UEG's educational platform for gastroenterology education bringing recordings, courses, guidelines, and more together in one place.

UEG Week Berlin 2025 Recordings Gut Guide online

Rate for better recommendations!

Your feedback helps us show you more of what you like and less of what you dislike!

More of this!

We will show you more similar content.

Not for me!

We won't suggest this again.

Bookmark!

Save this content for later!

New to myUEG? Create an account! Log In with myUEG account

UEG Poster

Standard Poster

Share via Email Share on Facebook Share on X Share on LinkedIn Share on Bluesky

This content is part of Gutflix. Log in with your myUEG account, or create one free, to watch it.

Not sure what you can access? Learn more about account types.

PREDICTION OF PANCREATIC NEUROENDOCRINE TUMOR GRADING USING AN ARTIFICIAL INTELLIGENCE-BASED VIDEO ANALYSIS MODEL APPLIED TO CONTRAST-ENHANCED ENDOSCOPIC ULTRASOUND VIDEOS

Matteo Tacelli 1, Adrien Meyer 2, Paolo Biamonte 1, Armine Vardazaryan 2, Rubino Nunziata 1, Leonardo Sosa Valencia 2, Gabriele Capurso 1, Nicolas Padoy 2, Paolo Giorgio Arcidiacono 1

Affiliations

1 San Raffaele Scientific Institute IRCCS, Milano, Italy

2 IHU-Strasbourg, Institute of Image-Guided Surgery, Strasbourg, France

Summary

AI Generated

Summary is not available for this content yet.

Download PDF

Was this helpful?

Thanks for your feedback.

This summary was generated by an AI large language model based on the content transcript. It is for informational purposes only and should not be considered a substitute for clinical judgment. Always rely on your professional expertise and the full clinical context when making clinical decisions.

Abstract

Introduction

Pancreatic neuroendocrine neoplasms (PNENs) are rare tumors from pancreatic endocrine cells, accounting for 1–2% of pancreatic malignancies. Their incidence is rising due to improved diagnostics and awareness. WHO grading—based on mitotic index and Ki-67—is the main prognostic factor. Grade 1 non-functional PanNETs <2 cm are often surveilled, while G2/G3 tumors need active treatment. Accurate preoperative grading is therefore essential. EUS-guided fine-needle aspiration (EUS-FNA) is the main diagnostic tool, though differentiating G1 from G2/3 tumors is difficult, with frequent grading misclassification. Contrast-enhanced EUS (CE-EUS) improves accuracy: low-grade PNENs show early hyperenhancement with washout, while high-grade tumors, with lower microvascular density, appear hypoenhancing. Still, CE-EUS interpretation is subjective. Artificial intelligence (AI) can enhance diagnostic precision.

Aims & Methods

This retrospective multicenter study aimed to evaluate AI in analyzing CE-EUS video sequences to predict PNEN grade. Conducted at San Raffaele Hospital, it included adults with PNEN confirmed by surgery or EUS-FNA, Ki-67 immunostaining, and CE-EUS videos (≥1 min, arterial + venous phases). Exclusions: MiNENs, missing grading, poor-quality videos, lack of consent, or non-pancreatic NETs. EUS was performed by experts using Pentax echoendoscopes and Sonovue®. Tumors were graded per WHO 2019. Videos were anonymized, ROIs manually selected by two experts, and transferred to IHU Strasbourg for AI analysis. CE-EUS videos were processed using a video transformer AI model. Patients were split into training (70%), validation (10%), and test (20%) cohorts. Multiple models were tested; the best-performing one was selected. Accuracy was assessed via precision (PPV), recall (sensitivity), and F1-score.

Results

115 patients (77 males, 66.9%) were included: 70 G1 and 45 G2 tumors. Median lesion size (G1: 17 mm vs. G2: 16 mm), tumor location (head: 41.2% vs. 32%), and stiffness (rigid: 52.9% vs. 55.5%) did not differ significantly. At diagnosis, 15.7% of G1 and 26.7% of G2 had metastases. AI analysis identified G1 tumors with 67% precision, 86% recall, F1-score 75%. The GradAINet model showed strong performance: sensitivity 0.817 (95% CI: 0.556–1.000), specificity 0.806 (CI: 0.588–1.000), PPV 0.759, NPV 0.856, and accuracy 0.811. By grade, G1 sensitivity was 0.796, specificity 0.817, PPV 0.859, NPV 0.741, accuracy 0.805. For G2–G3, sensitivity was 0.817, specificity 0.796, PPV 0.741, NPV 0.859, accuracy 0.805.

Conclusion

EUS-based grading shows 80.3% concordance with surgical findings, often underestimating tumor grade. This is the first study using deep learning on full CE-EUS video sequences, analyzing arterial and venous phases separately. GradAINet achieved high accuracy (0.811), outperforming static image methods. Further multicenter validation and integration of clinical/molecular data are needed to confirm generalizability.

Conference

UEG Week Berlin 2025

Submission format

Abstract

Session

EUS: Hot topics (Posters)

Citation

United European Gastroenterology Journal 2025; 13 (Supplement 8)

Published

2025

More Like This:

The era of TNT: What should we all be aware of?

Ana-Maria Bucalau

Where is my chest pain coming from?!

Isabella Dulmin

Surgical navigation

Suzanne S. Gisbertz

GASTRIC ISCHEMIA: THE FORGOTTEN CULPRIT IN ABDOMINAL PAIN

Lama Chouman

Summary: Hereditary GI cancer syndromes - Surveillance strategies

Maria Pellisé Urquiza

THE DARK SIDE OF THE ESOPHAGUS: A CASE REPORT OF ACUTE ESOPHAGEAL NECROSIS

Ashraf Morcos

UEG Presentation

Share via Email Share on Facebook Share on X Share on LinkedIn Share on Bluesky

The era of TNT: What should we all be aware of?

This content is part of Gutflix. Log in with your myUEG account, or create one free, to watch it.

Not sure what you can access? Learn more about account types.