Biopsy-Validated Deep Learning Mapping of Glioma Infiltration Beyond Contrast Enhancement Using Multiparametric MRI

A deep learning infiltrative Tumor Burden (iTB) map identifies glioblastoma infiltration beyond contrast enhancement, boosting independent test accuracy from 74.5% to 85.5% and AUC from 0.859 to 0.942.

• Why it matters: Glioblastoma infiltration often extends into non-enhancing FLAIR-hyperintense tissue that conventional MRI cannot visualize, leaving tumor burden hard to localize for planning and assessment.
• What they did: Using preoperative multiparametric MRI from 80 WHO grade II–IV glioma patients and 24 brain metastasis patients, a 3D convolutional neural network was trained on voxel-centered mpMRI patches with biopsy-validated histopathology ground truth.
• The result: Augmentation optimization significantly improved independent test-set performance (p = 0.008), and voxel-wise iTB maps showed low infiltration probabilities in brain metastases and extensive elevated probabilities in glioblastoma.