{"success":true,"data":{"author":"NeuroSpin, CEA (nidl team)","bucketAccessType":"public","bucketName":"nidl-neurospin/openmind","bucketPrefix":"","category":"Neuroscience","cloudProvider":"huggingface","createdAt":"2026-11-21T20:45:58.973Z","description":"Preprocessed versions of the OpenMind collection of head and neck 3D magnetic resonance imaging (MRI) pooled from OpenNeuro, produced at NeuroSpin (CEA, France) with the brainprep pipelines and packed as WebDataset shards for large-scale and self-supervised deep learning. quasiraw-v2.1.1 holds 71,490 brain images (42,941 T1-weighted, 23,935 T2-weighted, 4,614 FLAIR) from 33,738 subjects in 787 OpenNeuro datasets, loaded as 182 by 218 by 182 voxel volumes. cat12vbm-12.8.2 holds CAT12 voxel-based morphometry (VBM) grey matter maps and ROI morphometry such as total intracranial volume for 42,951 T1w scans, with CAT12 quality ratings. One shard folder per OpenNeuro dataset, an index.json.gz of byte offsets for random access, participants.tsv (age, sex, BMI, handedness, health status) and val_split.csv, a held-out validation split of whole datasets stratified on age, sex and BMI. Load with the nidl library's OpenMind PyTorch dataset, which streams by HTTP range requests. License CC BY 4.0; cite OpenMind (Wald et al., 2025) and the OpenNeuro datasets used.","displayTitle":"OpenMind Preprocessed Brain MRI: Quasi-Raw T1w, T2w and FLAIR plus CAT12 VBM Grey Matter Maps from 787 OpenNeuro Datasets (CEA NeuroSpin)","huggingFaceUrl":"https://huggingface.co/datasets/nidl-neurospin/openmind","id":"3ecb5041-7299-5820-9779-ae8ed5d8720c","license":"CC-BY-4.0","licenseEvidenceUrl":"https://huggingface.co/datasets/nidl-neurospin/openmind","objectCountEstimate":1639,"probeStatus":"ok","referenceUrl":"https://huggingface.co/datasets/nidl-neurospin/openmind","region":"us","schemaCard":{"accessHints":[{"language":"python-datasets","snippet":"from datasets import load_dataset\nds = load_dataset(\"nidl-neurospin/openmind\")"},{"language":"python-hub","snippet":"from huggingface_hub import snapshot_download\nsnapshot_download(repo_id=\"nidl-neurospin/openmind\", repo_type=\"dataset\")"}],"caveats":["HuggingFace metadata: author=nidl-neurospin, license:cc-by-4.0, size_categories:10K<n<100K, downloads=1182"],"format":"mixed","generatedAt":"2026-11-21T20:48:49.466Z","layout":{"example":"participants.tsv","pattern":"{repoId}/{filename}"},"mountPath":"/mnt/archive","sampleHeaders":[{"bytes":"dataset_id\tparticipant_id\tsession\trun\tage\tsex\tbmi\trace\thandedness\thealth_status\nds000001\tsub-01\tses-DEFAULT\t17328\t26.0\tfemale\tn/a\tn/a\tn/a\tn/a\nds000001\tsub-02\tses-DEFAULT\t22351\t24.0\tmale\tn/a\tn/a\tn/a\tn/a\nds000001\tsub-03\tses-DEFAULT\t71475\t27.0\tfemale\tn/a\tn/a\tn/a\tn/a\nds000001\tsub-04\tses-DEFAULT\t25273\t20.0\tfemale\tn/a\tn/a\tn/a\tn/a\nds000001\tsub-05\tses-DEFAULT\t18966\t22.0\tn/a\tn/a\tn/a\tn/a\tn/a\nds000001\tsub-06\tses-DEFAULT\t24505\t26.0\tfemale\tn/a\tn/a\tn/a\tn/a\nds000001\tsub-07\tses-DEFAULT\t33769\t24.0\tmale\tn/a\tn/a\tn/a\tn/a\nds000","encoding":"utf8","key":"participants.tsv","truncatedAt":512}],"sampleObjects":["cat12vbm-12.8.2/ds000001/ds000001-000000.tar","cat12vbm-12.8.2/ds000002/ds000002-000000.tar","cat12vbm-12.8.2/ds000003/ds000003-000000.tar","cat12vbm-12.8.2/ds000005/ds000005-000000.tar","cat12vbm-12.8.2/ds000006/ds000006-000000.tar"],"sidecars":[{"bytes":"---\nlicense: cc-by-4.0\npretty_name: OpenMind (preprocessed, NeuroSpin)\ntask_categories:\n- image-feature-extraction\ntags:\n- 3d\n- mri\n- neuroimaging\n- brain\n- medical\n- webdataset\n- self-supervised-learning\n- voxel-based-morphometry\n- cat12\nsize_categories:\n- 10K<n<100K\nconfigs:\n- config_name: quasiraw-v2.1.1\n  data_files:\n  - split: train\n    path: quasiraw-v2.1.1/*/*.tar\n- config_name: cat12vbm-12.8.2\n  data_files:\n  - split: train\n    path: cat12vbm-12.8.2/*/*.tar\n---\n\n# OpenMind: preprocessed brain MRI (NeuroSpin)\n\nPreprocessed versions of the [OpenMind dataset](https://huggingface.co/datasets/MIC-DKFZ/OpenMind)\n(Wald et al., 2025): head & neck 3D MRI pooled from ~800 [OpenNeuro](https://openneuro.org)\ndatasets. Here the brain T1w, T2w and FLAIR images were preprocessed at\n[NeuroSpin](https://joliot.cea.fr/drf/joliot/en/Pages/research_entities/NeuroSpin.aspx)\nusing [brainprep](https://github.com/brainprepdesk/brainprep) and packed as\n[WebDataset](https://github.com/webdataset/webdataset) shards, ready for large-scale\n(e.g. self-supervised) deep learning.