A large-scale collection of publicly available 3D medical imaging datasets aggregated for training universal tumor segmentation models. Total size: ~700 GB, 83 000+ annotated files across 13+ datasets and 10+ organ sites.

This dataset card documents the structure and contents. The actual imaging data is being uploaded progressively; in the meantime, individual datasets can be downloaded from the original sources listed below.

Dataset Summary

Dataset	Files	Size	Format	Modality	Anatomy / Task
BraTS 2023	13 605	41 G	NIfTI	MRI (T1, T1ce, T2, FLAIR)	Brain tumor (GLI, MEN, MET, PED, SSA)
MSD Task01 BrainTumour	1 269	7 G	NIfTI	MRI	Brain tumor
MSD Task03 Liver	664	27 G	NIfTI	CT	Liver + liver tumor
MSD Task05 Prostate	86	229 M	NIfTI	MRI	Prostate (peripheral + transition zone)
MSD Task06 Lung	316	9 G	NIfTI	CT	Lung tumor
MSD Task07 Pancreas	711	12 G	NIfTI	CT	Pancreas + pancreatic tumor
MSD Task08 HepaticVessel	1 492	9 G	NIfTI	CT	Hepatic vessels + tumors
AbdomenCT-1K	2 062	79 G	NIfTI	CT	Multi-organ abdominal (1 000 cases)
LUNA16	1 699	102 G	MHD/RAW	CT	Lung nodule screening (10 subsets)
LNDb	774	92 G	MHD/RAW	CT	Lung nodule database
COVID-19 CT Seg	80	1 G	NIfTI	CT	COVID-19 lung infection (20 cases)
LGG MRI	15 716	2 G	TIFF	MRI	Low-grade glioma (2D slices)
ULS23 (Parts 1--6)	~45 000	322 G	NIfTI	CT/MRI	Universal lesion segmentation

Upload Progress

Dataset	Status
COVID-19 CT Seg	Uploading (test sample available)
BraTS 2023	Pending
MSD Tasks 01--08	Pending
AbdomenCT-1K	Pending
LUNA16	Pending
LNDb	Pending
LGG MRI	Pending
ULS23 Parts 1--6	Pending

ULS23 Sub-datasets

Sub-dataset	Annotation Type	Files
KiTS21	Fully annotated	333
LIDC-IDRI	Fully annotated	2 246
LiTS	Fully annotated	888
MDSC Task06 Lung	Fully annotated	76
MDSC Task07 Pancreas	Fully annotated	283
MDSC Task10 Colon	Fully annotated	133
NIH LN ABD	Fully annotated	558
NIH LN MED	Fully annotated	379
DeepLesion3D	Novel data	750
Radboudumc Bone	Novel data	744
Radboudumc Pancreas	Novel data	124
CCC18	Partially annotated	1 211
DeepLesion	Partially annotated	25 634

