ADNI Mini
A 1,000-subject ADNI sMRI evaluation subset derived from medarc/adni_eval style metadata.
The dataset has a single eval split saved as Hugging Face Arrow shards with embedded processed T1w NIfTI image bytes. Brain masks are intentionally omitted.
Cohort
- 1,000 subjects / 1,000 scans
- One processed T1w scan per subject
- Sex: 500 Female / 500 Male
- Diagnosis: CN 333 / MCI 334 / AD 333
- MCI conversion labels: 167 non-converter / 167 converter within 36 months
- Tau status labels: 214 negative / 214 positive, 428 known total
- Amyloid status labels: 357 negative / 523 positive, 880 known total
- CSF Aβ labels: 712 known
- CSF p-tau labels: 709 known
- CSF t-tau labels: 711 known
Features
The feature schema matches medarc/adni_eval except that the mask feature is removed.
The image feature is nifti.
from datasets import load_dataset
ds = load_dataset("medarc/adni-mini", split="eval")
print(ds)
Local source selection: all_task_tau_optimized_exact_age_1000.csv.
- Downloads last month
- 229