G07 Reliability of diffusion tensor imaging measures

Abstract

Background Huntington9s disease (HD) has been associated with white matter (WM) abnormalities, evident using diffusion tensor imaging (DTI) analysis. However, the limited longitudinal work in this area has been inconclusive and further research into the progressive WM degeneration in HD is required. A key factor in the validity of results from longitudinal DTI research is the reliability of these measurements, which is not well-established. Aim To assess the test-retest reliability of DTI-based measures of brain structure. Methods Using repeat DTI data from 12 participants scanned at the London and Paris sites of the PADDINGTON study (scan interval range 0–24 days), the reliability of measures derived from different analysis methods was assessed by calculating intra-class correlations coefficients (ICC) for each measure. Results Fractional anisotropy (FA) across the whole brain showed a mean reliability of ICC=0.79. The regional distribution of reliability was also mapped, using voxelwise ICC, with WM regions showing higher ICC compared to grey matter or cerebro-spinal fluid. Tensor-based atlas registration showed substantial variation in reliability between different WM regions, however, assessed across all regions, reliability was high for FA (ICC=0.98) and the trace of diffusivity (ICC=0.98). Using manually delineated regions-of-interest, reliability was high for WM (ICC=0.96) and both the caudate (ICC=0.96) and putamen (ICC=0.97), while corpus callosum reliability was lower (ICC=0.83). Comparable measures of reliability were found for the trace of diffusivity. Conclusions DTI scanning generally produces highly reliable measures of brain structure. There is considerable heterogeneity in reliability between different brain regions and imaging methods. These results will be informative when designing longitudinal research to map progressive brain changes in HD and other degenerative disorders.