# ISLES 2018: CT perfusion of acute ischemic stroke with DWI-defined infarct core masks (103 patients, 156 cases)

## Facts

- Provider: ISLES challenge organizers (Mauricio Reyes, Ezequiel de la Rosa, Bjoern Menze et al.); mirrored by MedOtter
- License: CC-BY-4.0 (https://zenodo.org/records/17736412)
- Size: 2,066 files
- Format: JSON, NIfTI, JSON Lines (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/MedOtter/isles2018
- Access: public, readable without an account
- DOI: https://doi.org/10.5281/zenodo.17736412
- Category: Neuroscience
- Source: https://huggingface.co/datasets/MedOtter/isles2018
- Page: https://scigantic.com/archive/a01cbac6-da3d-5ef9-9bb7-302561f1444c
- Keywords: Neuroscience, CT, image segmentation, JSON, NIfTI, JSON Lines, Hugging Face

## About

The Ischemic Stroke Lesion Segmentation (ISLES) 2018 challenge data: CT perfusion (CTP) of acute anterior-circulation ischemic stroke for infarct core segmentation, from 103 patients and 156 CTP acquisitions pooled from the DEFUSE 2 study and a companion prospective cohort across four centers (three in the US, one in Australia). Limited z-coverage meant some patients were scanned as two slabs, each its own case folder; group on patient_id in cases.jsonl to avoid leakage. Each case has the baseline CT, the raw motion-corrected 4D CTP series (CT_4DPWI, about 1 s temporal resolution) and derived perfusion maps CT_CBF, CT_CBV, CT_MTT and CT_Tmax. TRAINING (94 cases, 63 patients) includes the OT infarct mask, drawn on MR diffusion-weighted images acquired within 3 hours of CTP and registered to CTP space; TESTING (62 cases) has no released ground truth. About 11.8 GB (measured 2026-11-17), files byte-identical to the organizers' Zenodo re-release. Load with nibabel or SimpleITK. Mirrored on Hugging Face by MedOtter from the authors' release, with a card documenting provenance and any normalization. The organizers' Zenodo record 17736412 (ISLES 2015 to 2018) is CC BY 4.0, and the archives also carry an ODbL notice; cite Cereda et al. 2016 and the ISLES 2018 challenge.

## Load the data

```python
from datasets import load_dataset
ds = load_dataset("MedOtter/isles2018")
print(ds)
```

## Data structure

File types in sampled files: JSON, NIfTI, JSON Lines

Example file: `cases.jsonl`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/a01cbac6-da3d-5ef9-9bb7-302561f1444c
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/a01cbac6-da3d-5ef9-9bb7-302561f1444c
- Over MCP: `get_schema_card` with id `a01cbac6-da3d-5ef9-9bb7-302561f1444c` at https://api.scigantic.com/mcp

## Related topics

- [Neuroscience datasets](https://scigantic.com/topics/neuroscience)
- [JSON datasets](https://scigantic.com/topics/json)
- [Image segmentation datasets](https://scigantic.com/topics/image-segmentation)
- [CT datasets](https://scigantic.com/topics/ct)
- [JSON Lines datasets](https://scigantic.com/topics/json-lines)
- [NIfTI datasets](https://scigantic.com/topics/nifti)
