Potential applications of artificial intelligence-assisted endoscopic ultrasound in pancreatic disease

Diagnosing pancreatic diseases can be challenging, even when using advanced tools such as endoscopic ultrasound. Doctors often face uncertainty when interpreting images, which can lead to repeated procedures, prolonged surveillance, or delays in definitive treatment. Four purposefully selected representative cases were used to illustrate potential clinical applications of artificial intelligence (AI) across different pancreatic disease scenarios. We then explored how AI outputs might have informed clinical management had they been available during the original evaluations. The models were originally developed to characterize pancreatic cysts, differentiate solid pancreatic lesions, assess lymph node malignancy, and estimate the grade of IPMNs. Overall, AI predictions were consistent with the final diagnoses in all four cases. Although these observations are only illustrative, they suggest that AI may provide useful additional information during the assessment of pancreatic disease. Given the retrospective nature of this report, prospective studies are required to determine whether these findings translate into meaningful clinical benefit.