Dolwani, Sunil  ORCID: https://orcid.org/0000-0002-3113-5472
2026.
AI-assisted colonoscopy and robotic colonoscopy: Hype versus reality [Abstract].
Journal of Digestive Endoscopy
17
(S 01)
, p. S68.
10.1055/s-0046-1828871
|
Abstract
Artificial intelligence (AI)-assisted colonoscopy has been considered an exciting advance in endoscopic research and innovation. Initial promising pilot studies on relatively small cohorts and case–control studies have since evolved into numerous randomized control trials (RCTs) most with parallel group and blinded as well as unblinded designs. Consistent and rigorous evaluation and detailed analysis of these suggests a significant gap between initial hope and expectation—the “hype” versus the actual reality of implementation in practice. It has also given rise to some previously unanticipated issues such as the impact on clinical procedural skills. CADe Computer (AI)-assisted polyp detection has been the subject of numerous mostly well-designed studies including RCTs. Consistent findings from these most recently rigorously evaluated by the National Institute for Health and Care Excellence (NICE) in the United Kingdom suggest that any incremental benefit of additional polyp detection with AI was mainly in smaller polyps less than 10 mm in size and the evidence was less convincing for larger polyps or those considered clinically significant. There did seem to be a gradient of benefit between nonexpert and expert colonoscopists with higher likely benefit in nonexperts. CADx Computer (AI)-assisted polyp diagnosis or characterization has seen significant advances in correlation and accurate categorization of polyps as “adenomatous” or “nonadenomatous” though any further benefit has been limited by the reduced certainty in characterization of sessile serrated lesions as opposed to hyperplastic polyps and no real advances in characterization of dysplasia. As these are considered important predictors of surveillance guidelines and “resect and discard” strategies, polyp characterization has thus far not advanced beyond a “neoplastic” versus “nonneoplastic” categorization. CADq Computer (AI)-assisted quality assurance is a more recent development with fewer RCTs looking at improvement in quality metrics with AI. Automated recording of procedure time, withdrawal time, completion rates, and proposing new metrics such as mucosal exposure time and inspection time seem to be very practical areas where AI seems to make a difference and aid the practicing colonoscopist in procedure-associated administrative tasks as well as quality assurance. Unintended Consequences The impact of AI assistance on skills acquisition and skills maintenance was not initially considered by most studies. However, in recent studies there is a suggestion of potential deskilling of colonoscopists with designs varying from pre- and post-AI detection rates to retrospective analyses. There is a suggestion of the impact being variable depending on the baseline level off skill and this is a subject of further studies. Robotic Colonoscopy The field of robotics has made significant advances in various areas of health care and in particular surgical approaches to precise lesion and organ resectional procedures. Due to the complexity of the colon being a flexible, stretchable, nonlinear complex tubular organ, robotic colonoscopy has been relatively slow to develop to a stage when it may be considered a close to clinical application device. Various types of robotic approaches have been proposed, designed, and tested to prototype and early clinical stages with a few of these having established small clinical studies in humans and very few used in practice limited to certain countries and settings. Some designs have tested fully autonomous versus semiautonomous approaches. The mechanisms of locomotion in different soft robots have varied from an eversion type to earthworm/inchworm design with variations in sensing mechanisms for pressure, obstructions, friction, and pneumatic and suction mechanisms. Propulsion of actuated rather than passive systems has also been variable with pneumatic and suction systems, electric motors, and magnetic fields all under active study. Sustainability has been an important consideration in these designs as almost is disposable rather than reusable in nature for the inserted components. Therapeutic capability has also been relatively limited with recent improvements in tip control mechanisms. Conclusion Both AI-assisted colonoscopy and robotic colonoscopy constitute exciting and innovative developments in colonoscopy and wider endoscopy. Their developments have reflected wider advances in these fields in areas of industrial and engineering applications. Further improvements are likely to reflect the same. However, clinical applications thus far of both these techniques have been relatively limited with the likelihood being that AI-assisted quality assurance may prove to be a pragmatic way of integration into general and large-scale endoscopy workflows to improve the quality and documentation of procedures.
| Item Type: | Short Communication |
|---|---|
| Date Type: | Published Online |
| Status: | Published |
| Schools: | Schools > Medicine |
| Publisher: | Medknow Publications |
| ISSN: | 0976-5042 |
| Last Modified: | 01 Sep 2026 11:16 |
| URI: | https://orca.cardiff.ac.uk/id/eprint/189289 |
Actions (repository staff only)
 |
Edit Item |
Altmetric
Altmetric