Commentary

The key message about the introduction of artificial intelligence (AI) in endoscopy is not that human detection no longer has a role, but exactly the opposite. If we want the best use of AI in our daily routine, all the colonoscopy quality indicators must also be at their best: excellent quality of bowel preparation, a high intubation rate, a very good withdrawal technique, adoption of dynamic position changes for adequate luminal distension, patience, and washing away the mucus and stool residues that preclude colon surface exploration.

Furthermore, endoscopists must be aware that the computer-aided detection (CADe) software used in colonoscopy is mainly based on subcentrimetic adenomas; thus several important lesion types (e. g., nongranular laterally spreading tumors and sessile serrated lesions) have been insufficiently represented in the dataset and might be easily missed by AI [1].

The real gain from the implementation of AI in our practice should be the promotion of a virtuous circle in which the endoscopy service and the endoscopists do their best to maximize the usefulness of this new, exciting, and still evolving tool; otherwise we will only have another ornament that needs dusting in the endoscopy room.

Publication History

Article published online:
22 September 2022

• References

• 1 Lafeuille P, Lambin T, Yzet C. et al. Flat colorectal sessile serrated polyp: an example of what artificial intelligence does not easily detect. Endoscopy 2022; 54: 520-521 DOI: 10.1055/a-1486-6220.