Imaging prediction of molecular subtypes in adult-type diffuse glioma considering spatial and temporal heterogeneity

Motivation: Imaging-based molecular prediction is crucial for identifying potential therapeutic targets in adult-type diffuse gliomas. Goal(s): To assess spatial and temporal heterogeneity using diffusion and perfusion MRI and to evaluate isocitrate dehydrogenase (IDH) mutation and epidermal growth factor receptor (EGFR) amplification status in primary and recurrent adult-type diffuse gliomas. Approach: Histogram features and temporally stable imaging features were identified. Models were constructed for IDH and EGFR prediction in cross-sectional and longitudinal sets. Results: The models achieved high performance for predicting IDH mutations, but they exhibited poor predictability in EGFR amplification, especially in recurrent tumors. Impact: This study enables reliable, non-invasive prediction of IDH mutation in gliomas using multiparametric MRI, informing personalized treatment. Caution is advised for EGFR prediction due to its spatial and temporal variability.