Human-machine collaboration in colorectal cancer screening: a narrative review of artificial intelligence-assisted colonoscopy

Colorectal cancer (CRC) is a globally prevalent malignant tumor, and colonoscopy serves as the gold standard for its early screening and diagnosis. However, traditional colonoscopy highly depends on the experience of endoscopists, which is associated with limitations such as a relatively high miss rate and inadequate inspection quality control. This narrative review summarizes the application progress, clinical value, and existing challenges of artificial intelligence (AI) in colonoscopy. Currently, AI has formed three mature application directions, namely computer-aided detection, diagnosis, and quality control training, which extensively cover key links of colonoscopy, including bowel preparation assessment, cecal landmark identification, withdrawal quality monitoring, auxiliary lesion detection, and polyp character evaluation. Preliminary evidence suggests that AI can promote the standardization of colonoscopy procedures and the refinement of quality control indicators, potentially bridge the diagnostic gap between endoscopists with different levels of experience, and reduce the misdiagnosis of minor lesions. However, current studies predominantly report positive effects, and there is a lack of long-term outcome data, standardized methodology, and discussion of human-AI interaction dynamics. In conclusion, AI is expected to become an important auxiliary tool for the early screening and diagnosis of CRC, but further rigorous evaluation of its clinical value and collaborative workflow is needed.