Digital morphometric characterization of mucosal surface lesion patterns under magnification colonoscopy

To study quantified mucosal surface patterns of digitized colonoscopic images obtained with 17-40x magnification and chromoscopy of hyperplastic, adenomatous lesions and adenocarcinomas confirmed by histology.

Morphometric measures were performed on at least 10 image files of each lesion and compared to normal mucosa at the same distance and area from each image. Thirty-seven hyperplastic lesions, 42 adenomas and 26 carcinomas from 105 patients were studied. Three morphometric characteristics were evaluated in lesions and normal mucosal surface area from each case: pit counts per area, pit diameter and pit area.

Pit counts per area and pit area of lesions as compared to the normal mucosa of the same cases showed statistically significant differences in the 3 groups of lesions studied (p < 0.001).

Real-time morphometric evaluation is feasible during videocolonoscopy with magnification roughly equivalent to a 5x microscopic field. Automated evaluation of morphometric characteristics as part of Bayesian Belief Network software may prove a useful information source for the colonoscopist's diagnostic decision.