Premium
Adaptive deformable model for colonic polyp segmentation and measurement on CT colonography
Author(s) -
Yao Jianhua,
Summers Ronald M.
Publication year - 2007
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.2717411
Subject(s) - segmentation , artificial intelligence , voxel , sørensen–dice coefficient , image segmentation , computer science , computer vision , nuclear medicine , medicine
Polyp size is one important biomarker for the malignancy risk of a polyp. This paper presents an improved approach for colonic polyp segmentation and measurement on CT colonography images. The method is based on a combination of knowledge‐guided intensity adjustment, fuzzy clustering, and adaptive deformable model. Since polyps on haustral folds are the most difficult to be segmented, we propose a dual‐distance algorithm to first identify voxels on the folds, and then introduce a counter‐force to control the model evolution. We derive linear and volumetric measurements from the segmentation. The experiment was conducted on 395 patients with 83 polyps, of which 43 polyps were on haustral folds. The results were validated against manual measurement from the optical colonoscopy and the CT colonography. The paired t ‐test showed no significant difference, and the R 2 correlation was 0.61 for the linear measurement and 0.98 for the volumetric measurement. The mean Dice coefficient for volume overlap between automatic and manual segmentation was 0.752 (standard deviation 0.154).