spc819/rsna2025-aneurysm-26class-seg

RSNA 2025 Aneurysm — 26-class Vessel-Anatomy and Aneurysm Segmentation Labels

26-class vessel-anatomy and aneurysm segmentation labels (13 vessel anatomy classes + 13 aneurysm location classes, values 0–26, see labels.json) for the RSNA 2025 Intracranial Aneurysm Detection challenge, placed back into the original image space (pure voxel placement, no resampling).

Paper: arXiv:2606.26706

Contents

Format

• —NIfTI, uint8, affine identical to the paired image (spacing / origin / direction).
• —Placement: pure voxel placement, no resampling; voxels outside the ROI are 0.
• —File names are SeriesInstanceUID.nii.gz, matching the image name of the same UID.

Reproducing the images

Per RSNA rules, the original DICOMs must be downloaded from Kaggle: https://www.kaggle.com/competitions/rsna-intracranial-aneurysm-detection/data

Processing code (DICOM → NIfTI): https://github.com/PengchengShi1220/RSNA2025_Intracranial-Aneurysm-Detection/blob/master/process_RSNA2025_all_data.py

See REPRODUCE.md for the exact pipeline.

The 27 multiframe cases

These are multiframe MR series whose DICOM headers required special handling (direction/spacing) during conversion. Their labels live in the coronal/sagittal space produced by dcm2niix, which differs from the axial output of process_RSNA2025_all_data.py. Do not overlay them on imagesTr/<uid> directly — reproduce their paired images via dcm2niix on the same multiframe DICOMs.

License

• —Labels: CC-BY-NC 4.0 (non-commercial use only, aligned with the RSNA 2025 competition winner license). These are derived annotations; the original images are not redistributed — download them from the official RSNA/AWS source (see above).

Please cite our work if it is helpful for your research:

@misc{shi2026intracranialaneurysmclassificationsegmentation,
      title={Intracranial Aneurysm Classification and Segmentation via Tri-Axial ROI and Multi-Task Learning},
      author={Pengcheng Shi and Kaiyuan Yang and Houjing Huang and Jiawei Chen and Yan Lu and Jiaqi Liu and Murong Xu and Minghui Zhang and Bjoern Menze and Xinglin Zhang},
      year={2026},
      eprint={2606.26706},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2606.26706},
}