Multiscale Spatial Characterization of Glioma MRI with MUSE and Radiomics

In 416 UCSF-PDGM glioma patients, native MUSE semivariance strongly correlates with half GLCM contrast at 9,371 aligned offsets (Pearson r = 0.954).

• Why it matters: Directional radiomics can lose physical lag distance and orientation, limiting how well glioma spatial heterogeneity is characterized and linked to molecular status like IDH.
• What they did: The study used postcontrast T1-weighted MRI and enhancing-tumor segmentations, comparing 3D MUSE variographic fields (328 directional lag distances) with directional GLCM features and testing IDH associations via patient-level permutation.
• The result: In 364 grade-4 tumors, MUSE remained associated with IDH after conventional adjustment (global permutation p = 0.00050) but not after primary directional GLCM adjustment (p = 0.256), while a residual 3D CNN reached out-of-fold IDH ROC-AUC 0.792.

AI summary of the abstract · may contain errors · not medical advice · read the paper.