{"success":true,"data":{"author":"Michal Nohel, Stefan Reguli, Romana Kaplanova et al. (University Hospital Ostrava); mirrored by MedOtter","bucketAccessType":"public","bucketName":"MedOtter/BEAMSTER","bucketPrefix":"","category":"Health & Biomedical","cloudProvider":"huggingface","createdAt":"2026-11-21T20:45:56.464Z","description":"BEAMSTER (Brain mEtAstases segMentation for STEreotactic Radiotherapy) is a retrospective MRI dataset from University Hospital Ostrava, Czech Republic (October 2019 to September 2024) of 140 patients treated with stereotactic radiotherapy for brain metastases. Each patient has one contrast-enhanced T1-weighted 3D MPRAGE volume and a binary metastasis mask drawn by a board-certified radiation oncologist and checked by a neuroradiologist. Dataset_A has 113 cases with 216 lesions (mean 6.7 cc) and Dataset_B 27 cases with 44 lesions (mean 0.5 cc), deliberately enriched for very small metastases; the two cohorts overlap in time and should not be used as a train/test split. Files: dataset/Dataset_A and Dataset_B NIfTI volumes and _segm masks, Spreadsheets/ with clinical data and voxel parameters, the deposit's README_original.txt and train.jsonl; about 4.4 GB (measured 2026-11-21). Read with nibabel or SimpleITK. Use for brain metastasis detection and segmentation, small-lesion sensitivity testing and radiosurgery planning. Mirrored by MedOtter from the authors' figshare deposit, CC BY 4.0, DOI 10.6084/m9.figshare.29365844.","displayTitle":"BEAMSTER: contrast-enhanced T1 MRI of 140 brain metastasis patients with stereotactic radiotherapy segmentations","doi":"10.6084/m9.figshare.29365844","huggingFaceUrl":"https://huggingface.co/datasets/MedOtter/BEAMSTER","id":"b577568a-e440-5960-a5a9-e6cd3dcf7d16","license":"CC-BY-4.0","licenseEvidenceUrl":"https://figshare.com/articles/dataset/29365844","objectCountEstimate":289,"probeStatus":"ok","region":"us","schemaCard":{"accessHints":[{"language":"python-datasets","snippet":"from datasets import load_dataset\nds = load_dataset(\"MedOtter/BEAMSTER\")"},{"language":"python-hub","snippet":"from huggingface_hub import snapshot_download\nsnapshot_download(repo_id=\"MedOtter/BEAMSTER\", repo_type=\"dataset\")"}],"caveats":["HuggingFace metadata: author=MedOtter, license:cc-by-4.0, size_categories:n<1K, format:parquet, downloads=231"],"format":"mixed","generatedAt":"2026-11-21T20:48:43.093Z","layout":{"example":"Spreadsheets/Table_clinical_data.csv","pattern":"{repoId}/{filename}"},"mountPath":"/mnt/archive","sampleHeaders":[{"bytes":"Patient ID,Sex,Age (years),MRI Scanner Type,Origin,Radiotherapy Dose (Gy),Number of Fractions,Isodose (%) / Prescription Isodose,\"Vital Status (0=Alive, 1=Deceased)\",Survival time (days from MRI),\"Prior surgery (0=No, 1=Yes)\",\"Treatment status (0=Relapse, 1=Primary)\"\r\nDataset_A_001,F,71,Avanto,Other,20 Gy,1,58,1.0,137.0,0,1\r\nDataset_A_002,M,68,Avanto,Melanoma,27 Gy,3,50,1.0,219.0,0,1\r\nDataset_A_003,M,64,Avanto,Gastrointestinal tract,27 Gy,3,53,1.0,364.0,0,1\r\nDataset_A_004,F,56,Avanto,Ca Mammae,27 Gy,3,50,1.","encoding":"utf8","key":"Spreadsheets/Table_clinical_data.csv","truncatedAt":512}],"sampleObjects":["README_original.txt","Spreadsheets/Table_clinical_data.csv","Spreadsheets/Table_clinical_data.xlsx","Spreadsheets/image_voxel_parameters.csv","Spreadsheets/image_voxel_parameters.xlsx"],"sidecars":[{"bytes":"---\nlicense: cc-by-4.0\ntask_categories:\n- image-segmentation\ntags:\n- medical\n- brain\n- brain-metastases\n- mri\n- 3d\n- radiotherapy\n- stereotactic-radiosurgery\npretty_name: BEAMSTER (brain metastases for stereotactic radiotherapy)\nconfigs:\n- config_name: preview\n  data_files:\n  - split: train\n    path: preview/train-*\ndataset_info:\n  config_name: preview\n  features:\n  - name: case_id\n    dtype: string\n  - name: subset\n    dtype: string\n  - name: sex\n    dtype: string\n  - name: age\n    dtype: int32\n  - name: scanner\n    dtype: string\n  - name: primary_origin\n    dtype: string\n  - name: image_file\n    dtype: string\n  - name: mask_file\n    dtype: string\n  - name: num_slices\n    dtype: int32\n  - name: preview_slice\n    dtype: int32\n  - name: lesion_volume_cc\n    dtype: