Reply to Tyagi et al.

We appreciate the thoughtful comments by Tyagi et al. regarding our recent study on artificial intelligence (AI) for submucosal vessel detection during third-space endoscopy [1] [2].

We acknowledge that our testing environment was artificial and focused specifically on the dissection phase within the submucosal space. However, we believe that in clinical practice, where cognitive load during endoscopic submucosal dissection is higher, the benefit of AI assistance would likely be even more pronounced than in our controlled study setting.

Regarding the concern about testing sequence, this was carefully considered during study design. While learning effects were anticipated for both the testing environment and the AI overlay, we determined that consecutive presentation of AI-assisted videos would allow participants to properly get accustomed to the visualization system. As detailed in the supplementary materials, only two videos were used to introduce the AI algorithm. Additionally, we observed operator fatigue toward the end of testing, which likely counterbalanced any learning curve effects. Conducting tests in a single session was necessary to maintain consistency, though complete standardization was challenging.

The reported intersection over union does not enable perfect vessel delineation in every scenario. However, informal testing feedback indicated that AI alerting interventionists to potentially dangerous situations prompted closer inspection, which is typically sufficient for vessel recognition. This differs from early cancer delineation tasks, where precise boundary marking is essential even under optimal visualization conditions.

We concur with the final point raised. Our preclinical video study utilized specific quality parameters designed to measure the effect of AI on endoscopist performance. This represents only an initial step toward comprehensive clinical evaluation, which would necessarily focus on clinically relevant end points such as complication rates and procedure duration.

We intend to incorporate these insights into future evaluations of this technology.

Publication History

Article published online:
29 July 2025

• References

• 1 Scheppach MW, Mendel R, Muzalyova A. et al. Use of artificial intelligence for submucosal vessel detection during third-space endoscopy. Endoscopy 2025; 57
• 2 Tyagi A, Rughwani H, Kalpala R. et al. Reflections on artificial intelligence for submucosal vessel detection during third-space endoscopy. Endoscopy 2025; 57