chronic phase, as these patients frequently show a degree of brain atrophy. This means that the changes in DTI measures such as lower FA may reflect partial volumes, resulting from contamination of the white matter measurements by CSF. The TBSS approach employed here attempts to limit the impact of this problem via the analysis involving ‘skeletonisation’ of the white matter tracts, restricting the subsequent statistical analysis to the centres of these tracts (Smith et al., 2006). This removes the white matter at the junctions with CSF and gray matter that is particularly prone to partial volume effects. Hence, the TBSS approach is arguably more robust to the effects of brain atrophy than approaches such as ROI or VBM analyses.
However, despite TBSS being able to deal with partial volume effects in the case of large white matter tracts, smaller fibre tracts such as the fornix are more problematic. For the projection of individual FA values onto the spatially invariant skeleton structure, in order to ‘fill’ the skeleton and construct that individual’s white matter tract skeleton, the search for the maximum FA value proceeds perpendicularly. For thin tracts such as the fornix, no wider than skeleton structure voxel dimensions, the FA value extracted is not likely to be the true ‘peak’ value, but is instead likely to additionally reflect some partial voluming from the surrounding CSF. In addition, as well as change in the DTI metrics, brain contusions or DAI may cause remote atrophy, which could exacerbate the partial volume effects in brains affected by these lesions. It follows that these effects may have played a role in some of the differences observed in the DTI metrics between the patient and control groups, as well as between the patients with and without microbleeds. By contrast, the DTI abnormalities observed here in patients without any macroscopic lesions relative to healthy controls should be relatively unaffected by atrophy-
related partial volumes, but still dependent on the accuracy of the projection of individual FA values onto the tract skeleton step of TBSS.
A related limitation is the use of scalar measures of FA and MD to provide insight into microstructural properties of white matter (Basser and Pierpaoli, 1996). Although in general FA should become lower and MD higher as the proportion of damaged relative to healthy axons increases within a voxel (Kou et al., 2010), these relationships are not quite as simple. A major limitation of scalar measures such as FA and MD is that they do not directly reflect features of tissue microstructure, but instead are sensitive to a variety of factors that affect its integrity and structural organization. The size and packing density of axons, as well as the permeability of cell walls and membranes, can affect these measures. The distribution of fibre orientations within an imaging voxel also affects the measured average anisotropy and diffusivity. Thus, a change observed in these statistics does not necessarily correspond to any specific changes in tissue microstructure (Alexander et al., 2010).
Investigating multiple DTI metrics between different participant subgroups naturally results in multiple comparisons being carried out. All TBSS analyses were corrected for comparisons carried out across multiple imaging voxels contained within the white matter skeleton. No additional correction was made to account for the separate analyses of the four DTI metrics and those of the different patient subgroups. This is based on the premise that all of these comparisons ask the same overall question, and their results lead to the same conclusion – relative to healthy individuals TBI patients show white matter disruption (indexed by lower FA and higher diffusivities), which is greater in those patients with macrostructural white matter injury. The results of all comparisons, including those producing non-significant results are included in the results section. It can be seen from the tables that list these results that their general pattern is very consistent, pointing to the conclusion stated above.
A limitation stemming from participant recruitment is that the TBI group consisted of predominantly males, whilst the control group had more equal proportions of males and females. As differences between male and female brains may exist that may in turn influence the DTI metrics (see e.g. Hsu et al., 2008), an attempt was made to statistically control for the effects of this possible confound on the relationship between group membership and white matter structure. It is acknowledged, however, that linear statistical approaches may not be able to deal with confounds involved in complex multifactorial relationships, such as those relating to
gender differences in brain structure. Therefore, in retrospect, it would have been preferable to selectively recruit more males than females for the control group in order to achieve better balanced gender ratios for the patient and control groups.
A surprising discrepancy between the TBI and control groups was that the patients significantly outperformed controls in current verbal and nonverbal reasoning ability. Furthermore, they did so whilst having, on average, a significantly lower WTAR-estimated premorbid level of intellectual ability than the control group. Whether this has any implications for the interpretation of the DTI comparisons between the patients and controls depends on the existence and causal direction of a relationship between IQ and white matter structure. Whilst it is of course possible that having a higher IQ leads to a higher structural integrity of white matter connections, potentially contributing to the group differences observed here in the DTI metrics, this is not likely to be of major concern considering the patients’ superior current intellectual function. In fact, that the patients have a higher current IQ than the controls potentially makes the current results more striking and the analysis more conservative. That the patients’ intellectual function should improve as a consequence of a TBI is peculiar, however, and a more likely explanation is that use of the WTAR has in this study produced articificially low estimates of the patients’ pre-injury IQ, especially given their currently high level of intellectual functioning. There are a number of possible reasons that could lead to such unreliability of the WTAR estimates.
