• No results found

Methodological limitations and related considerations

CHAPTER 3: Standard magnetic resonance imaging of traumatic brain injury and

3.4 Discussion

3.4.6 Methodological limitations and related considerations

and neuropsychological performance. Rorden and Kranath (2004) have considered a range of possible explanations, as summarised below.

First, lesion studies assume that discrete anatomical modules support specific cognitive functions (i.e. cognitive functions are assumed to be localised). It appears, however, that many brain functions are instead supported by distributed brain networks. Second, brain damage in TBI is not limited to the site of impact nor it is constrained by the boundaries of functional ‘nodes’. Third, plasticity of the human brain means that lesions that only partially damage a functional area known to be involved in a specific cognitive function may not have any obvious consequences. This complicates investigation of lesion-function relationships after TBI, as does the differential vulnerability of particular brain regions and the potential for disconnection through DAI. Together, these factors mean that a cognitive function that is primarily subserved by a certain brain region may also be impaired by damage to another area, which may in turn have an effect on the overall function of a large-scale network. Fourth, mapping anatomy- function correlations at group level requires aligning lesions from different individuals into standard brain space for the normalisation of brain size, shape and orientation. This process assumes that functional regions in brains of different individuals are in the same anatomical

locations, whilst, in fact, there is great inter-individual variability in this respect. After TBI, there may be a degree of reconfiguration of functional brain networks, making it even more problematic to infer the ‘normal’ anatomical correlates of a specific cognitive function. Finally, the predominantly fronto-temporal distribution of lesions in TBI may mean that whilst VLSM-type analyses can confirm relationships of cognitive indices with frontal and temporal nodes of distributed networks, they may ignore the contribution of other regions within these networks where no overt lesions are observed. For example, microstructural white matter injury can exist in the absence of damage visible on standard brain imaging after TBI (e.g. Rugg-Gunn et al., 2001); thus, normal scan results do not exclude the possibility that other types of structural abnormality are present and potentially associated with cognitive sequelae.

3.4.6.1 Distributed functional networks support complex cognitive functions. The

network-based approach to brain function, inspired by early connectionist accounts (see the next section), contends that cognitive functions result from the flow of information across large- scale networks made up of distinct cortical regions. It thus follows that cognitive dysfunction can result from damage anywhere within a distributed network, including its connections, and not just from focal damage to discrete anatomical regions (Bartolomeo, 2011; Mesulam, 1998). Conversely, performance on complex cognitive tasks, such as those engaging executive control processes, is unlikely to be supported by a single brain region (Bigler, 2001a).

This approach to understanding cognitive function offers a new perspective to the investigation of brain-behaviour relationships (Bartolomeo, 2011; Catani & ffytche, 2005). For example, the involvement of a fronto-parietal network in executive function has recently received considerable interest, particularly in functional neuroimaging research. Seeley et al. (2007) identified an ‘executive-control network’ linking the dorsolateral frontal and parietal cortices and found that performance on executive tasks correlated with lateral parietalnodes of the network. The historical reference to executive functions as ‘frontal lobe functions’ may, thus, be misleading. Frontal lesions may, of course, be associated with executive impairment after TBI, either because the site of damage is itself critical to the cognitive function or because of damage to the structural connections between the frontal regions and other parts of the network. The current investigation found minimal support for a direct relationship between the anatomical location of focal injury and specific impairments of cognitive function. As noted by Bigler (2001a), the specific anatomical location or extent of focal injury after TBI often does not

correspond to a patient’s profile of neuropsychological deficits.

3.4.6.2 Brain damage in TBI is not limited to the site of impact. Functional ‘nodes’ of

brain networks are connected by long white matter tracts, which are susceptible to DAI, with potential to result in widespread dysfunction. In recent years, the development of more advanced structural neuroimaging techniques, diffusion tensor imaging (DTI) in particular, has started to increase understanding of both the extent of injury to structural connections (see Chapter 4) and how their integrity may relate to cognitive function (see Chapter 5).

Whilst it is possible that some cognitive sequelae of TBI result from damage to particular cortical or subcortical brain regions, depending on the role of that area in specific cognitive functions, the diffuse trauma may play a greater role. Further, despite certain brain regions appearing to be particularly vulnerable to traumatic injury, including those near the frontal and temporal poles, it is highly unlikely that a single brain region is responsible for the frequently observed impairments of complex cognitive functions (Scheid & von Cramon, 2010).

As noted previously, damage to connections within a distributed network can have widespread effects on cognitive function. Wernicke (1885) was the first to argue against strict localisation of function, suggesting that associative connections underlie higher mental functions. However, the greater interest at the time in the cortical localisation perspective meant that Wernicke’s theory fell out of favour until Geschwind revived interest in the study of disconnection through his seminal paper, entitled ‘Disconnexion syndromes in animals and man’ (1965). Geschwind identified a number of syndromes that he believed to result from disconnection caused primarily by lesions of the association cortex, but also lesions of critical white matter tracts (see Catani and ffytche, 2005, for a review). Although he provided a useful framework for studying the clinical and behavioural correlates of anatomically specific lesions, Geschwind did not theorise about differential effects of cortical versus white matter lesions. More advanced structural imaging techniques, particularly DTI, combined with voxel-based data analysis techniques, might give a more detailed picture of the distribution and degree of axonal injury in TBI, enabling further insights into relationships between structural brain damage and cognitive function.

monitoring and management of TBI is the lack of efficient, valid and reliable assessment methods for diagnosis and treatment planning that could be used in everyday practice. This concern is compounded by the heterogeneity of neuropathology associated with TBI. With these issues in mind, Irimia et al. (2011) recently proposed a workflow for the multimodal assessment of TBI in clinical settings, to identify and quantify structural MR abnormalities including extra- and intra-cortical haemorrhages, oedema, focal lesions, and DAI, and to allow correlation of these metrics with outcome variables.

