Interview with Professor Seong-Eun Kim, Director of Ewha IBD Center

Colonoscopy can help prevent colorectal cancer by detecting and removing neoplastic polyps, often described as the “seeds” of the disease. In response to the rising incidence of colorectal cancer, the Korean government is considering adding colonoscopy at 10-year intervals to the National Cancer Screening Program beginning in 2028 and lowering the eligibility age to 45.

As the number of examinations grows, the medical community is increasingly focused on how to manage their quality. Simply undergoing colonoscopy is not enough to achieve its full preventive benefit. Could artificial intelligence, which automatically detects and analyses lesions in real time, help? To find out, we spoke with Professor Seong-Eun Kim of the Division of Gastroenterology at Ewha Womans University Mokdong Hospital, who recently studied the clinical utility of AI-assisted endoscopy.

Q. Could you introduce the findings from your recent study on AI-assisted endoscopy?

A. We randomly assigned 998 patients to two groups and compared the results of AI-assisted colonoscopy using computer-aided detection, or CADe, in 497 patients with those of conventional colonoscopy in 501 patients. The study was designed as a randomized clinical trial to ensure objectivity.

The adenoma detection rate, or ADR, was 52.3% in the AI-assisted group and 36.1% in the conventional colonoscopy group. ADR also remained higher with AI after adjusting for relevant variables such as sex and age.

Q. Can AI reduce the risk of colorectal cancer incidence and mortality?

A. Adenomas are precancerous lesions, some of which can progress to colorectal cancer. Studies have consistently reported that patients who undergo colonoscopy performed by physicians with higher ADRs have lower subsequent risks of developing and dying from colorectal cancer.

It is meaningful that AI can help physicians detect additional adenomas they might otherwise miss. However, further research is needed to determine how much the additional detection of adenomas with AI actually reduces colorectal cancer incidence and mortality. Sufficient long-term follow-up will be required to assess its preventive effects.

Q. What are the advantages of using AI in clinical practice?

A. Even highly experienced and skilled physicians cannot detect every lesion. Colonoscopy requires sustained concentration, and ADR may vary depending on when an examination is performed—that is, where it falls in the sequence of procedures.

Small polyps and lesions that do not differ greatly in appearance from the surrounding mucosa can be particularly difficult to identify. AI acts as a “second pair of eyes,” helping physicians find lesions they might otherwise miss.

Q. What roles can AI play beyond lesion detection?

A. AI can help differentiate lesion types that may have the potential to progress to cancer and check whether an examination has been performed thoroughly without missing any areas. Functions that automatically record endoscope insertion and withdrawal times and measure lesion size are already being used to provide additional information to clinicians.

AI technologies are also being developed to assess whether the bowel was adequately cleansed and whether all major anatomical landmarks were examined, thereby helping evaluate and manage examination quality.

I expect quality management to become an increasingly important function of AI-assisted endoscopy. If colonoscopy is added to the National Cancer Screening Program, demand will rise, making it even more important to maintain consistent examination quality.

Q. Are there any concerns about the growing use of AI?

A. Excessive reliance on AI could lead to “de-skilling,” in which clinicians’ fundamental procedural and decision-making abilities decline. AI can be particularly useful for less experienced clinicians, but ironically, they may also be the most vulnerable to de-skilling.

Not every area flagged by AI requires a biopsy, and performing unnecessary biopsies could adversely affect patients. The physician must make the final decision. AI should be used to complement clinicians’ expertise, not replace it. I always emphasise this point when educating and training junior physicians.

Q. What preparations are needed to ensure that clinicians and AI work effectively together?

A. When used appropriately, AI can help create a more systematic and reliable environment for endoscopic examinations. Colonoscopy is not an easy procedure for patients to undergo. Even if it is performed only once, it must be conducted thoroughly and evaluated properly. That is why quality management is so important.

We must continue building clinical evidence. If AI proves genuinely helpful in practice, I believe a reimbursement framework should also be established to support its use.