Alibaba DAMO Academy has made another breakthrough in the medical field, recently collaborating with institutions such as the Guangdong Provincial People's Hospital to successfully develop a new AI model for colorectal cancer screening — DAMO COCA. This model performed well on 27,000 plain CT images, accurately identifying 5 missed cases of colorectal cancer, with sensitivity and specificity reaching 86.6% and 99.8%, respectively. This means that the rate of misdiagnosis by doctors in colorectal cancer screening will be significantly reduced.

image.png

Colorectal cancer, which refers to colon and rectal cancers, is the second leading cause of cancer-related deaths worldwide. In particular, its incidence is increasing among people under the age of 30. Experts point out that early detection of colorectal cancer is crucial for improving patients' survival rates. Patients who are diagnosed early have a five-year survival rate of over 90%, while those diagnosed at later stages have a survival rate of only 14%. However, traditional methods of colorectal cancer screening, such as fecal occult blood tests and colonoscopies, often cause discomfort, leading many people to avoid screening.

DAMO Academy's innovative solution achieves non-invasive screening through "plain CT + AI". Plain CT is widely used in health check-ups and trauma assessments in China, generating an enormous number of images each year. With the DAMO COCA model, doctors can use data from these routine examinations for colorectal cancer screening without requiring additional preparations from patients. This method will significantly increase the acceptance of screening and help more people detect colorectal cancer risks in time.

To overcome the interference caused by intestinal contents in the images, the technical team at DAMO Academy adopted a deep learning architecture based on "localization first, then diagnosis" and specifically trained it for early tumors smaller than 3 centimeters. This two-stage learning strategy enables the AI model to effectively segment complex intestinal areas, thus detecting suspicious lesions more accurately.

Since its establishment in 2017, DAMO Academy has continuously explored the field of medical AI. Previously, it had developed an AI model for pancreatic cancer screening, DAMO PANDA, and an AI model for gastric cancer screening, DAMO GRAPE. With the launch of DAMO COCA, DAMO Academy's "plain CT + AI" multi-cancer screening technology route has officially gone into operation and has achieved significant progress in multiple cancer screenings.

Key Points:   

🩺 Alibaba DAMO Academy launched the colorectal cancer screening AI model DAMO COCA, improving screening accuracy.  

📈 The survival rate for early detection of colorectal cancer is as high as 90%, but traditional screening methods have low acceptance.  

🔍 The DAMO COCA model uses deep learning technology to identify early-stage colorectal cancer in routine CT images.