Abstract: Radiomics, by extracting and analyzing quantitative features of images and constructing machine learning models, has shown great potential in the process of tumor diagnosis, treatment and monitoring. Recently, radiomics research has begun to take shape, and the methods and procedures are also maturing and improving, but there are still some difficulties in clinical application. From the perspective of practicality, this paper briefly summarizes the various processes of radiomics and its corresponding analytical tools and methods, and focuses on the application, shortcomings and difficulties of radiomics in the study of tumor risk assessment, prognosis prediction, and therapeutic response prediction, with a view to providing new technologies and methods for accurate diagnosis and treatment of tumors.