Meeting Banner
Abstract #0170

Metrics that Matter: Improved statistical power to detect differences in tissue microstructure through dimensionality reduction

Maxime Chamberland1, Erika Raven1, Sila Genc1,2,3, Kate Duffy1, Greg Parker1, Chantal M.W. Tax1, Maxime Descoteaux4, and Derek K. Jones1,5

1School of Psychology, Cardiff University Brain Research Imaging Centre, Cardiff, United Kingdom, 2Department of Paediatrics, The University of Melbourne, Parkville, Australia, 3Developmental Imaging, Murdoch Childrens Research Institute, Parkville, Australia, 4Computer Science, University of Sherbrooke, Sherbrooke, QC, Canada, 5School of Psychology, Australian Catholic University, Victoria, Australia

Various diffusion metrics have been proposed for characterising tissue microstructure. However, it is unclear which metric best captures individual microstructural differences. One possible approach is to explore all available metrics. However, this increases the chance of Type I error and makes interpretation difficult. Using data-reduction approaches, we identified two principal components that capture 85% of the variance in diffusion measurements. The first captures properties related to hindrance, while the second reflects tissue complexity. We demonstrate the utility of this approach by showing significant correlations with age of these new metrics, whereas little to no effects were observed with any individual metric.

This abstract and the presentation materials are available to members only; a login is required.

Join Here