The present invention provides a radiomics-based tumor classification system, including: lesion area extraction, used to extract lesion data; heterogeneous area extraction, using a clustering method to extract all initial heterogeneous areas in the lesion data, and then Determine the final clustering result based on the connectivity and number of pixels of the clusters on the image, and find out multiple heterogeneous regions; parameter vector extraction, extract the shape, state and other parameters of each heterogeneous region; type group extraction, The clustering method is used to divide each heterogeneous area into clusters to obtain multiple categories. New lesion type determination is used to determine which type each heterogeneous type of new lesion belongs to. The new lesion typing result is used to output the new lesion typing result. The tumor typing system of the present invention has the advantages of non-invasiveness, non-intervention, repeatability, etc., and its analysis object is based on all lesion tissues, and the information is comprehensive.