DAMO Academy and Shengjing Hospital Launch Liver Cancer AI Diagnostic Model DAMO LiON, Published in Nature Medicine
en.Wedoany.com Reported - On August 24, Alibaba DAMO Academy, together with Shengjing Hospital of China Medical University and other institutions, announced the development of DAMO LiON (Liver DiagnOsis Network), an AI model for liver cancer diagnosis. The model can accurately identify tiny liver cancer lesions by analyzing contrast-enhanced CT images. The related research paper was published on August 19 in the internationally renowned academic journal Nature Medicine.
According to the paper published in Nature Medicine, DAMO LiON is an AI system based on contrast-enhanced CT that supports flexible multi-phase processing, clinical data integration, and workflow-compatible diagnosis of liver malignancies. The model was trained on data from 6,443 patients and retrospectively validated in multicenter and real-world cohorts encompassing 22,251 patients. Validation results showed that DAMO LiON achieved an AUC (Area Under the Curve) of 0.975 (95% CI: 0.971-0.979) for diagnosing malignant tumors, maintaining robust performance in patients with fatty liver (AUC 0.971) and cirrhosis (AUC 0.924).
The research team subsequently deployed DAMO LiON as an "AI safety officer" into routine clinical workflows in a real-world prospective single-arm trial involving 10,333 patients. The trial met its primary endpoint, with the AI achieving a diagnostic AUC of 0.952 (95% CI: 0.942-0.961). During the two-month trial, the AI analyzed contrast-enhanced CT images from over 10,000 patients, assisting physicians in identifying 51 suspicious lesions that had previously been missed, of which 15 were confirmed as malignant tumors. The vast majority of these missed lesions were approximately 1 cm in size, characterized by being "small, faint, and atypically located"—with an average diameter of about 1 cm, low contrast against liver tissue, or situated in relatively rare anatomical positions. These findings directly led to the revision of 37 radiology reports and the escalation of 22 cases to multidisciplinary team (MDT) discussions, helping patients receive timely surgical or pharmacological treatment.
Significant Technical Advantages, Effectively Improving Diagnostic Efficiency
DAMO LiON not only accurately identifies primary liver cancer but is also particularly adept at detecting easily overlooked liver metastases. Studies have shown that the AI model achieves higher accuracy in identifying malignant tumors than radiologists. With AI assistance, physicians' image reading time decreased by an average of 27%, and detection sensitivity for malignant tumors increased by 11.5%, effectively reducing missed diagnoses; junior physicians with AI assistance can reach the diagnostic level of senior physicians.
At the technical level, DAMO LiON employs an improved network architecture that effectively captures the relationship between lesions and the entire liver while preserving local texture and boundaries, enhancing performance on difficult cases such as fatty liver and cirrhosis. Additionally, the AI iteratively fuses images from different phases, capturing pixel-level differences to accurately identify tiny lesions that appear "in a flash" on contrast-enhanced CT scans.
DAMO Academy's Medical AI Endeavors Achieve Another Milestone
Since its establishment in 2017, DAMO Academy has been dedicated to medical AI, pioneering the use of AI to identify subtle lesions in medical images that are imperceptible to the human eye. Previously, DAMO Academy developed DAMO PANDA, an AI model for pancreatic cancer screening based on non-contrast CT, DAMO GRAPE for gastric cancer screening, and DAMO COCA for colorectal cancer screening. These achievements have been published in Nature Medicine three times, entered the innovation channel of the National Medical Products Administration, and received FDA "Breakthrough Device" designation in the United States. DAMO LiON represents another significant breakthrough for DAMO Academy in the field of medical AI.