The Impact of AI on Quality Metrics in Screening Colonoscopy

With the advent of artificial intelligence (AI) comes a new dimension in the fight against colorectal cancer. With increasing incidence rates among individuals under the age of 50, there is the challenge to identify new strategies to prevent the development of colorectal cancer through earlier detection. While still relatively early, AI’s rapid integration is reshaping not only the field of gastroenterology but also the approach to screening colonoscopies. 

A vast amount of data, particularly on the use of computer aided detection systems (CADe) has highlighted just how real time polyp detection and characterization may improve quality metrics in screening colonoscopies. These technologies aim to enhance endoscopy by increasing lesion recognition and improve outcomes.

A systematic review by Soleymanjahi et al. (2024) demonstrated that use of CADe was associated with improved average adenoma per colonoscopy (APC) as well as a small increase in advanced colorectal neoplasia (ACN) detection, however no actual difference in ACN per colonoscopy was identified. While there was a minimal increase in procedure time there was no overall difference in endoscopy performance. Similarly, another systematic review by Makar et al (2025) had similar findings. The group found that AI- assisted colonoscopy significantly improved adenoma detection rates (ADR) and APC but no significant differences in detection of advanced adenomas or sessile serrated lesions were seen regardless of endoscopist experience, system type or healthcare setting. Some reviews have also shown that AI-BPS (bowel preparation systems) may also standardize and outperform human bowel prep evaluation thereby reducing bias and limitations inherent in human ratings, but further research is needed to assess this metric.  Nevertheless, while these studies are hopeful for the future, others have raised concerns as it relates to cost of implementation, data privacy, and the need for extensive clinical validation.  There is also a valid concern that the prolonged exposure to AI may lead to an overall negative effect and potential deskilling of endoscopists.

Overall, the current data is in favor of adopting AI into clinical practice given improved quality metrics particularly adenoma detection. Despite this, AI implementation must be balanced against the potential risks and thus further research in computer aided quality improvement systems (CAQ)  along with thoughtful integration will be essential to maximizing the benefits of AI and minimizing harm.