Accuracy of artificial intelligence on histology prediction and detection of colorectal polyps

Leung_headshot Post written by Wai K. Leung, MD, from the Department of Medicine, Queen Mary Hospital, University of Hong Kong, Hong Kong.

This study reviewed and summarized the latest data on the use of artificial intelligence on histology prediction and detection of colorectal polyps. With the rapid development and increasing application of artificial intelligence (AI) in GI endoscopy, particularly for colorectal polyps, there is a pressing need to systematically review the current literature on the role of AI on colorectal polyp characterization and detection. The results would help to guide the future directions on how AI could integrate with routine colonoscopy services as well as the research gap on AI-assisted colonoscopy.

Based on the results of 18 studies, we found that the accuracy of AI on polyp histology prediction was 0.96, with a pooled sensitivity of 92.3% and specificity of 89.8%. Studies using AI with narrow-band imaging were superior to AI with white light for histology prediction (accuracy 0.98 vs 0.84, P<0.01). The negative predictive value of AI on histology prediction of diminutive polyps was 0.91. AI was also found to be superior to non-expert endoscopists on polyp histology prediction (accuracy 0.97 vs 0.90, P <0.01). Six studies were included in the meta-analysis of polyp detection by AI with an accuracy of 0.90. The pooled sensitivity was 95.0% and specificity was 88.0%.

It is interesting to note that despite the difference in AI models and study design, the AI performances are rather consistent on polyp histology prediction and detection. However, most of the included studies were retrospective in nature and could be subject to various biases. Future prospective multi-center studies are needed to verify the performance of AI on colorectal polyp characterization and detection.


Figure 5. The summary receiver operating characteristic curves for the diagnostic performance of artificial intelligence (AI) and nonexpert endoscopists.

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.

Leave a Comment

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s