Seminar 2008 05 13 Computational Intelligence
From CISSTwiki
CISST ERC Seminar
Computational Intelligence Supporting Anatomical Shape Analysis and Computer-Aided Diagnosis
Date: Tuesday, May 13, 2008
Time: 11:00am
Place: CSEB B17 (Lunch will be served at 12:00pm)
Speaker: Aaron Ward
Queen's University
Title: Computational Intelligence Supporting Anatomical Shape Analysis and Computer-Aided Diagnosis
Presentation: PDF, not yet uploaded
Abstract
Medical imaging technologies such as magnetic resonance imaging (MRI), computed tomography (CT), and ultrasound (US) provide useful views of internal body structures. Since the development of these imaging technologies, there has been an extensive effort to process the data they provide to transform it into higher-level information useful to medical professionals involved in diagnosis, intervention, and research.
One useful form of such information is the shape of anatomical structures. Based on good descriptions of shapes of structures of interest, medical researchers can learn about relationships between shape and pathology/injury in the population and diagnosticians can benefit from computer-aided diagnosis based on shape.
A system transforming medical image data into useful shape information requires solutions to several problems. First is the problem of segmentation, where the anatomical structure of interest is isolated and digitally extracted from the surrounding data in the medical image. Second is the problem of shape description, where the shape of the structure is encoded in some form that should permit effective qualitative and quantitative analysis. Third is the problem of shape correspondence, where a map is computed between parts of objects in a group; such a map leads to the ability to perform shape analysis on local shape regions. Satisfactory approaches to these problems make it possible to compute useful observations in the form of statistics on groups of structures, to provide effective visualizations of observed shape variability, and to train automatic classifiers to distinguish pathology groups based on structural shape.
This talk describes research into solutions to the problems of shape description and shape correspondence. It also describes medically-oriented research into the relationship between shape and pathology of musculoskeletal structures in the human shoulder. The central theme of this research is the use of computational intelligence techniques such as graph matching, machine and manifold learning, feature selection, and optimization to support anatomical shape analysis and computer-aided diagnosis.
Bio
Not Yet Entered