PANDAPro AI to detect pancreatic cancer on routine CT scans

Research of the Application of Pancreatic Cancer Screening Artificial Intelligence Model 'PANDAPro': A Single-Center,Realworld Clinical Trial

We will test whether the PANDAPro artificial-intelligence tool can spot pancreatic cancer on routine chest or abdominal CT scans in adults aged 18–90.

Quick facts

• Phase: Not applicable
• Study type: Interventional
• Enrollment: 100000 (estimated)
• Ages: 18 Years to 90 Years
• Sex: All
• Sponsor: Zhejiang University (Academic / other)
• Locations: 1 site (Hangzhou, Zhejiang)
• Trial ID: NCT06643715 on ClinicalTrials.gov

What this trial studies

The project upgrades the previously developed PANDA deep-learning model (PANDAPro) with additional training and multi-center external validation. In a first round, the team will retrospectively compare PANDA and PANDAPro on consecutive enhanced abdominal CTs from 2018–2022 using pathology as the reference standard. In a second round (expected Nov 1, 2025–Aug 1, 2027) PANDAPro outputs will be recorded in real time on non-contrast CTs to find lesions that imaging reports may have missed and recalled patients will receive secondary examinations and multidisciplinary review. The work uses large clinical datasets at Zhejiang University’s First Affiliated Hospital and collaborating centers to test whether the model can prompt and supplement routine PDAC diagnosis in clinical practice.

Who should consider this trial

Good fit: Adults aged 18–90 who have had chest or abdominal CT scans that include the pancreas and whose scans are of diagnostic quality without surgical alterations or severe artifacts are the intended candidates.

Not a fit: People whose CTs do not show the pancreas, emergency non-contrast CTs, scans with major motion/artifact problems, or those with prior operations that alter pancreatic anatomy are unlikely to benefit from this tool.

Why it matters

Potential benefit: If successful, PANDAPro could help detect pancreatic cancer earlier or catch cases that were missed on routine CT scans, enabling earlier follow-up and treatment.

How similar studies have performed: A prior PANDA deep-learning model showed promising retrospective performance, but large-scale, real-world prospective validations of such AI screening tools remain limited.

Eligibility criteria

Inclusion Criteria:

* Subjects who have undergone chest and/or abdominal CT scans at outpatient clinics, inpatient departments, or physical examination centers;
* Age at the time of the scan between 18-90 years old, with no restriction on gender;

Exclusion Criteria:

* Chest CT scans that do not cover the pancreas;
* Non-contrast CT scans performed in emergency settings;
* Patients who have undergone thoracic/abdominal surgeries affecting or altering the anatomical display of the pancreas (e.g., post-esophageal, gastric, pancreatic, vascular surgeries, or post-ERCP);
* Non-standard scans (e.g., hands placed on either side of the body or abdomen, severe respiratory motion artifacts, perfusion contamination, etc.);
* CT scans ordered by hepatobiliary and pancreatic surgeons or oncologists;
* Patients referred to a higher-level hospital due to a pancreatic mass found during local hospital examination;
* Patients who, for personal reasons, did not follow up with pancreatic cancer diagnosis or treatment at the hospital, or were lost to follow-up midway;
* Patients with concurrent malignancies in other locations or those undergoing comprehensive cancer treatment for malignant tumors;
* Imaging reports made by radiologists without referring to AI during the image interpretation;
* Patients who underwent enhanced CT, MRI, or PET-CT examinations concurrently.

Where this trial is running

Hangzhou, Zhejiang

• the First Affiliated Hospital, School of Medicine, Zhejiang University — Hangzhou, Zhejiang, China (Recruiting)

Study contacts

• Study coordinator: Qi Zhang

Conditions Pancreatic Cancer; Artificial Intelligence; PANDA PLUS; PANDA