VASC Seminar Announcement
            =========================

Date: Monday, 5/8/00
Time: 3:45-4:45
Place: NSH 3002

Speaker:  George Stetten
          CMU RI/U. Pitt Bioengineering
          http://www.stetten.com

Title: Fuzzy Medial-Based Segmentation using Truncated Wedges

Abstract: 

Image segmentation is generally regarded as the partitioning of space at the boundary of an object to 
represent the object's shape, pose, and topology. Some images, however, contain so much noise that 
distinct boundaries are not forthcoming even after the object has been identified. For example, we 
have used statistical methods based on medial features in Real Time 3D echocardiography to locate 
the left ventricular axis of the heart, even though the precise boundaries of the ventricle are simply not 
visible in the data. After locating the ventricular axis in this manner, we produce a fuzzy labeling of 
ventricular voxels to represent the shape of the ventricle without any explicit boundary. The fuzzy 
segmentation permits calculation of total ventricular volume as well as determination of local boundary 
equivalencies, both of which can be validated against manual tracings. 

The fuzzy segmentation method uses a medial-based compartmentalization that is generalizable to 
any shape. Each compartment forms a "truncated wedge" whose volume can be determined by 
integrating a function of the principal curvatures of the boundary patch and the distance to the 
medial manifold. The method is an extension of the lay concept of mining rights, by which the 
interior of the earth is parceled out to patches of real estate on the surface. This concept can be 
extended to any shape in m>2 dimensions, independent of topological genus. 

Papers on-line: 

Stetten, "Volume of Arbitrary Shapes from Boundary Curvature and Medial Scale," Technical Report 
# TR99-001, Department of Computer Science, University of North Carolina at Chapel Hill, Jan, 1999. 
http://www.stetten.com/george/publications/TR99_001.pdf 

Stetten and Pizer, "Medial Node Models to Identify and Measure Objects in Real-Time 3D 
Echocardiography," IEEE Transactions on Medical Imaging, Oct. 1999. 
http://www.stetten.com/george/publications/TMI_Stetten_99.pdf 

Stetten, "Automated Identification and Measurement of Cardiac Anatomy via Statistical Analysis of 
Medial Primitives," Ph.D. Dissertation, UNC, Chapel Hill., (see Chapters 7 and 8). 
http://www.stetten.com/george/publications/phd