Title: New Algorithmic Approaches for Problems in Biomedical Imaging

Speaker: Danny Ziyi Chen, Notre Dame

Time/Date:  Friday, Apr 13, 11-12
Location: Room 3584

Abstract:

New image acquiring modalities and technologies continue to 
revolutionize the fields of biological studies and medical care today. 
Quantitative biomedical image analysis plays 
a critical role in solving many problems that are faced by biologists, 
cardiologists, orthopedists, radiologists, and other physicians on the daily 
basis. The increasing sizes of image data, especially in 3-D and higher 
dimensions, present a significant challenge to conventional approaches for 
automated biomedical image analysis, which has often been a difficult or even 
unrealistic process due to its time-consuming and labor-intensive 
characteristics. 

Image segmentation, the problem of identifying objects of interest in 
volumetric image data, is a central problem in biomedical image analysis 
and computer-aided diagnosis.  Robust, efficient, and accurate automated 
segmentation methods are highly desirable for numerous biomedical studies 
and applications. In this talk, we present effective image segmentation 
techniques for detecting biomedical objects in 3-D and higher dimensional 
images.  The techniques are based on graph search frameworks or computational 
geometry paradigms.  In comparison with most known segmentation methods 
for volumetric images that suffer from their inability to attain optimal segmentation 
or lengthy computation time, our techniques produce, in an efficient manner, 
segmentation of optimal quality with respect to general cost functions 
on a wide range of biomedical objects with complex topological structures.
Segmentation results computed by our techniques on various biomedical 
objects (e.g., pulmonary fissures, airway trees, aorta, coronary vessels, 
retina, knee cartilage, blood clots, etc) and different imaging modalities 
are shown.