Endoscopic ultrasonography (EUS) is the most sensitive modality for accurately establishing a tissue diagnosis in patients with solid pancreatic masses. However, small lesions can be challenging to detect, particularly for less experienced endosonographers. Therefore, outcomes of EUS are operator dependent. We validated the performance of novel artificial intelligence (AI)-enhanced EUS for detection of solid pancreatic lesions.

In this single-center, prospective, nonrandomized, comparative study, high-risk patients aged ≥18 years referred for pancreatic cancer screening or with suspected (solid and cystic) pancreatic lesions owing to symptoms, radiological, or laboratory findings were evaluated in real time using AI-EUS software. The model included 32 713 EUS frames (training/testing phases) of normal, solid, and >10-mm cystic pancreatic lesions from 202 patients. Clinical validation was conducted prospectively when EUS findings were evaluated concurrently in real time by two independent expert examiners, one using conventional EUS and another with AI-EUS, both blinded to the alternative assessments. The primary outcome was detection of solid pancreatic masses.

308 patients were evaluated (January–July 2024). AI-EUS performance was not significantly different to that of conventional EUS performed by experts (97.1% vs. 100%; risk difference 2.9%, 95%CI –1.2 to 6.8; P = 0.25). Final pathology of 105 pancreatic solid masses revealed neoplasia in 93 (88.6%) and benign lesions in 12 (11.4%).

The performance of AI-EUS was not significantly different to that of experienced endosonographers for detection and segmentation of solid pancreatic masses. By standardizing performance, AI-EUS may have the potential to optimize clinical outcomes in pancreatic cancer.

Received: 06 March 2025

Accepted after revision: 12 September 2025

Accepted Manuscript online:
15 September 2025

Article published online:
13 October 2025

© 2025. The Author(s). This is an open access article published by Thieme under the terms of the Creative Commons Attribution License, permitting unrestricted use, distribution, and reproduction so long as the original work is properly cited. (https://creativecommons.org/licenses/by/4.0/).

Georg Thieme Verlag KG
Oswald-Hesse-Straße 50, 70469 Stuttgart, Germany