Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.666788
Title: Correcting for motion between acquisitions in diffusion MR imaging
Author: Bai, Y.
Awarding Body: University College London (University of London)
Current Institution: University College London (University of London)
Date of Award: 2009
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Abstract:
The diffusion tensor (DT) and other diffusion models assume that each voxel corresponds to the same anatomical location in all the measurements. Movements and distortions violate this assumption and typically the images are realigned before model fitting. We propose a set of model-based methods to improve motion correction and avoid the errors that the traditional method introduces. The new methods are based on a three-step procedure to register DWI datasets, and use different reference images for DWIs with different gradient directions for registration, so the registrations take into account the contrast differences of measurements. Performance of the model-based registration techniques depends critically on outlier rejection. We develop new methods for fitting the diffusion tensor to diffusion MRI measurements in the presence of outliers by drawing on the RANSAC algorithm from computer vision. We compareone popularly used outlier rejection method RESTORE in the diffusion MRI literature with our new method. Then, we combine outlier rejection methods with model-based registration schemes, and compare the performance of motion correction with other methods. After aligning the dataset, we also update diffusion gradients for the registered datasets from both traditional and our methods, according to the transformations used in registrations. We develop and discuss a variety of registration evaluation methods using both synthetic and human-brain diffusion MRI datasets. Experiments demonstrate both quantitative and qualitative improvements using our new model-based methods.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.666788  DOI: Not available
Share: