From pjrhusbands at lbl.gov  Mon Jul 23 16:37:16 2007
From: pjrhusbands at lbl.gov (Parry Husbands)
Date: Mon Jul 23 16:38:52 2007
Subject: [BANANA] Berkeley Lab - Scientific Computing Seminar - Thursday,
 July 26, 2007
Message-ID: <op.tvxykefg88l0aa@talkabout.lbl.gov>

Date: Thursday, July 26, 2007
Time: 11:00am-12:00pm
Location: Building 50A-5132

Seminar Speaker: Heng Huang
                  Department of Computer Science and Engineering
                  University of Texas at Arlington
                  http://ranger.uta.edu/~heng/

Title: Shape Modeling and Alignment: Applications to
        Biomedical Informatics

Abstract:

Shape modeling and surface representation combine physical measurement
of objects with mathematical models and are important in a large number
of scientific and engineering areas. Biomedical informatics is one of
the most important applications. In this talk, I will present two
recent results on shape modeling and registration with several
applications in biomedical informatics.

In the first part, I will show a novel surface registration algorithm
for arbitrarily shaped but simply connected (i.e., contiguous) 3D objects.
Based on the rotational properties of harmonic analysis, a new
Parameterization Rotation Theorem is introduced and proved and a fast
surface alignment algorithm is developed. This algorithm can accurately
and efficiently generate surface correspondences between objects for
which spherical harmonics (SPHARM) was used for surface modeling. Its
computational complexity is improved from the O(n^3) of previous methods
to O(n^2). These techniques have been applied to many biomedical  
applications,
including improving morphological understanding of anatomic structures,
dynamic shape analysis (e.g. of beating hearts), cardiac electromechanical
model, protein-protein docking, and assisting in medical tasks like
pacemaker placement.

In the second part, I will present a new hemisphere-based harmonic
surface model. I define a set of complete hemispherical harmonic basis
functions on a hemisphere domain and use these to support a novel  
parametric
shape description method. This method is very useful in efficiently and
flexibly representing the surfaces of certain anatomical structures (like
heart ventricles) in biomedical applications.

Heng Huang received his Ph.D. degree in Computer Science at Dartmouth
College in 2006. Before he joined University of Texas at Arlington (UTA),
he worked as a postdoctoral researcher at the Center for Biological and
Biomedical Computing in the Biology department at Dartmouth College.
Currently he is an assistant professor in the computer science and
engineering department at UTA and leads the BIOVIZON (biomedical computing
and visualization) Lab, where his research focus is on the areas of  
biomedical
image analysis, bioinformatics, pattern recognition, and scientific  
visualization.

Sponsor of Seminar: Chris Ding