Artificial Intelligence-Assisted Endoscopic Ultrasound-Guided Ablation of Pancreatic Neuroendocrine Tumors: Toward Precision Diagnosis, Risk Stratification, and Personalized Therapy

Pancreatic neuroendocrine tumors (pNETs) are increasingly detected at an early stage because of the wider use of cross-sectional imaging and endoscopic ultrasound. Their management remains challenging, particularly for small functioning tumors and selected non-functioning lesions, where the risks of pancreatic surgery must be balanced against tumor biology, symptoms, progression risk, and patient preference. Endoscopic ultrasound (EUS)-guided ablation, particularly radiofrequency ablation, has emerged as a minimally invasive, organ-preserving option for carefully selected patients with small pNETs, especially insulinomas and low-risk non-functioning lesions. However, current evidence is limited by small cohorts, heterogeneous techniques, variable follow-up protocols, and uncertainty regarding long-term oncological outcomes. Artificial intelligence (AI) may enhance this evolving field by supporting EUS-based lesion detection, characterization, grading prediction, risk stratification, patient selection, procedural planning, and post-ablation surveillance. AI-assisted models using EUS images, radiomics, pathology, and multimodal clinical data may help identify patients most likely to benefit from ablation while avoiding inappropriate local therapy in biologically aggressive disease. This review summarizes the current role of EUS-guided ablation for pNETs and explores the emerging potential of AI to support precision diagnosis, individualized risk assessment, and personalized minimally invasive therapy.