Revolutionary AI Model Enhances Colorectal Cancer Detection

Alibaba has launched a groundbreaking AI model, DAMO COCA, designed to enhance colorectal cancer detection using non-contrast CT scans. This innovative approach promises to make cancer screening more accessible and patient-friendly, potentially increasing global screening rates. With impressive accuracy rates and the ability to detect lesions often missed by traditional methods, DAMO COCA represents a significant advancement in early cancer diagnosis. This model is part of a broader initiative aimed at leveraging AI for multiple cancer screenings, addressing a critical need in global healthcare, especially in underserved regions. Learn more about this transformative technology and its implications for cancer detection.

By Narendra Chaudhary Apr 29, 2026, 09:38 IST

Transforming Cancer Detection with AI

In a significant advancement for artificial intelligence in the medical field, Alibaba has unveiled an innovative AI model that promises to revolutionize the detection of colorectal cancer. This model, developed by the Alibaba DAMO Academy in partnership with Guangdong Provincial People’s Hospital and other organizations, is named DAMO COCA. It utilizes non-contrast CT scans to accurately identify cancer without the need for bowel preparation.

A New Era of Non-Invasive Screening

Conventional methods for colorectal cancer screening, like colonoscopy, are effective but often avoided due to their invasive nature and the necessity for bowel preparation. DAMO COCA aims to shift this paradigm by facilitating non-invasive cancer screening through standard CT scans. This breakthrough could enhance accessibility and patient comfort, potentially leading to increased screening rates globally.

Research-Backed Accuracy

Research published in the Annals of Oncology indicates that the AI model boasts a sensitivity of 86.6% and an impressive specificity of 99.8%. Experts highlight that this means the system is highly proficient in accurately identifying both cancerous and non-cancerous cases, thereby minimizing false positives. Notably, DAMO COCA has shown enhanced detection capabilities in regions where traditional screening often overlooks lesions, underscoring its potential for earlier and more reliable cancer detection.

Last April, Damo’s Panda model received a “breakthrough device” designation from the US Food and Drug Administration, allowing for an expedited review and approval process.

Leveraging Deep Learning and Big Data

The study analyzed over 27,000 CT scans, making it one of the largest datasets utilized for AI research in colorectal cancer. The model employs a two-stage deep learning framework that enhances the interpretation of the intestines' complex structure. This sophisticated architecture enables the AI to identify subtle irregularities that might be missed during manual evaluations, providing a valuable resource for radiologists and healthcare professionals.

Part of a Larger Initiative

DAMO COCA is not an isolated innovation; it is part of Alibaba’s extensive “CT + AI” initiative aimed at early detection of multiple cancers. Previous models from DAMO Academy have already targeted pancreatic and gastric cancers, indicating a comprehensive effort towards AI-enhanced diagnostics.

Significance for Global Healthcare

Colorectal cancer ranks among the top causes of cancer-related fatalities worldwide, primarily due to late-stage detection. Many individuals exhibit no symptoms in the early phases, and when symptoms do arise, they may include persistent diarrhea or constipation, blood in the stool, narrow stools, abdominal discomfort, and unexplained weight loss or anemia. Key risk factors encompass age (over 50), high intake of red and processed meats, low fiber diets, obesity, smoking, excessive alcohol consumption, and personal or family histories of polyps or cancer.

By eliminating obstacles such as cost, discomfort, and preparation requirements, AI-driven CT screening could significantly enhance early diagnosis rates. For nations with limited access to specialized screening facilities, this technology presents a scalable and efficient alternative, particularly benefiting underserved areas where early detection resources are scarce.