Post written by Yuichi Mori, MD, PhD, from the Clinical Effectiveness Research Group, University of Oslo, and the Department of Transplantation Medicine, Oslo University Hospital, Oslo, Norway, and the Digestive Disease Center, Showa University Northern Yokohama Hospital, Yokohama, Japan.

The primary aim of this study was to focus attention on the major challenges that computer-aided prediction of histology faces.

With the increased introduction of sophisticated computer-aided detection tools in colonoscopy and mammography, a concern arises in relation to overdiagnosis and overtreatment in cancer screening programs.

Computer-aided prediction of histology would be a welcome mitigation measure to reduce overtreatment by identifying what should be removed and what should not be removed. However, there is a serious criterion standard challenge in the development process of computer-aided prediction tools that has never been stressed in the research community.

This Thinking Outside the Box article clarifies this challenge and explores solutions that would enable optimal development of a reliable computer-aided prediction tool. Overcoming this challenge is quite important because the use of computer-aided prediction of histology is expected to be the standard of care in line with the rapid introduction of computer-aided detection tools in colonoscopy practice and colorectal cancer screening.

Concept of semisupervised learning for endoscopic images with uncertain pathologic diagnoses. Unlabeled data (images without reliable pathologic diagnoses) are automatically classified into clusters with use of the information obtained from labeled data (images with reliable pathologic diagnoses) and similarity of the images between each data.

Read the full article online.

The information presented in Endoscopedia reflects the opinions of the authors and does not represent the position of the American Society for Gastrointestinal Endoscopy (ASGE). ASGE expressly disclaims any warranties or guarantees, expressed or implied, and is not liable for damages of any kind in connection with the material, information, or procedures set forth.