Research of the Application of Artificial Intelligence Model "PANDA"

This study focuses on potential cases of clinically missed PDAC. It aims to evaluate the pancreatic cancer screening model PANDA, based on deep learning and non-enhanced CT scans, in a prospective real-world cohort from multiple clinical scenarios at the First Affiliated Hospital of Zhejiang University. The study will track patients with negative imaging reports but positive PANDA model findings, verifying their pathology through gold standard examinations to assess PANDA's efficacy. It aims to validate the model's utility, applicability, sensitivity, and specificity.

Based on these objectives, the study will undertake the following:

• Utilize PANDA's output to categorize enrolled patients into nonPDAC, PDAC, and normal groups. It will compare these results with imaging findings. Patients identified as PANDA-positive for PDAC but without corresponding imaging evidence of pancreatic lesions, or those with imaging suggesting pancreatic findings but lacking subsequent clinical intervention, will be categorized for follow-up. These patients will be recalled to the hospital for further examination and diagnosis at Zhejiang University's First Affiliated Hospital. For PDAC-positive cases identified during secondary examinations, standard clinical procedures such as MDT will be followed for confirmation of pathology. Patients identified as PDAC-negative during secondary examinations will undergo extensive follow-up for up to two years to determine outcomes, thus validating PANDA's sensitivity and specificity. Patients identified by PANDA as nonPDAC-positive but lacking corresponding pancreatic findings in imaging will undergo a review by hepatobiliary pancreatic surgeons to confirm accuracy. Those reported as normal by PANDA but with imaging suggesting pancreatic abnormalities will undergo secondary review by surgical experts to confirm or rule out false negatives by PANDA.
• For true positive PDAC cases identified by PANDA, medical records will be collected (tumor marker levels, patient symptoms, resectability grading, TNM staging, etc.) for comparison with corresponding indicators from PDAC patients identified through the Standard Order of Clinic SOC. This aims to validate PANDA's capability in early detection and identification of lesions in pancreatic cancer development.

Inclusion criteria

• The participants have undergone chest and/or abdominal plain CT scans at outpatient, inpatient, or physical examination centers

Exclusion criteria

• Chest CT scan without pancreatic coverage
• Patients undergoing thoracic/abdominal surgical procedures that affect or alter the anatomical display of the pancreas (esophageal/gastric/pancreatic/vascular/ERCP postoperative, etc.)
• Scanning non-standard examinations, such as significant respiratory motion artifacts