Analysis of Longitudinal Shape Variability via Subject Specific Growth Modeling

The Faculty of Informatics is pleased to announce a seminar given by Guido Gerig

DATE: Tuesday, June 11th 2013
PLACE: USI Lugano Campus, room SI-006, Informatics building (Via G. Buffi 13)
TIME: 10.30

ABSTRACT:
Statistical analysis of longitudinal imaging data is crucial for understanding normal anatomical development as well as disease progression. Availability of longitudinal, serial 3D imaging spawns of a new class of image processing and analysis methodologies centered about spatio-temporal analysis of individual subjects and population-based analysis of sets of 4D trajectories encoding anatomical changes. The fundamental analysis task is challenging due to the difficulty in modeling longitudinal changes, such as growth pathological change trajectories, and comparing changes across different populations. Our approach for analyzing shape variability over time and for quantifying spatiotemporal population differences estimates 4D anatomical growth models for a reference population (an average model) and for individuals in different groups, measuring shape variability through diffeomorphisms that map the reference to the individuals. Conducting our analysis on this 4D space enables statistical analysis of deformations as they are parameterized by momenta vectors that are located at homologous locations in space and time. Alternatively, we also develop a generative model which describes shape change over time by extending simple linear regression to the space of shapes represented as "currents" in the large deformation diffeomorphic metric mapping (LDDMM) framework. A crucial aspect in our shape modeling is the notion of correspondence-free shape matching via "currents", to overcome limitations in landmark selection on complex 3D shapes or sets of 3D shapes.
We evaluate our method on a synthetic shape database and data from two clinical studies that seek to quantify brain growth differences in infants at risk for autism and evaluate neurodegeneration in normal aging and Huntington's disease, respectively. The presented  methodologies are generic and will find applications in a wide range of areas where we would like to extract spatiotemporal models from  time-discrete image data.

BIO:
Guido Gerig received his Ph.D. in 1987 from the Swiss Federal Institute of Technology, ETH Zurich, Switzerland, mentored by Prof. Olaf Kuebler who later became president of ETH/EPFL. He continued research in image analysis at ETH as a postdoc,with a habilitation and as assistant professor at the EE department. In 1998, he was appointed Taylor Grandy Professor at the University of North Carolina at Chapel Hill with joint appointments in the departments of Computer Science and Psychiatry.
Guido Gerig joined the SCI Institute at the University of Utah in 2007, where he has a faculty position at the School of Computing, with adjunct positions at the departments of Bioengineering and Psychiatry. He serves as associate director of the SCI Institute and the director of the Utah Center for Neuroimage Analysis (www.ucnia.org). He was appointed as a fellow of the American Institute for Medical and Biological Engineering (AIMBE) in 2010 and is a fellow of the MICCAI society.

Guido Gerig conducts research in medical image analysis driven by several clinical neuroimaging. Methodologies include image pre-processing, registration, atlas building, segmentation, shape analysis, and statistical analysis of shapes and appearances. Current key research topics are segmentation of volumetric image data, spatiotemporal shape modelling from longitudinal image data, building of population atlases of volumetric images and embedded structures, and new methodologies for statistical analysis of diffusion tensor imaging (DTI). Driving clinical problems are studies of early brain development in subjects at risk for mental illness, longitudinal infant studies of autism, assessment of pathology and changes due to  therapeutic intervention in traumatic brain injury (TBI), and analysis of anatomical shape changes during pre-clinical Huntington's disease. New tools and methods are freely shared with the scientific community as open source, open platform software (ITK) and made available to public via the NIH NITRC resource.

HOST: Prof. Kai Hormann