Research and education use only
A research demonstration, not a medical device. This model is not validated or approved for clinical use here; do not use it for diagnosis, treatment or any other clinical decision. Segmentations can be wrong; always review them against the images.
Your images stay on this computer
Files are read and processed inside this browser tab on your GPU or CPU. No image data or results are uploaded. Only the page, the model files and the demo cases are downloaded, and the model is cached for later visits.
What you need
- A recent desktop Chrome or Edge (Firefox and Safari may work on the CPU).
- 8 GB RAM or more. WebGPU is fastest; on the CPU a scan takes one to several minutes.
- A non-contrast axial head CT in Hounsfield units.
CLAIM aSAH Charité Berlin · nnU-Net multiclass
Six classes on non-contrast CT after aneurysmal SAH: SAH, intraventricular hemorrhage, ventricles, intracerebral hemorrhage, aneurysm and subdural hematoma.
models/claim/ folder; convert the CC BY-NC 4.0 weights from Zenodo with tools/claim_export.py to enable this panel.Drop a head CT here
A zip of a DICOM study, a DICOM folder, loose DICOM files, a .nii / .nii.gz or a .nrrd volume. The 2D model works slice by slice at 0.445 mm in plane; it was trained on 5.5 to 6.5 mm slices, so thin slices are averaged to 5 mm. The demo cases include the expert labels, so Dice is shown.
Citation, licence and disclaimer
Model. Kiewitz J, Aydin OU, Hilbert A, Gultom M, Nouri A, Khalil AA, Vajkoczy P, Tanioka S, Ishida F, Dengler NF, Frey D. Deep learning-based multiclass segmentation in aneurysmal subarachnoid hemorrhage. Front Neurol 2024;15:1490216. doi:10.3389/fneur.2024.1490216 (preprint: medRxiv 2024.06.24.24309431)
Code and weights. CLAIM project page · github.com/claim-berlin/aSAH-multiclass-segmentation · weights on Zenodo doi:10.5281/zenodo.12579491. Licensed CC BY-NC 4.0 (attribution, non-commercial use only); the nnU-Net 2D network is converted to ONNX without changes to the weights. nnU-Net: Isensee F, Jaeger PF, Kohl SAA, Petersen J, Maier-Hein KH. Nat Methods 2021;18:203–211 (Apache-2.0).
Demo data. Three cases from the BHSD dataset as distributed in the CLAIM repository: Wu B, Xie Y, Zhang Z, et al. BHSD: a 3D multi-class brain hemorrhage segmentation dataset. In: Machine Learning in Medical Imaging (MLMI 2023), LNCS 14348. Springer; 2024. doi:10.1007/978-3-031-45673-2_15. The reference labels shown with the demo are those supplied with the CLAIM demo data.
How it runs here. As nnUNetv2_predict for the 2D configuration: clip to −3…88 HU, cubic-spline resampling in plane to 0.4453 mm, z-score (mean 32.47, SD 27.41), 512 × 512 tiles with Gaussian weighting, argmax. Checked against nnUNetv2_predict with the published weights on the three demo cases (see the README). The authors report similar accuracy for their 2D and 3D models and used the 2D model for external validation.
Disclaimer. For research and education only. Not a medical device; not cleared or approved by the FDA, CE or any regulator; not validated for clinical use; must not be used for diagnosis, triage, treatment or any other clinical decision. Outputs can be wrong and must be reviewed against the images by a qualified reader. Provided "as is" without warranty of any kind. Images are processed locally in this browser and are not uploaded.