Meeting Banner
Abstract #3576

Regularisation of Fractional Anisotropy Using Neighbourhood Information

Marta Morgado Correia1, Virgina FJ Newcombe1, Thomas Adrian Carpenter1, Guy B. Williams1

1Wolfson Brain Imaging Centre, University of Cambridge, Cambridge, UK

Notwithstanding its successes, the traditional tensor model for diffusion imaging is a very simple model which ignores the uncertainty associated to the data caused by noise and partial volume averaging. In this study we propose the use of Bayes decision rule in a regularisation algorithm which takes into consideration this uncertainty and aims at producing more reliable and robust measures of fractional anisotropy (FA). Results show that the proposed method reduces the variability observed between voxels belonging to the same population of fibres, and it increases FAs ability to differentiate between tissue types.