全球约30%人口患有代谢功能障碍相关脂肪性肝病（MASLD），若趋势持续，到2040年可能接近50%。该疾病长期隐匿无症状，常在进展至纤维化、肝硬化或癌症后才被发现，尤其在南亚和东南亚与2型糖尿病和代谢综合征并行上升，形成公共卫生风险。传统筛查依赖血检、影像和活检，受成本与可及性限制，且往往错过早期干预窗口。

阿里巴巴达摩院开发的AI模型MAOSS利用既有临床数据，尤其是大量非增强CT扫描，实现“机会性筛查”。系统结合影像细微模式与年龄、BMI及血液指标，识别早期脂肪肝并评估进展风险。其训练基于2,071例确诊患者数据，在超过1,000例活检验证队列中，传统方法识别中高风险比例约17%，而AI方法提升至52%，同时有效排除低风险人群，减少不必要随访并缓解医疗资源压力。

该技术建立在多国多癌筛查实践基础上，并与大型语言模型结合，形成“检测+解释”的双层智能：前者发现人眼不可见模式，后者生成风险说明与临床指导。局限包括回顾性研究性质、数据质量、监管及医生信任等，需前瞻性与跨区域验证。总体趋势显示医学正由症状后反应转向早期风险预测，AI嵌入常规医疗流程，可能改变慢性病管理路径。

Roughly 30% of the global population has metabolic dysfunction-associated steatotic liver disease (MASLD), with projections approaching 50% by 2040. The disease is largely asymptomatic and often detected only after progression to fibrosis, cirrhosis, or cancer, especially across South and Southeast Asia where it rises alongside Type 2 diabetes and metabolic syndrome, creating a public health risk. Traditional screening relies on blood tests, imaging, and biopsy, constrained by cost and access, and frequently misses the early intervention window.

Alibaba DAMO Academy’s AI model, MAOSS, leverages existing clinical data, particularly large volumes of non-contrast CT scans, enabling “opportunistic screening.” By combining subtle imaging patterns with age, BMI, and blood markers, it detects early fatty liver and estimates progression risk. Trained on 2,071 confirmed cases and validated on over 1,000 biopsy-proven patients, traditional pathways identified intermediate- to high-risk cases about 17% of the time, while the AI approach increased this to 52%, also effectively ruling out low-risk patients to reduce unnecessary follow-ups and resource strain.

Built on multicancer screening deployments across several countries, the system integrates with large language models to form a two-tier intelligence: detection of patterns invisible to humans and translation into actionable clinical guidance. Limitations include retrospective design, data quality, regulation, and physician trust, requiring prospective and cross-regional validation. The broader shift moves medicine from reactive symptom-based care toward early risk prediction, embedding AI into routine workflows to reshape chronic disease management.