Metabolic associated fatty liver disease （MAFLD） is a common chronic liver disease worldwide， and timely and precise intervention can delay disease progression and significantly reduce the risk of serious complications such as liver fibrosis， liver cirrhosis， and liver cancer. Although traditional liver biopsy combined with metabolic markers is the gold standard， it may cause complications such as pain and bleeding as an invasive examination， which has promoted scientific research to shift its focus to the construction of noninvasive assessment systems. In recent years， noninvasive diagnostic technologies based on multi-dimensional detection strategies have been continuously updated， including serological models， imaging techniques， and clinical algorithms. This article systematically reviews the screening methods for MAFLD during the fibrotic stages F1—F3， especially deep learning models based on artificial intelligence， in order to provide ideas for the early screening of MAFLD， as well as a scientific reference for optimizing disease management strategies.