White matter structure associated with executive function

verbal memory performance, the groups showed distinct relationships between white matter structure and set-shifting ability, one of the three executive functions investigated. As expected, after TBI increasing mean and radial diffusivities (indexing lower integrity) were associated with executive impairment of set-shifting ability and these effects were seen in white matter tracts connecting the frontal regions. However, no such relationships were observed for controls (further discussed in section 5.4.4).

These results reveal that white matter tracts show variable structure-function relationships in the healthy and damaged brain, suggesting that highly complex relationships exist between white matter microstructure and high-order cognitive functions. Therefore, it appears that determining these relationships in the healthy brain and then using this information to guide an ROI analysis in a TBI group (e.g. Niogi, Mukherjee, Ghajar, Johnson, Kolster, Lee et al., 2008) may not be optimal. Although Niogi, Mukherjee, Ghajar, Johnson, Kolster, Lee et al. (2008) showed that regional DTI abnormalities after TBI can relate to a specific cognitive deficit in the absence of a relationship with a deficit in another domain, this approach does not exclude the possibility that the structure of white matter tracts that were not investigated would have also shown a relationship with performance on either of their two cognitive measures, in addition to the ‘specific’ ROI. Their demonstration of a ‘correlational dissociation’ does, however, suggest that the structure of certain critical tracts may be more relevant than that of others in terms of specific cognitive functions, which the current results lend support to.

The flexible voxelwise approach employed here appears particularly suitable for the investigation of complex relationships between white matter microstructure and cognitive

function. First, a standard ROI approach requires a priori knowledge of the likely locations of effects of interest, which is both difficult and restrictive, particularly at this relatively early stage of research. Second, if the ROIs to be investigated after brain injury are first defined in healthy individuals, it is assumed by default that similar relationships exist in the healthy and injured brain. This is problematic as the current understanding of these relationships is limited and the subtle white matter damage after TBI often diffuse, limiting the usefulness of an ROI approach. The TBSS approach used in the present study does not require the placement of specific ROIs and allows the relationship between the DTI and cognitive variables to be modelled within the framework of a general linear model, which can be adapted to look at various DTI metrics across all major white matter tracts. As discussed in Chapter 4, TBSS is also more robust to the effects of brain atrophy on white matter tracts than some widely used alternatives as it restricts the analysis to tract centres only and thus reduces the impact of atrophy around the edges avoiding such partial volume effects.

Increases in MD and radial diffusivity, potential indices of white matter injury, have been previously reported following TBI (Kraus et al., 2007; Sidaros et al., 2008; Kennedy et al., 2009). The involvement of tracts connecting the frontal with posterior medial and parietal regions suggests that axonal damage after TBI can disrupt the integrity of what has been dubbed the brain’s ‘structural core’. This core is densely connected with the temporal and frontal cortices via white matter tracts (Hagmann et al., 2008). The most prominent ‘node’ of this structural core is the precuneus/posterior cingulate cortex that also corresponds to the posterior component of the human ‘default mode network’ (DMN), a functional network found engaged in a variety of neuroimaging paradigms across a wide range of cognitive states (Honey et al., 2009; Chapter 1). Here, reduced structural integrity (indexed by high radial diffusivity) of white matter tracts interconnecting these posterior with frontal regions was found to be associated with worse set- shifting ability after TBI. This is consistent with the proposal that executive dysfunction following brain injury is partly the result of disconnection between frontal and more posterior brain regions involved in executive processes (Miller & D’Esposito, 2005). Converging evidence comes from studies of normally ageing older adults that have shown age-related decline in the structural coherence of frontal connections to predict executive dysfunction (Davis et al., 2009; Kennedy & Raz, 2009; O’Sullivan et al., 2001; Perry et al., 2009).

slowed information processing on a range of tasks, including the ‘executive’. A related consideration comes from functional neuroimaging research into cognitive efficiency (Rypma et al., 2006). In particular, Rypma et al. suggest that patterns of neural interaction within the fronto- parietal attentional network may critically mediate individual differences in cognitive efficiency. Specifically, individual variability in processing speed across a range of tasks may depend on prefrontal activity through the PFC exerting more control over the more posterior brain regions in slow-performing than in fast-performing individuals. Therefore, the structural integrity of connections between the frontal and more posterior parietal regions could be more important after brain injury for achieving optimal cognitive efficiency than it is in the healthy brain. This could partly explain the distinct relationships observed in the TBI and control groups between radial diffusivity in the specific tracts and set-shifting performance. The predominantly right- lateralized pattern of these results could also be related to the known involvement of the right- lateralized fronto-parietal network in attentionally demanding tasks such as those requiring set- shifting (Fox et al., 2005).

5.4.4 Unexpected findings and limitations. Whilst the involvement of white matter