RSNA 2005 

Abstract Archives of the RSNA, 2005


SSK17-04

Digital Bowel Cleansing for CT Colonography Based on Hessian-based Structure-Analysis Method

Scientific Papers

Presented on November 30, 2005
Presented as part of SSK17: Physics (Computer-aided Detection with Colonography)

Participants

Wenli Cai PhD, Presenter: Nothing to Disclose
Michael Ethan Zalis MD, Abstract Co-Author: Nothing to Disclose
Janne Johannes Nappi PhD, Abstract Co-Author: Nothing to Disclose
Hiroyuki Yoshida PhD, Abstract Co-Author: Nothing to Disclose
Gordon Harris PhD, Abstract Co-Author: Nothing to Disclose

PURPOSE

An essential step for visualizing the colonic lumen in fecal tagging CT colonography (CTC) is a digital bowel cleansing that removes residual stool and fluid enhanced by an oral contrast agent. The purpose of this study was to develop and evaluate a novel method for digital bowel cleansing using Hessian-based structure-analysis method.

METHOD AND MATERIALS

Nine patients underwent CTC examinations with reduced cathartic cleansing and orally administered iodine-based fecal tagging. CT scanning was performed in both supine and prone positions with multi-detector CT scanner. In each scan, axial CT slices were interpolated to yield an isotropic volume data set, and the initial colonic wall and the tagged feces were segmented by thresholding based on the histogram of the volume. Folds and polyps covered by the tagged feces were enhanced using Hessian-based structure-analysis method that was designed to extract arciform and spherical structures. The initial colonic wall was deformed based on the structure-analytical response value to obtain the final colonic wall. The mucosa covered by tagged feces was reconstructed to recover the smooth transition from the segmented colonic wall and the luminal air. The resulting electrically-cleansed colon was subjected to our computer-aided detection (CAD) of polyps that was trained on 121 cleansed CTC cases to yield a 95% by-polyp sensitively with 1.5 FPs per dataset. The CAD output was compared with that obtained from the previously-reported digital bowel cleansing method.

RESULTS

Our CAD scheme yielded an average of 5.4 FPs per dataset when it was applied to the colon obtained by the previous bowel cleansing method. The average number of FPs was reduced by 35%, resulting in 3.5 FPs per dataset, when the new bowel cleansing method was used. All (2) colonoscopy-confirmed polyps were detected by CAD with either cleansing method. The reduction of FPs occurred because of the reduced subtraction artifacts as well as the reduction of incomplete cleansing due to inhomogeneous tagging.

CONCLUSION

Our new digital bowel cleansing method can provide a cleansed colon with substantially reduced subtraction artifacts and incomplete cleansing.

Cite This Abstract

Cai, W, Zalis, M, Nappi, J, Yoshida, H, Harris, G, Digital Bowel Cleansing for CT Colonography Based on Hessian-based Structure-Analysis Method.  Radiological Society of North America 2005 Scientific Assembly and Annual Meeting, November 27 - December 2, 2005 ,Chicago IL. http://archive.rsna.org/2005/4419429.html