The extent to which a certain level of neuropathology affects cognitive function can also differ across individuals. One possible factor contributing to this could be ‘reserve’, that is, an individual’s estimated maximum cognitive or brain-based capacity (e.g. Brickman et al., 2011; Tate, Neeley et al., 2011). Thus individuals with greater ability or physiological resilience may be better able to absorb some loss without major effects on observed functioning. ‘Cognitive reserve’ has been estimated from years of education or performance on tests of premorbid intellectual function (e.g. Brickman et al., 2011), whilst ‘brain reserve’ has been quantified as total intracranial volume measured from MRI scans (Tate, Neeley et al., 2011). Although Tate, Neeley et al. (2011) did not find total intracranial volume to directly predict a dementia diagnosis in a group of 194 older adults (>65 years), when low volume/reserve was combined with the presence of a genetic risk factor (ApoE-ε4 allele) and male gender, together these three predicted dementia classification. It would be interesting in future research to study whether, and to what extent, cognitive or brain reserve affect the relationships between neurological, cognitive, and functional sequelae of TBI.

3.4.6.4 Associated methodological considerations. The MRI sequences used here

were chosen for their known potential to detect brain contusions and traumatic microbleeds. Other structural neuroimaging techniques, such as FLAIR or DTI, are potentially more sensitive to the presence, location and extent of white matter damage, out of which the role of DTI in TBI will be discussed in detail in the next two empirical chapters.

Although voxelwise methods, such as VLSM, may perform better than simply mapping out visible lesions when the aim is to infer relationships between the anatomical location of brain contusions and cognitive function, they are not without their own problems. Here, the small sample size in particular that was available for the voxelwise analyses reduced statistical power

to detect possible true effects. It was also necessary to use parametric methods (t-tests), even though voxelwise lesion data are unlikely to be normally distributed in such a mixed sample of TBI patients. The nonparametric alternative available in MRIcron (the Brunner-Munzel rank order test) was not used as it is not recommended where there are fewer than 10 patients in each group (Medina et al., 2010).

Investigating the relationships between multiple measures of neuropsychological function and brain structure also raises a multiple comparisons issue. In the present study this was dealt with by applying Bonferroni correction within the family of brain volume indices to account for the multiple between-groups comparisons carried out on normalised volumes of grey matter, white matter and total brain tissue. When correlations were tested between the neuroimaging measures (grey matter volume, white matter volume, total brain volume, lesion volume, and number of microbleeds) and the six cognitive variables, Bonferroni correction was again applied, variable-wise, to account for the multiple tests carried out between each neuroimaging measure and the six cognitive variables. This ensured, in each case, that the familywise error rate for the planned comparisons, both between groups on brain volume and between the neuroimaging and neuropsychological measures, was kept at p < .05. It is acknowledged here, though, that whilst the Bonferroni correction controls for the increased probability of false positives associated with multiple comparisons, this control may come at the cost of increased probability of false negatives and thus reduced statistical power.

Furthermore, the groups in the various analyses were not perfectly matched. Whereas good matching was achieved for age, gender distributions were unequal (as discussed above). Groups also differed in the indices of general intellectual function: whereas the WTAR indicated higher premorbid IQ for controls, patients outperformed controls on current verbal reasoning ability, indexed by WASI Similarities.

Efforts were made to statistically control for these potential confounds, as relevant to each of the separate analyses carried out. Thus, the possible effects of age were controlled for when investigating the relationship between brain volume and cognitive function, because normal aging has been demonstrated in several studies to be associated with structural brain changes including reduced brain volumes (Fjell & Walhovd, 2010). Advancing age has also been shown to contribute to the degree of cognitive impairment observed during the first one to five years post-TBI (Millis et al., 2001). Moreover, people who sustain a TBI at an advanced age

tend to have longer PTA (Sherer

,

Struchen, Yablon, & Nick 2008), and appear to be more cognitively impaired than younger survivors of TBI (Himanen et al., 2006). Schönberger et al. (2009) recently demonstrated that advanced age at the time of TBI is associated with larger grey and white matter lesion volumes and lower grey matter volume, independently of other possible factors such as cause of injury and time since injury. There are also well-established gender differences in brain size, females appearing to have smaller brains (e.g. DeCarli, Massaro et al., 2005). For this reason an additional brain volume analysis was performed including the male participants only, which showed similar results.

Injury severity is also likely to be associated with brain atrophy after TBI. For example, longer PTA duration is associated with larger grey and white matter lesion volumes and smaller residual white matter volume (Schönberger et al., 2009), as well as with overall brain tissue atrophy, measured by increased ventricle to brain ratio (Bigler et al., 2006; Wilde et al., 2006). Similarly, longer time since TBI is associated with larger grey and white matter lesion volumes as well as smaller residual grey matter volumes across several brain regions (Schönberger et al., 2009). Both these variables were controlled for in the present analyses were they were identified as potential confounds.

Although structural abnormalities visible on conventional MRI acutely after TBI have previously been shown to predict symptoms and disability at one year post-injury (e.g. Hiekkanen, Kurki, Brandstack, Kairisto, & Tenovuo, 2009), these relationships are not necessarily apparent at later stages. Scheid and von Cramon (2010) retrospectively analysed structural neuroimaging and neuropsychological data from 320 post-acute/chronic TBI patients, and as here, did not find strong relationships between any particular structural neuroimaging indices and cognitive function.

3.4.7 Conclusions. The purpose of the current chapter was three-fold. Its first aim was