Please use this identifier to cite or link to this item:
|Scopus||Web of Science®||Altmetric|
|Title:||HOMOR: Higher Order Model Outlier Rejection for high b-value MR diffusion data|
|Citation:||Neuroimage, 2012; 63(2):835-842|
|Publisher:||Academic Press Inc Elsevier Science|
|Kerstin Pannek, David Raffelt, Christopher Bell, Jane L. Mathias, Stephen E. Rose|
|Abstract:||Diffusion MR images are prone to artefacts caused by head movement and cardiac pulsation. Previous techniques for the automated voxel-wise detection of signal intensity outliers have relied on the fit of the diffusion tensor to the data (RESTORE). However, the diffusion tensor cannot appropriately model more than a single fibre population, which may lead to inaccuracies when identifying outlier voxels in crossing fibre regions, particularly when high b-values are used to obtain increased angular contrast. HOMOR (higher order model outlier rejection) was developed to overcome this limitation and is introduced in this study. HOMOR is closely related to RESTORE, but employs a higher order model capable of resolving multiple fibre populations within a voxel. Using high b-value (b=3000 s/mm2) diffusion data from a population of 90 healthy participants, as well as simulations, HOMOR was found to identify a decreased number of outlier voxels compared to RESTORE primarily within areas of crossing, bending and fanning fibres. At lower b-values, however, RESTORE and HOMOR give similar results, which is demonstrated using diffusion data acquired at b=1000 s/mm2 in a mixed cohort. This study demonstrates that, although RESTORE is suitable for low b-value data, HOMOR is better suited for high b-value data.|
|Keywords:||Brain; Humans; Diffusion Magnetic Resonance Imaging; Artifacts; Brain Mapping; Movement; Image Processing, Computer-Assisted; Adolescent; Adult; Aged; Middle Aged; Female; Male; Young Adult|
|Rights:||© 2012 Elsevier Inc. All rights reserved.|
|Appears in Collections:||Psychology publications|
Files in This Item:
There are no files associated with this item.
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.