Alibaba DAMO Academy releases AI model, doubles detection rate for high-risk populations

On the morning of March 9, news reports that Alibaba DAMO Academy, in collaboration with Shengjing Hospital affiliated with China Medical University, Nanjing University Affiliated Gulou Hospital, and other institutions, has developed an AI model for fatty liver screening called MAOSS. Using routine examinations such as non-contrast CT scans and serum markers, it can not only accurately stage liver fat but also assess liver fibrosis progression, increasing the detection rate of high-risk patients from 16.6% to 52.4%. The related paper was published in Nature Communications.

DAMO Academy algorithm expert Gao Yuan explained that traditionally, non-contrast CT scans have limited ability to identify early fatty liver and liver fibrosis. Leveraging years of experience in the “non-contrast CT + AI” field, DAMO Academy used AI to automatically extract high-dimensional features such as liver texture, density, and morphological changes. By training on large-scale, gold-standard in vivo biopsy data and combining multimodal data like serology and imaging reports, they achieved, for the first time, simultaneous assessment of liver fat content and fibrosis staging from non-contrast CT scans.

It is also reported that the MAOSS model can effectively predict the progression of cirrhosis. Retrospective follow-up analysis shows that patients identified as high risk by MAOSS have a 45.5% chance of developing cirrhosis within two years, significantly higher than the 11.8% in the low-risk group.

Compiled from Sina Technology