AI-Powered Precision Decision-Making for Pancreatic Diseases

MEHTOD： This multicenter clinical trial evaluates the reliability and effectiveness of an AI system for patients with pancreatic diseases in a real-world clinical environment. The study calculates the AI system's classification accuracy using pathological diagnosis (biopsy/surgery results) or long-term follow-up as the "gold standard" for comparison. Additionally, the safety and clinical utility of the management strategies recommended by the AI are assessed by measuring the risk of missing malignant lesions, the rate of unnecessary surgeries for pancreatic diseases, and the level of agreement with traditional clinical decisions. STUDY DESIGN All contrast-enhanced CT images from patients with pancreatic diseases are analyzed by the AI system to generate a classification result (Intervention, Intensive Surveillance, or Routine Surveillance). Simultaneously, clinical doctors review the same data and categorize patients into these three groups to determine their actual care plan: 1. INTERVENTION: Patients assessed by doctors as needing "Intervention" are recommended for further surgical evaluation or treatment. 2. INTENSIVE SURVEILLANCE: Patients assessed by doctors as needing "Intensive Surveillance" receive a personalized, high-frequency follow-up plan until the study endpoint. 3. ROUTINE SURVEILLANCE: Patients assessed by doctors as needing "Routine Surveillance" undergo follow-up for at least one year. If abnormalities arise during this period, the patient is transferred to the appropriate "Intervention" or "Intensive Surveillance" protocol. OUTCOMES： The study compares the performance of the AI system against clinical doctors regarding classification accuracy, the risk of missed diagnoses, unnecessary surgery rates, and decision consistency. These metrics are used to validate the AI system's value, safety, and utility in the clinical management of pancreatic diseases.