BRaTS 2021 Dataset

The dataset used in the BraTS challenge consists of a large collection of multi-institutional, clinically acquired multi-parametric MRI (mpMRI) scans of patients with gliomas. These data include pathologically confirmed diagnoses as well as information on MGMT promoter methylation status. The dataset is divided into training, validation, and testing subsets to evaluate the performance of brain tumor segmentation models.

The images are provided in NIfTI format (.nii.gz) and include multiple imaging modalities: native T1, post-contrast T1-weighted (T1Gd), T2-weighted, and FLAIR. The scans are acquired using different clinical protocols and imaging devices, reflecting real-world variability. Each case is manually annotated by one or more expert neuroradiologists following a standardized protocol. The annotations identify key tumor sub-regions, including enhancing tumor (ET), peritumoral edema (ED), and necrotic tumor core (NCR). Additionally, all data have been preprocessed to ensure consistency, including spatial co-registration, uniform resolution, and skull stripping.