\n\n| Folder | Pipeline | Status |\n|---|---|---|\n| `quasiraw-v2.1.1/` | brainprep quasi-raw v2.1.1 (T1w, T2w, FLAIR) | available |\n| `cat12vbm-12.8.2/` | CAT12 12.8.2 VBM via brainprep v2.1.0 (T1w): grey matter maps + ROI morphometry | available |\n\nAll folders share the same sample identifiers: (`dataset_id`, `subject`, `session`, `run`),\nwith `dataset_id` the shard folder name, so derivatives of the same scan can be joined\nacross folders.\n\n## Repository layout\n\n```\nREADME.md\nval_split.csv                 # recommended validation split, for all folders (see below)\nparticipants.tsv              # per scan: age, sex, bmi, race, handedness, health_status\nquasiraw-v2.1.1/              # see below\ncat12vbm-12.8.2/              # see below\n```\n\nEach preprocessing folder holds one folder of WebDataset shards per OpenNeuro dataset and an\n`index.json.gz` with the byte offsets of every file of the shards, used by\n[nidl](#usage-with-nidl) for random access.\n\n## Validation split\n\n`val_split.csv` (columns `dataset_id, participant_id, session, run`) defines a held-out\nvalidation set made of **whole datasets** (no subject or dataset is shared with training),\nstratified on age, sex and BMI. It holds ~10% of the images. Only the rows listed in the \ncsv belong to validation. The others belong to training.\n\nThe split is given by scan identifiers, shared by all folders: it applies to\n`quasiraw-v2.1.1` and `cat12vbm-12.8.2` alike.\n\n## Usage with nidl\n\n[nidl](https://github.com/neurospin-deepinsight/nidl) provides an `OpenMind`\nPyTorch `Dataset` that reads the shards by byte offset (from `index.json.gz`): random access,\nany sampler and any number of `DataLoader` workers, modality pairing and the subject-level\ntrain / val split. By default the images are streamed from this repository with HTTP range\nrequests (nothing is written to disk but the index and the split); `streaming=False`\ndownloads the needed shards first.\n\n```bash\npip install \"git+https://github.com/neurospin-deepinsight/nidl\"\n```\n\n```python\nfrom torch.utils.data import DataLoader\nfrom nidl.datasets import OpenMind\n\nroot = \"openmind\"  # local mirror of this repository, files are fetched on demand\n\n# Quasi-raw: one T1w image per item, subject-level split from <root>/val_split.csv\ntrain = OpenMind(root, preprocessing=\"quasiraw\", modality=\"t1\", split=\"train\", apply_mask=True)\nval = OpenMind(root, preprocessing=\"quasiraw\", modality=\"t1\", split=\"val\", apply_mask=True)\nprint(train[0].shape)  # (1, 182, 218, 182)\n\n# Quasi-raw: paired T1w / T2w acquired in the same session\npaired = OpenMind(root, modality=(\"t1\", \"t2\"), group_by=\"session\", split=\"train\")\nsample = paired[0]  # {\"t1\": array, \"t2\": array}\n\n# CAT12 VBM: grey matter map and ROI morphometry\nvbm = OpenMind(root, preprocessing=\"cat12vbm\", split=\"train\", return_morphometry=True)\ngm, morphometry = vbm[0]\nprint(gm.shape, morphometry[\"total_volumes\"][\"TIV\"])  # (1, 113, 137, 113) ...\n\nloader = DataLoader(train, batch_size=8, shuffle=True, num_workers=8)\n```\n\n`nidl` does not read the CAT","encoding":"utf8","key":"README.md","name":"README.md","size":15211,"truncated":true}],"sourceObjectKey":"participants.tsv","starterCell":"import pandas as pd\ndf = pd.read_csv(\"hf://datasets/nidl-neurospin/openmind/participants.tsv\", sep='\\t')\ndf.head()","starterSamplePath":"cat12vbm-12.8.2/ds000001/ds000001-000000.tar","totalObjects":1639},"summary":"Preprocessed versions of the OpenMind collection of head and neck 3D magnetic resonance imaging (MRI) pooled from OpenNeuro, produced at NeuroSpin (CEA, France) with the brainprep…","title":"OpenMind Preprocessed Brain MRI: Quasi-Raw T1w, T2w and FLAIR plus CAT12 VBM Grey Matter Maps from 787 OpenNeuro Datasets (CEA NeuroSpin)","updatedAt":"2026-11-21T21:53:43.281Z","visibility":"public","websiteUrl":"https://github.com/neurospin-deepinsight/nidl"}}