale (GCS) score, and 22 healthy control subjects. White matter connections between 90 gray matter brain regions were defined using tractography. Weighted brain structural matrices were constructed for each subject, and network measures were calculated. An objective cognitive deficits score of non-verbal abstract reasoning in TBI patients was calculated by subtracting pre-injury (obtained in adolescence as part of the aptitude tests of the Israeli Defense Forces draft board) from post-injury performance on a Raven progressive matrices test (RPM-R). In order to assess the injury severity effect on network measures, we performed ANOVA between controls, mild TBI and moderate-severe TBI on cluster coefficients, strength, efficiency, betweenness centrality and characteristic path length. These network measurements were analyzed in both global and local levels. To examine the effect of injury severity on cognition, Pearson`s correlations were computed between network properties and deficit scores of the RPM-R task in TBI patients.
Results: Global analysis revealed differences between groups in strength, efficiency and cluster coefficient. Thus, when injury severity increases, graph measures decrease. Moreover, these measures differed between the 3 groups within particular hub regions, including the insula, frontal superior medial cortex, precuneus, frontal superior orbital cortex and caudate nucleus. Local analysis revealed a severity effect in strength, mainly in frontal and cingulum regions. Additionally, we found that reduced network efficiency in the left precuneus was associated with a greater deficit in nonverbal abstract reasoning performance.
Conclusion: Our findings support the notion that injury severity affects network measures of structural connectivity, in particular disconnection of network hubs, which in turn may contribute to cognitive impairments.