RÉSUMÉ
Solid tumors have spatial and temporal heterogeneity,which limits the use of traditional invasive biopsy but gives huge potential for medical imaging,because medical imaging can capture intra-tumoural heterogeneity by a non-invasive way.During the past decades,medical imaging innovations with updating hardware,new imaging agents and standard protocols,make the field move towards the era of quantitative analysis,promote the development of imaging date,automatic algorithm,reproducible analysis and intelligent recognition tools,which make it possible to use high-throughput extracted information in clinic (radiomics).A systematic methodology,including diagnosis based on radiology images and evaluation of treatment and prognosis of tumors,can be constructed by getting the image features in the multi-source data platform,combining with gene and clinical information,mining and screening the feature sets which highly correlate to the diagnosis of malignant tumors.Radiomics is not only a chance for imaging,but also facing sever challenges,which needs further validation in multicentric settings and in multidisciplinary laboratory.