RHUH SAH dataset in the browser

Open non-contrast CT scans of aneurysmal SAH with expert bleeding masks from Río Hortega University Hospital, Valladolid (Cepeda & García-García, Zenodo 2023). Load a case and its mask, run SAHVAI (BrainHemoAI), CLAIM, BLAST-CT or the Valladolid SAH Swin UNETR on it and compare with the expert. Everything runs on this computer.

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.
Compute
R

RHUH SAH dataset Cepeda & García-García · Valladolid, Spain

Open non-contrast CT scans of aneurysmal SAH with expert manual bleeding masks. Load one case and its mask, run any of the four models on it and compare with the expert segmentation.

Río Hortega University Hospital · Zenodo
  1. Download subarachnoid_hemorrhage_rhuh.rar (649 MB) from Zenodo record 8228847 and extract it (browsers cannot open RAR archives).
  2. Drop the two NIfTI files of one case here (the CT and its mask), or choose them below.
  3. Pick a model and press Run. The expert mask is drawn in white and Dice and volumes are compared as "any blood".

Drop one RHUH case here (CT + expert mask, .nii / .nii.gz)

CT: none · mask: none

or try a hosted RHUH case with its expert mask:
About these scans. RHUH volumes are registered to the MNI template (193 × 229 × 193 at 1 mm) and skull-stripped: brain only, background 0, intensities close to Hounsfield units. The three models were trained on whole-head CT, so the page detects this, sets the background to air (−1000 HU) and, for SAHVAI, takes the brain mask from the scan. Treat the results as approximate. On cases 001–003 the any-blood Dice against the expert mask was 0.65, 0.27 and 0.62 for SAHVAI (BrainHemoAI), 0.12, 0.02 and 0.37 for CLAIM, and 0.22 for BLAST-CT (case 001). The Valladolid SAH Swin UNETR (option 4) works in exactly this template space and scored 0.77, 0.70 and 0.82, but it comes from the same hospital and these cases may have been used to develop it.

Citation, licence and disclaimer

Dataset. Cepeda S, García-García S. Non-contrast CT scans dataset for subarachnoid hemorrhage: bleeding segmentation [data set]. Zenodo; 2023. Version 1. doi:10.5281/zenodo.8228847. Licensed CC BY 4.0. Río Hortega University Hospital, Valladolid, Spain. Cases 001–003 are hosted here unchanged as demos, under the same licence.

Related work by the dataset authors. García-García S, Cepeda S, Müller D, Mosteiro A, Torné R, Agudo S, de la Torre N, Arrese I, Sarabia R. Mortality prediction of patients with subarachnoid hemorrhage using a deep learning model based on an initial brain CT scan. Brain Sci 2024;14(1):10. doi:10.3390/brainsci14010010 · García-García S, Cepeda S, Arrese I, Sarabia R. A fully automated pipeline using Swin transformers for deep learning-based blood segmentation on head computed tomography scans after aneurysmal subarachnoid hemorrhage. World Neurosurg 2024. doi:10.1016/j.wneu.2024.07.216 · Code: smcch/Subarachnoid_Hemorrhage_segmentation_and_mortality_prediction (Academic Non-Commercial Source Code License). Their Swin UNETR segmentation model runs in this browser on the SAH Swin UNETR page and as option 4 here.

Data you drop here stays in this browser. The comparison is between a public research dataset and research models; it is not a validation of any model for clinical use.

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.

Log