AMOS (abdominal multi-organ segmentation)
Multi-center abdominal CT and MRI from patients with a range of diseases, with voxel labels for 15 abdominal organs
Open the source page ↗ See it in the directory
At a glance
- What it is
- Multi-center abdominal CT and MRI from patients with a range of diseases, with voxel labels for 15 abdominal organs
- Area
- Body
- Modality
- CT, MR
- Format
- NIfTI
- Size
- 500 CT and 100 MRI scans (240 train, 120 validation, 240 test); one 24.2 GB zip
- Normals included
- Partial (mixed or unlabeled)
- License
- CC BY 4.0
- Commercial use
- Allowed by the license, with its attribution terms
- Access
- Open download, no account
- Status
- Verified: source page reviewed by Hounsfield Lab
- Added
- 2026-10-26
- Last checked
- 2026-11-11
- Source
- zenodo.org/records/7262581
How to access
Open the Zenodo record (DOI 10.5281/zenodo.7262581) and download amos22.zip and the metadata CSV. No account is needed. The record links separate Zenodo records for the unlabeled CT and MRI scans.
Limitations
Labels are organ anatomy (spleen, kidneys, gallbladder, esophagus, liver, stomach, aorta, IVC, pancreas, adrenals, duodenum, bladder, prostate/uterus), not disease, and scans are not labeled normal or abnormal. NIfTI in the nnU-Net folder layout, not DICOM. Built for the MICCAI 2022 AMOS challenge; confirm which splits carry labels before you benchmark. Organ sizes are long-tailed, so small structures such as the adrenals are hard.
Citation required?
Yes. Cite Ji et al., AMOS: A Large-Scale Abdominal Multi-Organ Benchmark for Versatile Medical Image Segmentation, NeurIPS 2022 (arXiv:2206.08023), and the Zenodo DOI 10.5281/zenodo.7262581. CC BY 4.0 requires attribution.
Related open-source tools
- TotalSegmentator: Segments more than 100 anatomical structures in CT and MR
- NV-Segment-CT: CT segmentation built from MONAI VISTA3D: 117 structures automatically, or 132 promptable by point or class. Research use only, not for clinical care. Needs a CUDA GPU; weights are about 872 MB
- Merlin: 3D vision-language model for CT that uses structured EHR and report supervision
- Pillar-0 (finetune code + HF weights): Open 3D radiology foundation model; finetuning framework, weights on Hugging Face
- nnU-Net: Self-configuring segmentation framework
- TorchIO: Medical image loading, preprocessing and augmentation for PyTorch
- SimpleITK: Simplified interface to ITK for image I/O, filtering, segmentation and registration in Python, R, Java, C# and other languages
More body datasets
- C4KC-KiTS: CT · CC BY 3.0
- Colorectal-Liver-Metastases: CT · CC BY 4.0
- CT Colonography: CT · CC BY 3.0 and CC BY 4.0 (see the file table)
- CT-ORG: CT · CC BY 3.0
- Healthy-Total-Body-CTs: CT · NIH Controlled Data Access Policy (CT images); CC BY 4.0 (organ segmentations and demographics)
- LDCT-and-Projection-data: CT · Chest, liver, and phantom images CC BY 4.0; head images under the NIH Controlled Data Access Policy
Hounsfield Lab links to the source and does not host images. The license and access summary reflects the source page on the last-checked date; terms can change, so read the source before you download. Not legal advice.