Alibaba’s DAMO Academy Unveils AI That Spots Esophageal Cancer in Ordinary Lung CT Scans
TREE NEWS reports: Alibaba’s DAMO Academy, together with Sichuan Cancer Hospital and Sun Yat-sen University Cancer Center, has released DAMO EAGLE, an AI model that screens for esophageal cancer using plain, non-contrast CT scans — no intubation, no contrast injection. The research was published in Nature Medicine on September 24. The model reads the esophagus in scans originally taken to check the lungs, flagging high-risk patients who then still require endoscopy and pathology for a definitive diagnosis.
It is the fourth cancer-screening model from DAMO Academy, following pancreatic, gastric and colorectal cancer tools. The pancreatic model, PANDA, received FDA Breakthrough Device designation in April 2025.
Why This Matters Clinically
China accounts for nearly half of global esophageal cancer cases: roughly 200,000 new cases and 172,000 deaths in 2024. About 70% of patients are diagnosed at an advanced stage. When the lesion is still in the mucosal layer, endoscopic microsurgery can cure it with a five-year survival rate above 95%. Once it spreads, survival falls to 10–15%.
Endoscopy is the gold standard, but supply is nowhere near demand. More than 100 million people are estimated to be high-risk, while only about 30 million upper gastrointestinal endoscopies are performed annually. Training an endoscopist takes three to five years, and many patients — including physicians’ own relatives — fear the procedure.
DAMO EAGLE sidesteps that bottleneck. In a simulated study in Suining, Sichuan, using CT-first screening followed by endoscopy, the detection rate for malignant lesions rose from 1.7% to 5.2%, and the number of endoscopies needed to find one malignant case dropped from 59 to 19. In validation across three countries, 12 centers and more than 80,000 patients, sensitivity reached nearly 90% for esophageal cancer, with specificity of 99.2% in clinical settings and 99.94% among 10,959 asymptomatic health-check participants. Sensitivity for precancerous lesions — a weaker signal — was 52.5%.
Market Implications
- Healthcare AI and imaging: The model validates a scalable “opportunistic screening” pathway that reuses existing CT infrastructure. This strengthens the commercial case for AI-assisted diagnostics globally, particularly for vendors of CT scanners, PACS software and hospital IT systems in China and beyond.
- Alibaba and Chinese tech: DAMO Academy’s expanding portfolio — five Nature Medicine publications and an FDA breakthrough designation — positions Alibaba as a serious player in medical AI, an area where Beijing has signaled policy support. This could support sentiment around Alibaba’s cloud and AI ambitions.
- Medtech and endoscopy: The tool is designed to triage, not replace endoscopy. Rising screening volumes could increase demand for endoscopy equipment and consumables over time, even as it reduces unnecessary procedures.
- Cost dynamics: Specificity matters economically. Each false positive can mean an extra 1,000+ yuan in follow-up costs. High specificity is what makes reimbursement and health-insurance adoption feasible.
Key Takeaways for Investors
- AI-driven cancer screening is moving from research to real-world validation, with regulatory signals (FDA) and top-tier journal publication adding credibility.
- The economic model — reusing existing CT scans rather than adding new tests — is capital-light and could scale rapidly in China’s tier-2 and tier-3 hospitals.
- Watch for prospective clinical trials and reimbursement decisions; these, not model accuracy alone, will determine commercial adoption.
- Precancerous lesion sensitivity of 52.5% and low endoscopy completion rates among AI-flagged patients remain open questions that could slow deployment.
