Multi-center glioblastoma MRI dataset with expert-validated segmentations and MGMT profiling

A dataset of 337 glioblastoma patients’ MRI with expert-refined tumor segmentations (mean Dice 0.94) and MGMT methylation prediction baseline accuracy 0.60 supports radiomic and deep learning research.

• The dataset comprises T1 pre-/post-contrast, T2, and FLAIR sequences from eight hospitals, preprocessed with skull stripping and MNI152 registration.
• Automated segmentation by SegResNet was refined by neuroradiologists, achieving a mean Dice of 0.94.
• Baseline MGMT methylation prediction accuracy was 0.60, highlighting a challenging radiogenomic task.

Automated summary

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