Directory Structure

OmniTumorData/
│
├── abdomenct_1k/
│   ├── AbdomenCT-1K-ImagePart1/          # CT volumes (*.nii.gz)
│   ├── AbdomenCT-1K-ImagePart2/
│   ├── AbdomenCT-1K-ImagePart3/
│   └── Mask/                              # Segmentation masks (*.nii.gz)
│
├── brats/
│   ├── BraTS-GLI/
│   │   ├── ASNR-MICCAI-BraTS2023-GLI-Challenge-TrainingData/
│   │   │   └── BraTS-GLI-XXXXX-XXX/      # Per-case directories
│   │   │       ├── BraTS-GLI-*-t1c.nii.gz
│   │   │       ├── BraTS-GLI-*-t1n.nii.gz
│   │   │       ├── BraTS-GLI-*-t2f.nii.gz
│   │   │       ├── BraTS-GLI-*-t2w.nii.gz
│   │   │       └── BraTS-GLI-*-seg.nii.gz
│   │   └── ASNR-MICCAI-BraTS2023-GLI-Challenge-ValidationData/
│   ├── BraTS-MEN/
│   ├── BraTS-MET/
│   ├── BraTS-PED/
│   └── BraTS-SSA/
│
├── covid19_ct/
│   ├── COVID-19-CT-Seg_20cases/           # CT volumes
│   ├── Infection_Mask/                    # Infection segmentation
│   ├── Lung_Mask/                         # Lung segmentation
│   └── Lung_and_Infection_Mask/           # Combined masks
│
├── LGGMRI/
│   └── kaggle_3m/
│       └── TCGA_XX_XXXX_XXXXXXXX/        # Per-patient directories
│           ├── TCGA_*_NNN.tif             # MRI slices
│           └── TCGA_*_NNN_mask.tif        # Segmentation masks
│
├── lndb/
│   ├── LNDb-NNNN.mhd + .raw              # CT volumes
│   ├── masks/                             # Nodule masks (*.nii.gz)
│   └── predictedNodulesB/                 # Predicted nodules
│
├── luna16/
│   ├── subset0/ ... subset9/              # CT volumes (*.mhd + *.raw)
│   ├── seg-lungs-LUNA16/                  # Lung segmentation masks
│   └── evaluationScript/                  # Evaluation tools + annotations
│
├── MSD_Task01_BrainTumour/
│   ├── imagesTr/                          # Training images (*.nii.gz)
│   ├── labelsTr/                          # Training labels (*.nii.gz)
│   └── imagesTs/                          # Test images
│
├── MSD_Task03_Liver/                      # Same structure as Task01
├── MSD_Task05_Prostate/
├── MSD_Task06_Lung/
├── MSD_Task07_Pancreas/
├── MSD_Task08_HepaticVessel/
│
├── uls23_part1/
│   └── ULS23/
│       ├── novel_data/
│       │   ├── ULS23_DeepLesion3D/
│       │   ├── ULS23_Radboudumc_Bone/
│       │   └── ULS23_Radboudumc_Pancreas/
│       └── processed_data/
│           └── fully_annotated/
│               └── KiTS21/
│                   ├── images/            # CT volumes (*.nii.gz)
│                   └── labels/            # Segmentation masks
│
├── uls23_part2/
│   └── ULS23/processed_data/fully_annotated/
│       ├── LIDC-IDRI/
│       └── LiTS/
│
├── uls23_part3/
│   └── ULS23/processed_data/fully_annotated/
│       ├── MDSC_Task06_Lung/
│       ├── MDSC_Task07_Pancreas/
│       ├── MDSC_Task10_Colon/
│       ├── NIH_LN_ABD/
│       └── NIH_LN_MED/
│
├── uls23_part4/
│   └── ULS23/processed_data/...
│
├── uls23_part5/
│   └── ULS23/processed_data/...
│
├── uls23_part6/
│   └── ULS23/processed_data/partially_annotated/
│       ├── CCC18/
│       └── DeepLesion/
│           └── images/                    # 25 634 CT crops (*.nii.gz)
│
└── uls23_labels/
    └── annotations/ULS23/
        ├── novel_data/*/labels/
        ├── processed_data/fully_annotated/*/labels/
        └── processed_data/partially_annotated/*/labels/

Data Sources

Dataset	Source	License
BraTS 2023	Synapse	CC-BY-SA 4.0
MSD (Tasks 1--8)	Medical Decathlon	CC-BY-SA 4.0
AbdomenCT-1K	GitHub	CC-BY 4.0
LUNA16	Zenodo	CC-BY 4.0
LNDb	Zenodo	CC-BY-NC-SA 4.0
COVID-19 CT Seg	Zenodo	CC-BY-NC-SA 4.0
LGG MRI	Kaggle	CC-BY-NC-SA 4.0
ULS23	Zenodo	CC-BY 4.0

Usage

This dataset is designed for training the OmniTumor 3D adapter architecture for universal tumor segmentation.

import nibabel as nib

# Load a volume
vol = nib.load("OmniTumorData/MSD_Task03_Liver/imagesTr/liver_0.nii.gz")
data = vol.get_fdata()  # shape: (H, W, D)

# Load corresponding label
seg = nib.load("OmniTumorData/MSD_Task03_Liver/labelsTr/liver_0.nii.gz")
mask = seg.get_fdata()  # 0=background, 1=liver, 2=tumor

License

Each constituent dataset retains its original license (see table above). This aggregation is provided for research purposes only. Not intended for clinical use.

Downloads last month: -