float32\n  - name: n_components\n    dtype: int32\n  - name: spacing_mm\n    list: float32\n  - name: image\n    dtype: image\n  - name: mask\n    dtype: image\n  - name: overlay\n    dtype: image\n  - name: overlay_zoom\n    dtype: image\n  splits:\n  - name: train\n    num_bytes: 29643563\n    num_examples: 140\n  download_size: 29639302\n  dataset_size: 29643563\n---\n\n# BEAMSTER\n\n**B**rain m**E**t**A**stases seg**M**entation for **STE**reotactic **R**adiotherapy — a\nretrospective MRI dataset with expert segmentations. Re-host of the author deposit\n[figshare 10.6084/m9.figshare.29365844](https://doi.org/10.6084/m9.figshare.29365844) v1\n(**CC BY 4.0**), from University Hospital Ostrava, Czech Republic, Oct 2019 - Sep 2024.\n\n140 patients, one contrast-enhanced T1w 3D MPRAGE volume each, with a binary\nbrain-metastasis mask drawn by a board-certified radiation oncologist (13 years'\nexperience) for stereotactic radiotherapy planning and independently verified by a\nboard-certified neuroradiologist.\n\n## Cohorts\n\n| subset | cases | lesions | mean lesion vol | note |\n|---|---|---|---|---|\n| `Dataset_A` | 113 | 216 | 6.7 cc | treated Oct 2019 - Apr 2022, 1/3/5 fractions |\n| `Dataset_B` | 27 | 44 | 0.5 cc | treated Jul 2021 - Sep 2024, **deliberately enriched with very small lesions** |\n\nThere is **no official train/val/test split** — `train.jsonl` carries all 140 rows with\n`split: \"train\"`. `subset` is a cohort label, not a split.\n\n> **Do not use A vs B as a train/test partition.** Their accrual windows overlap by\n> 9 months at the same institution, and the paper never states the cohorts are\n> patient-disjoint. The de-identified IDs make this unverifiable.\n\n## Layout\n\n```\ndataset/Dataset_A/Dataset_A_YYY.nii.gz        ce-T1w volume\ndataset/Dataset_A/Dataset_A_YYY_segm.nii.gz   binary metastasis mask\ndataset/Dataset_B/...                          same convention\nSpreadsheets/Table_clinical_data.{csv,xlsx}    140 rows of clinical data\nSpreadsheets/image_voxel_parameters.{csv,xlsx} per-case dimensions + voxel size\nREADME_original.txt                            the deposit's own README, verbatim\ntrain.jsonl                                    one row per case (see columns below)\n```\n\n`train.jsonl` columns: `case_id`, `subset`, `image`, `mask`, `shape`, `spacing_mm`,\n`orientation`, `lesion_volume_cc`, `n_components`, `fg_voxels`, `split`, plus the\nclinical fields `sex`, `age`, `scanner`, `primary_origin`, `rt_dose_gy`, `n_fractions`,\n`isodose_pct`, `vital_status`, `survival_days`, `prior_surgery`, `treatment_status`.\n`vital_status` (1 case) and `survival_days` (15 cases) are nullable.\n\n## Verified properties\n\n- Image and mask share an **identical shape and affine in all 140 cases** — no\n  resampling is needed to pair them.\n- Masks are strictly `{0, 1}` but stored as **float32** — cast to uint8 in a loader.\n- All volumes are 512x512 in-plane, z = 201-506. **72 distinct voxel spacings**;\n  slice thickness is 1.0 mm (137 cases), 1.25 mm (2), 0.625 mm (1).\n- Orientation is **LPS for 139 cases and LAS for `Dataset_B_001`**. Image and mask\n  agree within every case, so this only matters to code that hardcodes axis order.\n- No empty masks. ID sequences have gaps: `Dataset_A` spans 001-123 (10 missing),\n  `Dataset_B` spans 001-030 (3 missing).\n\n## Caveats that affect evaluation\n\n1. **Masks are a","encoding":"utf8","key":"README.md","name":"README.md","size":6333,"truncated":true}],"sourceObjectKey":"Spreadsheets/Table_clinical_data.csv","starterCell":"import pandas as pd\ndf = pd.read_csv(\"hf://datasets/MedOtter/BEAMSTER/Spreadsheets/Table_clinical_data.csv\")\ndf.head()","starterSamplePath":"Spreadsheets/Table_clinical_data.csv","totalObjects":289},"summary":"BEAMSTER (Brain mEtAstases segMentation for STEreotactic Radiotherapy) is a retrospective MRI dataset from University Hospital Ostrava, Czech Republic (October 2019 to September…","title":"BEAMSTER: contrast-enhanced T1 MRI of 140 brain metastasis patients with stereotactic radiotherapy segmentations","updatedAt":"2026-11-21T22:19:48.835Z","visibility":"public","websiteUrl":"https://doi.org/10.6084/m9.figshare.29365844"}}