Heterogeneity Analyzed by
CT
‐Based Habitat Analysis for Clinical Management of Cancers: A Narrative Review
Yang Zhang, Zhi‐Yan Jin, Min‐Ying Lydia Su ABSTRACT
Tumors are highly heterogeneous, and whole‐lesion radiomics analysis is a popular method for extracting texture features that reflect this heterogeneity, which can be used to build models for cancer diagnosis, staging, therapy response evaluation, and prognosis prediction. However, there is no spatial information for understanding which components provide the most critical information. Recently, habitat analysis has gained widespread attention, and it is used to divide a tumor into distinct sub‐regions, called habitats, that share common characteristics. CT reveals heterogeneity in attenuation, texture, entropy, and enhancement patterns, which can be used to generate habitats. In the past 2 years, many studies exploring the role of CT habitats have been reported. The divided habitats can be used to build models for the classification of histological or molecular subtypes, the evaluation and prediction of treatment response, and prognostication of progression‐free survival or overall survival. This review aims to summarize studies on the application of CT‐habitat analysis for cancer management. In addition to clinical applications, various methods for habitat segmentation will be reviewed. Lastly, how spatial information revealed by the habitat can guide tissue biopsy for tissue‐level confirmation studies will be described, which is critical for understanding the underlying biology and paving the way for future clinical implementation.