Journal: Endoscopy
This multicenter prospective study evaluated a novel AI-based algorithm for polyp size estimation during colonoscopy in 217 patients undergoing colorectal cancer screening or surveillance.
Among 465 polyps, the AI demonstrated an overall size class determination accuracy of 85.8%, with:
- Highest accuracy for diminutive polyps (≤5 mm): 93.3%
- Lower accuracy for large polyps (≥10 mm): 55.1%
The AI tool correctly assigned 90.8% of patients to appropriate surveillance intervals and showed a mean absolute error of 1.13 mm in measuring polyps ≤10 mm.
These results suggest that the AI model performs comparably to expert endoscopists and could enhance the accuracy and efficiency of colorectal cancer screening.