BraTS2020: Fixed Patient Splits for Brain Tumor Segmentation

A fixed patient split for the BraTS2020 brain tumor segmentation challenge. The split is provided in JSON, TXT, and CSV formats, likely defining which patient scans belong to training, validation, and test sets. The original data source and author are not specified in the provided metadata.

Use Cases

• Reproduce published BraTS2020 segmentation experiments based on the defined patient split.
• Benchmark new segmentation algorithms against a standard training and validation cohort.
• Ensure fair model comparison by using a consistent test set partition.
• Manage data loading pipelines by pre-defining patient IDs for different dataset folds.

Strengths

• Provides a fixed split, which is essential for reproducible machine learning research.
• Offers the split in multiple common formats (JSON, TXT, CSV) for flexibility.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Row count and the specific patient IDs included in the split are unknown.
• Column-level documentation is absent; field semantics must be inferred after download.

Provenance

Source

Kaggle

Collection Method

Likely derived from the official BraTS2020 challenge data.