Department of Electrical & Computer
                                  Engineering
                            Old Dominion University
                                GRADUATE SEMINAR
                             Electrical Engineering
                      Attacks Biology with Image Analysis
                                       by
                               Dr. Scott T. Acton
               Department of Electrical and Computer Engineering
                             University of Virginia
                              Charlottesville, VA
                           Friday, February 10, 2012
                                3:00 p.m. KH 224
                          Host: Dr. Khan Iftekharuddin
The first part of the talk describes our work in tracking leukocytes (white
blood cells) in vivo. Active contour methods are highlighted. Specifically, the
Poisson inverse gradient method of contour/surface initialization and the
vector field convolution method of contour evolution are presented. Results in
leukocyte studies, as well as anatomical studies show the efficacy of these
approaches.

The second portion of talk depicts ongoing work in neural image analysis. The
knowledge of neuron structure is a central part of the understanding of the
brain. Towards this end, development of a library describing the shapes and
connectivity of neurons is necessary. Such an atlas is called the neurome.
Construction of this library for a range of organisms will require automated
image analysis. 

The talk describes our proposed automatic segmentation method, Tree2Tree, and
our proposed neuron matching method, Path2Path. Tree2Tree is a graph based
algorithm that extracts the complex shape of the neuron from 3D intensity
images of neurons. The challenge lies in computing a consistent estimate of the
neuronal branching and connectivity from the low contrast images, characterized
by ambiguous edge information and inconsistent brightness patterns. Image
processing and analysis methods associated with Tree2Tree, and the path-based
matching of Path2Path, will be described. Preliminary results show that the
Tree2Tree algorithm segments the neuron with high sensitivity and accuracy and
that Path2Path may hold promise in matching neurons based on morphology,
hierarchy and spatial position.

More details at http://www.ece.odu.edu