ARTFEED — Contemporary Art Intelligence

Alibaba's Coca AI Model Outperforms Radiologists in Colorectal Cancer Detection

ai-technology · 2026-04-28

Alibaba Group Holding has released an AI model named Coca, developed by its research arm Damo Academy, which claims higher sensitivity than radiologists in detecting early-stage colorectal cancer from non-contrast CT scans. In a study involving over 27,000 people, Coca identified five previously missed cases, achieving 86.6% sensitivity and 99.8% specificity. Damo Academy stated that Coca outperformed 10 radiologists by 20.4% in sensitivity. The model was developed with Chinese institutions including Guangdong General Hospital. Current diagnostic methods like colonoscopy are invasive, while Coca offers a non-invasive, cost-effective alternative. The findings were published in the Annals of Oncology peer-reviewed journal last week.

Key facts

  • Alibaba's Damo Academy released Coca AI model for colorectal cancer detection.
  • Coca identified five previously missed cases from scans of over 27,000 people.
  • Coca achieved 86.6% sensitivity and 99.8% specificity.
  • Coca outperformed 10 radiologists by 20.4% in sensitivity.
  • Developed with Chinese institutions including Guangdong General Hospital.
  • Coca is non-invasive and cost-effective compared to colonoscopy.
  • Study published in Annals of Oncology last week.
  • Alibaba owns the South China Morning Post.

Entities

Institutions

  • Alibaba Group Holding
  • Damo Academy
  • Guangdong General Hospital
  • Annals of Oncology
  • South China Morning Post

Locations

  • China

Sources