Please use this identifier to cite or link to this item: http://scholarbank.nus.edu.sg/handle/10635/136511
Title: MULTISCALE KERNELS FOR DIFFEOMORPHIC BRAIN IMAGE AND SURFACE MATCHING
Authors: TAN MING ZHEN
Keywords: Image Processing, Brain Registration, Multiscale Kernels
Issue Date: 20-Jan-2017
Source: TAN MING ZHEN (2017-01-20). MULTISCALE KERNELS FOR DIFFEOMORPHIC BRAIN IMAGE AND SURFACE MATCHING. ScholarBank@NUS Repository.
Abstract: This thesis focuses on the multiscale based improvements of the large deformation diffeomorphic metric mapping (LDDMM) framework, with specific emphasis on human brain registration. Motivated by the concept of multiscale Gaussian reproducing kernels and their proven efficacy in mapping objects of multiple scales of variation, we discuss two novel multiscale implementations for the LDDMM framework, with applications in both surface and volume-based matching functionals. In the first method, we demonstrated marked improvements in accuracy and computational costs via a coarse-to-fine multiresolution surface registration framework that utilises a mixture of weighted Gaussian kernels. The method incorporates sparsity without the use of augmented sparse-penalty functions and was also shown to alleviate the undesired in-folding effects of the popular currents surface matching functional. In the second method, we proposed the use of frame-based reproducing kernels that construct hierarchical multiscale structures based on the multiresolution analysis framework for wavelets. These reproducing kernels are easy to tune and demonstrate improved accuracy for image-based registrations involving multiple scales of variation. Importantly, these frame-based kernels were shown to be feasible for practical integrated whole-brain registration frameworks involving anatomical landmarks, manually-delineated curves and surface representations alongside volume-based brain mapping.
URI: http://scholarbank.nus.edu.sg/handle/10635/136511
Appears in Collections:Ph.D Theses (Open)

Show full item record
Files in This Item:
File Description SizeFormatAccess SettingsVersion 
TanMZ.pdf11.17 MBAdobe PDF

OPEN

NoneView/Download

Page view(s)

17
checked on Jan 14, 2018

Download(s)

18
checked on Jan 14, 2018

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.