| --- |
| license: other |
| pretty_name: ADNI sMRI Eval v0.2 |
| --- |
| |
| # ADNI sMRI Eval v0.2 |
|
|
| Single-split ADNI structural MRI evaluation dataset. The repository contains one Hugging Face `DatasetDict` split named `eval`. |
|
|
| This release keeps all usable processed T1w subject-sessions with matching ADNIMERGE clinical metadata, SynthSeg QC/volume outputs, and brain masks. It is not balanced and includes CN, MCI, and AD labels, with ADNIMERGE `Dementia` mapped to AD. |
|
|
| Summary from `eval_report.json`: |
|
|
| - sessions: 6,844 |
| - subjects: 2,042 (mean 3.4 sessions/subject, max 13) |
| - diagnosis counts: CN 2,560; MCI 2,503; AD 1,781 |
| - sex counts: Female 3,136; Male 3,708 |
| - age: mean 75.4 (range 50.4–95.6) |
|
|
| Load directly from the Hub: |
|
|
| ```python |
| from datasets import load_dataset |
| |
| dataset = load_dataset("medarc/adni_eval") |
| eval_ds = dataset["eval"] |
| ``` |
|
|
| Or snapshot the repository and load from disk: |
|
|
| ```python |
| from datasets import load_from_disk |
| |
| dataset = load_from_disk("/path/to/adni_eval") |
| eval_ds = dataset["eval"] |
| ``` |
|
|
