Alibaba DAMO Academy announced that it has jointly developed an AI model for fatty liver screening called MAOSS with multiple institutions, including the Shengjing Hospital of China Medical University and the Gulou Hospital of Nanjing University. The research findings were published in the prestigious international journal Nature Communications in February of this year.

The prevalence of fatty liver disease in the population exceeds 30%. Due to its subtle early symptoms, it is often overlooked and can progress to liver fibrosis or cirrhosis. Traditional examinations, such as B-ultrasound, have limited sensitivity, while specialized examinations are costly, leading to a high rate of missed diagnoses among high-risk patients in clinical practice.

image.png

Core breakthroughs and advantages of the MAOSS model:

  • Deep mining of unenhanced CT scans: DAMO Academy used the "unenhanced CT + AI" technology, enabling AI to automatically extract high-dimensional features such as liver texture and density. This is the first time that liver steatosis and fibrosis staging can be simultaneously assessed using only an unenhanced CT scan.

  • Diagnostic accuracy surpasses doctors: In multi-center validation, the area under the curve (AUC) of MAOSS for liver steatosis staging reached 0.904-0.917, significantly higher than the average level of radiologists (0.709).

  • Double the detection rate for high-risk cases: For the critical window period to prevent cirrhosis (stage 2 fibrosis), the model can identify 52.4% of high-risk patients, compared to only 16.6% identified by traditional clinical pathways, more than doubling the detection rate.

  • Early warning for cirrhosis risk: Follow-up data show that patients classified as high-risk by MAOSS had a 45.5% chance of developing cirrhosis within two years, far exceeding that of the low-risk group.

Experts from DAMO Academy stated that the model can utilize existing unenhanced CT data from physical exams and outpatient visits, achieving the "front-end prevention" of chronic liver disease management without increasing additional costs for patients. In the future, grassroots hospitals may use this AI technology to provide high-risk warnings during routine health check-ups, enabling early detection and reversal.