1-Page Late Breaking Paper (Poster)
09:05 - 09:55 | Thu 16 Feb | Ballroom D | ThRAF
Accurate prediction of polyp histology during colonoscopy allows endoscopists to implement resect-and -discard or diagnose-to-leave strategies for diminutive colorectal polyps (≤ 5mm) and make on-site recommendation for the next surveillance interval, saving time and cost. This study presents the evaluation of a computer-aided (CAD) method transferring low-level Convolutional Neural Network (CNN) features learned from non-medical domain for classification of colorectal polyps evaluated based on Preservation and Incorporation of Valuable Endoscopic Innovations (PIVI) initiative, of which the objectives are to identify clinical importance relating to endoscopic technologies. Two preliminary experiments were conducted according to the two statements of PIVI initiative and the proposed CAD method resulted in great potential for real-time endoscopic assessment use.
No information added