| الوصف: |
Background Multiple sclerosis (MS) is a highly variable disease of the central nervous system with inflammatory and neurodegenerative components, associated with both physical and cognitive disability. Abnormalities are visible on routine magnetic resonance imaging (MRI) of the brain, with 'white matter hyperintensities' (WMHs) representing sites of previous inflammation. Techniques for measuring WMHs have not been standardised, although manual outlining is conventionally taken to be the reference standard, despite its subjective element. WMHs have been found to only partly explain the degree of cognitive impairment, forming part of the 'clinico-radiological paradox'. Research interest has largely moved to advanced imaging techniques, one such technique being diffusion tensor imaging (DTI). Through sensitivity to water molecule movement, DTI reflects the integrity of white matter tracts and thus its measures may be relevant to both the inflammatory and degenerative disease components. Aims The work described in this thesis aims to improve our understanding of the true relationship between measures of white matter damage and cognitive impairment in people with MS, to determine the optimum measurement technique(s) for quantifying WMHs, including developing and testing a novel visual rating scale, and to assess whether information provided by DTI can strengthen the association of imaging and clinical findings. Methods A systematic review of the literature and meta-analysis relating WMHs to cognition was conducted, focussing on image analysis technique. Three separate methods for quantifying WMHs were then investigated. The reproducibility of manual outlining was assessed using scans available from 43 people with secondary progressive MS (SPMS). An automated software method was optimised for the same cohort, based on the results of the manual outlining. A novel semi-quantitative visual rating scale was developed, with validation using the same scans within a larger, more varied cohort. All available information regarding the participants studied was then used to construct a linear regression model predicting cognitive outcomes and determining the utility of the various imaging markers derived from conventional imaging techniques. A non-linear relationship for WMHs was also considered. White matter DTI metrics in the same smaller cohort of 43 people were then investigated, primarily considering tissue outwith WMHs, as well as that within major tracts and the novel diffusion marker `peak width of skeletonised mean diffusivity'. The additional explanatory power of DTI metrics within the linear models developed previously was then determined. Results High variability was found in the literature regarding imaging marker measurement and reporting of technique reproducibility. Manual outlining was found to be associated with considerable measurement error, dependent on observer and cohort factors. It was possible to optimise the automated software for a particular cohort, either for volumetric or spatial outputs. Visual rating of MS imaging features was found to be feasible and measures of WMH burden were closely related to fully quantitative measures. The overall association of WMHs to cognitive function was similar to that found in the published literature, with no additional association following addition of DTI metrics. A trend towards a greater effect of WMH volume at higher levels was found, consistent with a non-linear relationship between imaging metrics and cognitive phenotype. Conclusions Substantial heterogeneity in the reporting of the reproducibility of WMH measurement supports a move towards benchmarking against reference datasets. Poor reliability of the current reference standard, manual segmentation, should be recognised as a key limitation for the field. Rich information can be captured quickly using visual rating of imaging features. The close correlation of visual ratings of WMHs with quantitative measures may represent a practical alternative in the appropriate circumstances. Combining visual rating features provided additional explanatory power, supporting a multidimensional substrate for the cognitive phenotype. Finally, both automated and visual rating analyses support a non-linear relationship between disease burden and cognitive performance in MS. |
