{"success":true,"data":{"author":"Santiago Cepeda et al., Rio Hortega University Hospital, Valladolid (converted by MedOtter)","bucketAccessType":"public","bucketName":"MedOtter/RHUH-GBM","bucketPrefix":"","category":"Health & Biomedical","cloudProvider":"huggingface","createdAt":"2026-11-14T14:15:12.128Z","description":"RHUH-GBM is a longitudinal multi-parametric MRI dataset of glioblastoma patients from Rio Hortega University Hospital (Valladolid, Spain), published on The Cancer Imaging Archive. Its distinguishing feature is that each of the 40 patients is imaged at three timepoints, preoperative, early postoperative (within 72 hours) and at recurrence, with expert-validated tumor sub-region segmentations at every timepoint, which covers the post-resection settings where automated tools trained on preoperative data such as BraTS tend to fail. Each study has T1, contrast-enhanced T1, T2, T2-FLAIR and ADC volumes, skull-stripped and co-registered to the SRI24 atlas, giving 600 imaging volumes and 120 segmentations labelled necrosis, peritumoral region and enhancing tumor. This MedOtter repository carries the TCIA NIfTI release unchanged in layout. As measured 2026-11-14 it has 40 patient folders of NIfTI (.nii.gz) files, a Parquet index and subjects_manifest.json, about 3.1 GB in all; open volumes with nibabel, SimpleITK or MONAI. Used for postoperative tumor segmentation, residual tumor and recurrence assessment, and survival modelling. TCIA releases the collection under CC BY 4.0; cite Cepeda et al., Data in Brief 2023.","displayTitle":"RHUH-GBM: longitudinal glioblastoma MRI with pre-, post-operative and recurrence segmentations","doi":"10.7937/4545-c905","id":"1f9455d1-526f-5f7e-86c7-52356ea1880e","license":"CC-BY-4.0","licenseEvidenceUrl":"https://www.cancerimagingarchive.net/collection/rhuh-gbm/","objectCountEstimate":724,"probeStatus":"ok","referenceUrl":"https://doi.org/10.1016/j.dib.2023.109617","region":"us","schemaCard":{"accessHints":[{"language":"python-datasets","snippet":"from datasets import load_dataset\nds = load_dataset(\"MedOtter/RHUH-GBM\")"},{"language":"python-hub","snippet":"from huggingface_hub import snapshot_download\nsnapshot_download(repo_id=\"MedOtter/RHUH-GBM\", repo_type=\"dataset\")"}],"caveats":["HuggingFace metadata: author=MedOtter, license:cc-by-4.0, size_categories:n<1K, format:parquet, downloads=1458"],"format":"mixed","generatedAt":"2026-11-14T14:15:16.009Z","layout":{"example":"data/train-00000-of-00001.parquet","pattern":"{repoId}/{filename}"},"mountPath":"/mnt/archive","sampleHeaders":[{"bytes":"50415231150415900815b4034c1550150012000088043009000000524855482d303030312e0d0000322e0d0000332e0d0000342e0d0000352e0d0000362e0d0000372e0d0000382e0d0000391d0d0431302e0d002e820000312e820000312e820000312e820000312e820000312e820000312e820000312e820000312e820000322e820000322e820000322e820000322e820000322e820000322e820000322e820000322e820000322e820000322e820000332e82000433313d1100332e820000332e820000332e820000332e820000332e820000332e820000332e820038333909000000524855482d30303430150015c60115cc012c15f001151015061506000063f06203000000f00101061f000004412008c2300c044114456118c6711c08822449a228cab22c0cc3344de338cef33c100445512449d2344d144555556559d6755d18866559a669dab66d1cc7755de779def77d20088661288ae2388e24499665699ae6799e15041518151c4c150615001200000c2c0000000001000000020000001500154e15222c15f0011510150615060000272c03000000f00101021f2449926a03001504156015644c1506150012000030bc050000007072656f700c0000006561726c795f706f73746f7013000000666f6c6c6f7775705f726563757272656e63651500154e15222c15f0011510150615060000272c03000000f0","encoding":"hex","key":"data/train-00000-of-00001.parquet","truncatedAt":512}],"sampleObjects":["RHUH-0001/0/RHUH-0001_0_adc.nii.gz","RHUH-0001/0/RHUH-0001_0_flair.nii.gz","RHUH-0001/0/RHUH-0001_0_segmentations.nii.gz","RHUH-0001/0/RHUH-0001_0_t1.nii.gz","RHUH-0001/0/RHUH-0001_0_t1ce.nii.gz"],"sidecars":[{"bytes":"---\nlicense: cc-by-4.0\ntask_categories:\n- image-segmentation\ntags:\n- medical\n- mri\n- mpmri\n- brain\n- glioblastoma\n- glioma\n- brain-tumor\n- tumor-segmentation\n- longitudinal\n- nifti\n- tcia\n- rhuh-gbm\npretty_name: RHUH-GBM (Longitudinal mpMRI + Tumor Segmentation)\nsize_categories:\n- n<1K\ndataset_info:\n  features:\n  - name: patient_id\n    dtype: string\n  - name: timepoint\n    dtype: int32\n  - name: timepoint_name\n    dtype: string\n  - name: shape\n    dtype: string\n  - name: num_slices\n    