MRI habitat-derived radiomics Identifies novel glioblastoma subtypes with distinct tertiary lymphoid structure-associated gene profiles

Motivation: The lack of noninvasive biomarkers to identify tumor subtypes with distinct tertiary lymphoid structure (TLS) -associated transcriptomic profiles hinders the precise prediction of neoadjuvant immunotherapy efficacy in glioblastoma (GBM). Goal(s): To identify glioblastoma subtypes with distinct TLS-associated transcriptomic profiles by habitat-derived radiomics features based on T2-weighted imaging. Approach: Paired analysis of MRI and transcriptomic profiles in GBM patients was conducted to identify tumor subtypes, the radiomics-based model for discirminating subtypes was developed and validated. Results: Two GBM subtypes were identified based on TLS-associated gene characterizations with differential survivals. A habitat-derived radiomics model was developed and validated to classify subtypes with excellent performance. Impact: We developed and validated a classifier based on habitat-derived radiomics features to identify novel GBM subtypes with distinct TLS-associated gene profiles. This may provide a new in vivo approach for precise evaluation of neoadjuvant immunotherapy response in GBM noninvasively.