Alibaba DAMO Academy: PANDA AI pancreatic cancer screening on plain CT

Alibaba’s DAMO Academy uses AI on non-contrast (“plain”) CT scans to screen for pancreatic cancer early. Its Nature Medicine paper reports 92.9% sensitivity and 99.9% specificity for lesion detection, with sensitivity 34.1% higher than the average radiologist. The tool received FDA Breakthrough Device designation in April 2025, and a pilot in Ningbo analyzed more than 180,000 CT scans and found 24 pancreatic cancers, 14 of them early-stage.

Summary

Item	Detail
Organization	Alibaba DAMO Academy
Industry	Healthcare
Use cases	Business intelligence, data-driven analysis
Functions	R&D
Reported result	Screening cost about $25; real-world validation on 20,530 people with 92.9% sensitivity / 99.9% specificity; Ningbo pilot found 24 pancreatic cancers, 14 early-stage
Moments rating	S-tier quick win
AI use-case score	100

Why it matters

Plain CT is cheap and already performed routinely in check-ups. Adding AI turns an existing scan into a screening opportunity instead of requiring a new, expensive test.

Case details

Background

Pancreatic cancer has few early symptoms. China Daily, citing The New York Times, reports a five-year survival rate below 10%, and the study’s authors note that liquid biopsy detects only about 50-60% of pancreatic cancers. DAMO PANDA, developed by Alibaba’s DAMO Academy, combines plain CT with deep learning to find pancreatic lesions.

What they did

The model was trained on data from 3,208 patients, and the study was published in Nature Medicine in 2023 (Cao et al.). PANDA went live in November 2023 in Chinese hospitals and health check-up centers. According to Yicai, its “one scan, multiple screenings” technology is now connected to some check-up chains, imaging IT vendors and several hospitals in Shanghai, Zhejiang and elsewhere. A hospital in Ningbo uses it as a clinical trial; other pilot sites include the Shanghai Institute of Pancreatic Diseases, Jiaxing Second Hospital and dozens of local community hospitals. DAMO is also working with the World Health Organization to bring the tool to low-resource regions.

Results and data

• Real-world validation on 20,530 people: 92.9% sensitivity and 99.9% specificity for lesion detection, 34.1% and 6.3% higher respectively than the average radiologist. Plain CT plus PANDA was non-inferior to contrast-enhanced CT radiology reports.
• FDA Breakthrough Device designation in April 2025.
• The Ningbo pilot analyzed more than 180,000 CT scans and helped find 24 pancreatic cancers, 14 of them early-stage. Reports put identification accuracy at 93% and the cost of plain CT plus AI screening at about $25.
• The sources do not give a cumulative number of people served.

What to take from it

• Reusing plain CT scans already taken in check-ups for “opportunistic” screening adds no extra step for the patient, which is what keeps costs low.
• Paper first, then regulatory designation, then multi-site pilots is a common route for medical AI into clinical use.

Limits

FDA Breakthrough Device designation is a fast-track review status, not market approval. The paper’s data come from specific populations; other hospitals, scanners and patient groups need fresh validation. Deployment also involves medical device registration, integration with hospital systems and a care pathway for positive results.

Source

Original source: China Daily · DAMO PANDA report

Additional source: Targeted Oncology: DAMO PANDA earns FDA Breakthrough Device designation

Additional source: Yicai: DAMO PANDA earns FDA Breakthrough Device designation

This is an industry research entry: a publicly documented third-party AI case, not a Moments client case. For projects Moments delivered itself, see client cases.