Multiparametric MRI deep learning pipeline segments glioblastoma with high agreement to reference masks in 50 patients

Deep learning segmentation of glioblastoma on multiparametric MRI achieved median Dice scores of 0.90 (whole tumor), 0.94 (tumor core), 0.86 (enhancing tumor) in 50 UCSF-PDGM subjects.

• The pipeline combined deep learning segmentation with anatomical parcellation, generating regional tumor burden 'hit-plots' that map tumor compartments to brain structures.
• In a longitudinal LUMIERE subject (Patient-048, six timepoints), the pipeline recovered concordant volumetric trajectories and evolving regional-burden patterns against independent segmentations.
• The study used standard multiparametric MRI sequences (T1, T1-Gad, T2, FLAIR) without atlas coregistration.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.