Alibaba DAMO Academy Develops MAOSS Model to Significantly Improve Detection Rate of High-Risk Fatty Liver Patients

Recently, Alibaba DAMO Academy collaborated with several hospitals to develop MAOSS, an AI model for fatty liver screening. Using plain CT images and serum indicators, the model can simultaneously determine fatty liver stages and liver fibrosis progression, increasing the detection rate of high-risk patients from 16.6% to 52.4%. Related papers have been published in Nature Communications. Fatty liver affects over 30% of people, but conventional tests lack sensitivity and specialized exams are costly. MAOSS extracts high-dimensional liver features and uses multi-modal data training; its AUC for fatty liver staging ranges from 0.904 to 0.917, higher than radiologists' 0.709. With AI assistance, doctors' diagnostic AUC rises to 0.798, effectively reducing missed diagnosis risks. The model uses existing plain CT data without extra costs, providing early warnings for chronic liver disease intervention. DAMO Academy also focuses on screening for pancreatic cancer, gastric cancer, etc., exploring multi-disease screening methods.