Alibaba DAMO Academy Launches Colorectal Cancer Screening AI Model DAMO COCA

DAMO Academy's Third Cancer AI Model Debuts, Precisely Targeting Colorectal Cancer

On April 28, Alibaba DAMO Academy, in collaboration with Guangdong Provincial People's Hospital and other institutions, officially released its colorectal cancer screening AI model — DAMO COCA. The model can accurately identify colorectal cancer from routine non-contrast CT images without patients even being aware, proposing for the first time internationally an opportunistic colorectal cancer screening method that requires no bowel preparation. This is the third cancer screening AI model released by DAMO Academy, following its pancreatic cancer and gastric cancer screening models.

Core Capability: Catching 5 Missed Cancer Cases Among 27,000 People

The core highlight of DAMO COCA lies in its outstanding detection performance. In a large-scale retrospective validation covering 27,000 individuals, the model successfully identified 5 previously missed colorectal cancer cases, achieving an overall sensitivity of 86.6% and a specificity as high as 99.8%. This means the model can not only efficiently detect potential colorectal cancer patients but also rarely produces false positives, reaching an extremely high standard of clinical utility.

Traditional colorectal cancer screening methods primarily rely on colonoscopy, which requires patients to undergo strict bowel preparation in advance — a cumbersome process with a poor patient experience that deters many from screening altogether. DAMO COCA takes a completely different approach: it analyzes routine non-contrast CT images taken for other medical reasons, requiring no additional examination procedures and truly realizing the concept of opportunistic "incidental cancer screening."

Technical Approach Validated: 'Plain CT + AI' Builds a Multi-Cancer Screening System

Notably, the release of DAMO COCA marks the official validation of DAMO Academy's original "plain CT + AI" multi-cancer screening technical approach. Previously, DAMO Academy had successively launched AI screening models for pancreatic cancer and gastric cancer, both based on the same technical pathway — using deep learning models to automatically identify early cancer signals from the most common non-contrast CT images in everyday clinical settings.

The core advantages of this technical approach are threefold:

• High data accessibility: Non-contrast CT is one of the most widely used imaging examinations in clinical practice, generating hundreds of millions of imaging datasets annually, providing a natural foundation for large-scale screening.
• Zero patient burden: No additional examinations or special preparations are needed. Cancer clues are "mined" from existing images, and patients remain completely unaware throughout the process.
• Multi-cancer scalability: The same technical framework can be adapted for different cancer types, laying the groundwork for building a one-stop multi-cancer joint screening system.

The successive deployment of models for three cancer types proves that this technical approach possesses strong replicability and clinical translation potential.

Industry Significance: AI Redefines the Early Cancer Screening Paradigm

Colorectal cancer ranks among the most prevalent malignant tumors globally. In China, more than 500,000 new colorectal cancer cases are diagnosed annually, with a low early detection rate — many patients are already at intermediate or advanced stages at the time of diagnosis. The opportunistic screening model represented by DAMO COCA has the potential to fundamentally change this situation.

From a broader perspective, DAMO Academy's multi-cancer screening system represents an important trend in the AI medical imaging field: evolving from single-disease assisted diagnosis toward multi-cancer, full-process, low-cost universal screening. As the "plain CT + AI" model covers more cancer types, patients in the future may be able to complete preliminary screening for multiple cancers with just one routine CT examination.

Outlook: The Era of Multi-Cancer Joint Screening Is Accelerating

With the release of DAMO COCA, DAMO Academy has completed its deployment across three high-incidence cancer types: pancreatic cancer, gastric cancer, and colorectal cancer. It is foreseeable that the team may next extend the technology to liver cancer, lung cancer, and other cancer types, gradually building an AI screening network covering major malignant tumors.

Of course, transitioning from the laboratory to large-scale clinical application still faces multiple challenges, including model generalizability validation, medical device regulatory approval, and hospital system integration. However, it is undeniable that DAMO Academy's "plain CT + AI" approach has provided a novel and highly promising solution for early cancer screening — using AI to make cancer screening as simple as getting a CT scan.