dtype: int32\n  - name: selected_slice\n    dtype: int32\n  - name: labels_present\n    dtype: string\n  - name: t1ce\n    dtype: image\n  - name: flair\n    dtype: image\n  - name: segmentation\n    dtype: image\n  - name: overlay\n    dtype: image\n  splits:\n  - name: train\n    num_bytes: 6251229\n    num_examples: 120\n  download_size: 6256291\n  dataset_size: 6251229\nconfigs:\n- config_name: default\n  data_files:\n  - split: train\n    path: data/train-*\n---\n\n# RHUH-GBM — Rio Hortega University Hospital Glioblastoma dataset\n\nLongitudinal multi-parametric MRI (mpMRI) of glioblastoma patients from Rio\nHortega University Hospital (Valladolid, Spain), with expert tumor sub-region\nsegmentations at **three timepoints per patient**: preoperative, early\npostoperative (< 72 h), and follow-up at recurrence. This is the **NIfTI**\nrelease from TCIA — images are skull-stripped and co-registered to the SRI24\natlas, and the segmentations are aligned to them.\n\n> Distinguishing feature vs. preop-only glioma datasets (e.g. BraTS): RHUH-GBM\n> provides **post-resection and recurrence** scans with expert-validated masks —\n> exactly the timepoints where automated tools usually fail.\n\n## Dataset Details\n\n| Field | Value |\n|---|---|\n| Modality | Brain mpMRI — T1, T1-Gd (T1CE), T2, T2-FLAIR, ADC |\n| Body part | Brain (glioblastoma, WHO grade 4) |\n| Task | 3D multi-class tumor sub-region segmentation |\n| Patients | 40 |\n| Timepoints / patient | 3 (preop / early postop / recurrence) |\n| Studies (timepoints) | 120 |\n| Imaging volumes | 600 (5 sequences x 120 studies) |\n| Segmentations | 120 (one per study) |\n| Volume geometry | 240 x 240 x 155 (one study, RHUH-0028/0, is 230 x 230 x 138) |\n| Format | NIfTI (`.nii.gz`) |\n| License | CC BY 4.0 |\n\n## Label Scheme\n\n| Value | Tumor sub-region |\n|---|---|\n| 0 | Background |\n| 1 | Necrosis (necrotic tumor core) |\n| 2 | Peritumoral region (edema / non-enhancing signal alteration) |\n| 3 | Enhancing tumor |\n\nEvaluation regions (BraTS-style): **WT** (whole tumor) = 1+2+3,\n**TC** (tumor core) = 1+3, **ET** (enhancing tumor) = 3.\n\n> **Label encoding note (verified against the released masks).** The upstream\n> preprocessing pipeline documents enhancing tumor as BraTS label **4**, but the\n> distributed NIfTI masks use **3** — value 4 never appears in any of the 120\n> masks. Loaders should treat **3** as enhancing tumor for RHUH-GBM.\n\n## Ground Truth\n\nA **single, expert-corrected segmentation tier** (no separate automated tier).\nMasks were initialized with a DeepMedic CNN and then **reviewed and manually\ncorrected by two neurosurgeons** specializing in neuroimaging. Every study has\nexactly one segmentation; no GT-tier filtering is required.\n\n## Structure\n\n```\nRHUH-NNNN/<tp>/RHUH-NNNN_<tp>_t1.nii.gz\nRHUH-NNNN/<tp>/RHUH-NNNN_<tp>_t1ce.nii.gz\nRHUH-NNNN/<tp>/RHUH-NNNN_<tp>_t2.nii.gz\nRHUH-NNNN/<tp>/RHUH-NNNN_<tp>_flair.nii.gz\nRHUH-NNNN/<tp>/RHUH-NNNN_<tp>_adc.nii.gz\nRHUH-NNNN/<tp>/RHUH-NNNN_<tp>_segmentations.nii.gz   # expert GT\nsubjects_manifest.json                               # per-study paths + legends\n```\n\n`<tp>` is the timepoint index: `0` = preoperative, `1` = early postoperative\n(< 72 h), `2` = follow-up / recurrence. `subjects_manifest.json` lists, for every\nstudy, the five modality paths and the segmentation path, plus the label and\ntimepoint legends — so loaders need not re-derive them.\n\n## Cohort Overlap\n\n**No known overlap with BraTS2023 or UCSF-PDGM.** RHUH-GBM is a single-institution\nSpanish cohort (Rio Hortega U. Hospital, 2018-2022) and is not among the\ncontributing sites of either dataset; the released data carries no B","encoding":"utf8","key":"README.md","name":"README.md","size":6478,"truncated":true}],"sourceObjectKey":"data/train-00000-of-00001.parquet","starterCell":"from datasets import load_dataset\nds = load_dataset(\"MedOtter/RHUH-GBM\")\nprint(ds)","starterSamplePath":"RHUH-0001/0/RHUH-0001_0_adc.nii.gz","totalObjects":724},"summary":"RHUH-GBM is a longitudinal multi-parametric MRI dataset of glioblastoma patients from Rio Hortega University Hospital (Valladolid, Spain), published on The Cancer Imaging Archive.","title":"RHUH-GBM: longitudinal glioblastoma MRI with pre-, post-operative and recurrence segmentations","updatedAt":"2026-11-14T17:51:26.324Z","visibility":"public","websiteUrl":"https://www.cancerimagingarchive.net/collection/rhuh-gbm/"}}