A salient explanation is that good performance on this test is dependent upon correct pronunciation of English words, whilst English was not the first language of all participants (see Chapter 2; Green et al., 2008). In addition, some have questioned the use of scores on a word pronunciation tests as an index of the kind of intellectual ability that putatively is resistant to the effects of head injury (Riley & Simmonds, 2003). It has been suggested, for example, that the level of impairment on the WTAR can vary as a function of severity of injury (Mathias et al., 2007). It is also possible that there was some improvement in cognitive function after TBI, which could partly explain why the patient group on average were not found impaired in current reasoning ability (see Green, Melo et al., 2008, for further discussion). The observations by Mathias et al. (2007), suggesting that greater injury severity after TBI may sometimes be associated with worse WTAR performance, highlight that some caution may be necessary when premorbid IQ is estimated in a mixed sample of patients of various levels of injury severity, as
this could potentially lead to underestimation of premorbid intellectual ability in some more severely injured patients.
4.4.7 Conclusions. To summarise, widespread FA, MD, and axial diffusivity differences were found between TBI patients and healthy controls using TBSS. This study also demonstrates the presence of white matter abnormalities in patients with no evidence of microbleed evidence of DAI on gradient-echo MRI and in patients with clinically mild TBI. These results suggest that DTI is extremely sensitive to microstructural white matter damage following TBI and that these abnormalities of white matter tract structure are more widespread across the brain than previously demonstrated by ROI studies that have limited their analysis to a fraction of the total white matter.
Importantly, these results also suggest that DTI can detect white matter abnormalities in cases where tools used in standard clinical care fail to do so despite there being factors that may make the presence of a TBI possible. However, whilst use of DTI for research purposes is on the increase, it remains to be further validated clinically before it can be used to support diagnosis of TBI (Saatman et al., 2008). Given the susceptibility of axonal fibres to damage through TBI, not necessarily visible on conventional MRI as identifiable white matter lesions, DTI could be particularly useful in identifying and understanding mild and very mild TBI, where subtle forms of structural brain damage are possible (see Bigler & Bazarian, 2010).
Many patients with mild TBI who have normal conventional scans experience cognitive problems, some of which are persistent (Lipton et al., 2008; Nakayama et al., 2006; Rugg-Gunn et al., 2001). These impairments can have many important consequences such as limiting the return to productive employment or education (Drake, Gray, Yoder, Pramuka, & Llewellyn, 2000). Early and accurate identification of TBI patients who are likely to have long-lasting problems would thus not only be valuable clinically but also socioeconomically. More generally, improved early detection of axonal damage after TBI could inform subsequent care and rehabilitative interventions. Possible better outcomes that could be achieved by early and well targeted interventions include, for example, reduced length of stay in acute care, improved cognitive function and increased functional independence (Edwards, McNeil, & Greenwood, 2003; Mackay, Bernstein, Chapman, Morgan, & Milazzo, 1992; Wagner et al., 2003).
suited to analysing the complex patterns of disruption to the structure of the brain’s white matter tracts following TBI. Regional white matter damage that DTI (but not necessarily standard MRI) can detect has previously been found to correlate with cognitive dysfunction after TBI (e.g. Kraus et al., 2007). The voxelwise approach used here may, however, be better suited than ROI approaches to investigating the relationships between DTI indices of white matter tract structure and their relevance for high-order cognitive function, likely to be supported by spatially distributed brain networks. This will be the focus of the next chapter that explores the relationships of the DTI metrics studied here with measures of cognitive function.
CHAPTER 5: Relationships between white matter tract structure and cognitive functions
Diffusion tensor imaging and tract-based spatial statistics (TBSS) combined may provide a sensitive way of investigating the complex relationships between axonal injury and cognitive functions that are supported by large-scale neural networks. It has not been previously investigated across the whole brain how damage to specific white matter tracts may affect cognitive performance following TBI. Here, the TBSS approach described in the previous chapter was extended to explore whether the structure of specific white matter connections is associated with performance within particular cognitive domains. These relationships were investigated in the same sample of 28 TBI patients and 26 age-matched controls as in Chapter 4. Specific patterns of white matter abnormality predicted performance on some tasks. The structure of the fornix was correlated with associative learning and memory performance across the patient and control groups, as was the structure of fronto-temporal/parietal connections with immediate logical memory. By contrast, the structure of frontal connections showed relationships with set-shifting ability, which differed in the two groups: patients showed a relationship between stronger directionality of diffusion within these tracts and better performance, whilst controls did not. In addition, patients showed a relationship between greater axial diffusivity in a region that contains descending perceptual-motor fibres and faster responses on a simple choice-reaction task. These results highlight the complexity of relationships between structural properties of white matter and different cognitive functions. Although widespread and sometimes chronic white matter abnormalities are identified after TBI, their impact on high-level cognitive function is likely to depend on damage to key pathways that link nodes in the distributed brain networks supporting the specific cognitive processes involved.
5.1 Introduction
5.1.1 Outline. Whilst Chapter 1 provided a general overview of the anatomical