You are invited to come hear a lecture given by Dr. Jiang Li, Ph.D. in the Doppman Conference Room.Ê The talk is entitled: ÒMachine Learning Methodologies in CT Virtual Colonoscopy Computer-Aided Colonic Polyp DetectionÓ.
Ê
ÊÊÊÊÊÊÊÊÊÊÊÊÊÊÊ
When:Ê July 6, 2006
Ê
Where:Ê Doppman Conference Room
(Located in the Diagnostic Radiology Department, 1C520)
Ê
Time:Ê 2:00 pm to 3:00 pm
Ê
Ê
Abstract:
Ê
Although colon cancer is the second leading cause of cancer death in the US, it is also one of the most treatable forms of cancer. Early detection of colon polyps is the best way to approach treatment. Computed tomographic colonography (CTC) computer aided detection (CAD) system is a new tool in development to detect colon polyps in virtual colonoscopy. The CTC CAD system developed in our group first segments colon using a region growing algorithm, and the colon surface is extracted by an iso-surface technique. Polyp candidates along the colon surface are then identified and segmented. We calculate more than 100 quantitative features for each of the segmented candidates. These features are then run through a feature selection procedure, and the selected features are finally used to classify the polyp candidates by a committee of support vector machines. A large population study showed that our system achieved fairly high performance for detecting larger polyps, but more challenges should be tackled when the goal is to locate the so-called "medium"-sized polyps, 6-9mm in diameter. During the presentation, I will briefly introduce the system diagram and some methodologies for improving the system. In particular, I will present a multiobjective optimization scheme to improve the initial polyp detection on the colon surface, a wavelet analysis method for false positive reduction, and an efficient algorithm for the feature selection.
Ê
Ê
Ê
Ê
for further information contact:
Ê
Sharon Robertson
Program Support Assistant
DRD/CC/NIH
9000 Rockville Pike
10 Center Drive
Bldg. 10/Room 1C655B, MSC 1182
Bethesda, MDÊ 20892-1182
Phone: 301-402-5076
FAX: 301-496-9933
E-mail: robertsons@cc.nih.gov
Ê