Nowadays, the research on traditional radiomics has gradually matured. However, it usually regards the tumor as a whole, and high-throughput data are often generated in the entire tumor region, which cannot express clear spatial heterogeneity. In order to explore the potential biological information within tumors and realize individualized precise diagnosis and treatment, habitat analysis technology emerges at the historic moment, which provides a new way of thinking to identify tumor microenvironment. On the basis of traditional radiomics, the tumor cell population with similar characteristics is clustered, and the tumor is segmented into multiple sub-regions. Therefore, the study of tumor is no longer limited by the subjective differences of observers in the description of imaging features, and the information of tumor spatial heterogeneity is ideally obtained.

Neoplasms; Tumor microenvironment; Radiomics; Machine learning