QBI Seminar: Diffeomorphic metric mapping, brain shape analysis and imaging genetics

Event Details

Date:

Wednesday, 27 May 2015 - Wednesday, 27 May 2015

Time:

12:00 pm - 1:00 pm

Room:

QBI Level 7 Auditorium

UQ Location:

Queensland Brain Institute (St Lucia)

URL:

http://www.qbi.uq.edu.au/neuroscience-seminars

Event category(s):

Full Description:: Xianfeng Yang
Queensland Brain Institute, University of Queensland
Title: Diffeomorphic metric mapping, brain shape analysis and imaging genetics
Abstract:
Large deformation diffeomorphic metric (LDDMM) is a reference transformation model in computational anatomy which has been shown as an effective computational paradigm to measure variability of complex brain anatomy. However, its high computational complexity prevents the broad use in clinical and research studies. Many efforts trying to speed up LDDMM fall short of the validity under large deformation setting. To overcome such limitation, we proposed a rigorous formulation based on one-parameter subgroup of diffeomorphism in context of LDDMM. This method can significantly increase registration speed without compromising accuracy, and can improve the performance of metric based shape analysis as well. Capability of this approach is demonstrated in imaging genetics studies related to Alzheimer's disease. By hippocampal shape phenotypes derived from this approach, we have revealed the longitudinal effects of AD risk genes on hippocampal degeneration from early to later life. These shape phenotypes also led to discovery of novel risk loci of Alzheimer's disease which are in line with previous biochemical findings. In contrast, traditional approaches such as volumetric analysis, voxel-wise analysis and PCA cannot identify these effects.