Dynamics of the structural connectome in traumatic brain injury

Traumatic Brain Injury (TBI) is a leading cause of death and disability globally, with survivors often experiencing ongoing and debilitating cognitive impairments (e.g., slowed processing speed, poor attention, and executive functioning deficits). These impairments are often linked to focal lesions in regions of the cerebral cortex thought to uphold each cognitive function. However, the spectrum of impairments experienced by individual patients are not fully explained by focal lesions of the grey matter; instead, emerging theories suggest that many cognitive burdens result from disconnections in the white matter of the brain. With the advent of diffusion MRI (dMRI), new techniques are available to study how TBI disrupts the white matter pathways that connect brain regions (structural connectomics). Structural connectomics allows the quantification of network disruption in TBI patients using graph theoretical analyses, with studies reporting alterations in brain network integration and segregation. These studies suggest that graph metrics may be used as a ‘biomarker’ for TBI patients’ cognitive impairments, by linking changes in brain derived graph metrics to cognitive symptoms. However, challenges remain in ascribing behavioural relevance to graph metrics in this newly emerging field.

This thesis critically evaluates the use of graph theoretical measures of the structural connectome in moderate-severe TBI, and their use at a single-subject level. First, a meta-analysis of studies comparing healthy controls and TBI patients using graph metrics is used to demonstrate that communication metrics are most robustly linked to brain injury. This review also highlights issues with the over-interpretation of the relationship between graph metrics such as path-length and the efficiency of cognitive processes. Second, a study in healthy adults shows that communication metrics are related to processing speed. This relationship between cognitive performance and measures of network alteration is underpinned by biologically plausible models of cognition and brain structure. Third, a profile of graph theoretical properties and alterations in six TBI patients is explored using a personalised connectomics approach. Spiderplots are used to represent graph metric alterations in each patient compared to healthy controls. Profiling individual patients in this way provides new insights into how graph metrics relate to lesion characteristics and TBI subtypes. Taken together, this thesis explores 1) how structural network topology is altered in patients with TBI, 2) how graph metrics can be interpreted, 3) how a personalised connectomics approach to TBI can be implemented, and 4) the methodological considerations for studying TBI using graph theory. The collective results of thesis indicate that graph metrics display potential for characterising network alterations in patients with brain injury; specifically, a profiling approach can account for heterogeneity in the TBI population, informing clinical decision making.