AMOS 2022: abdominal multi-organ segmentation benchmark (300 CT + 60 MRI)

Provider
Yuanfeng Ji et al., AMOS 2022 challenge (redistributed by Yongcheng Yao)
License
CC-BY-4.0 (source)
Size
5 files
Stored at
Hugging Face (https://huggingface.co/datasets/YongchengYAO/AMOS22-Lite)
Access
Free and public. No account with the data provider is needed.
DOI
10.5281/zenodo.7262581
Category
Health & Biomedical
Source
https://arxiv.org/abs/2206.08023

About AMOS 2022: abdominal multi-organ segmentation benchmark (300 CT + 60 MRI)

AMOS (A large-scale abdominal multi-organ benchmark for versatile medical image segmentation) is the MICCAI 2022 challenge dataset of abdominal CT and MRI collected across multiple centres, vendors and disease cohorts, annotated voxel-wise for 15 organs: spleen, both kidneys, gallbladder, esophagus, liver, stomach, aorta, inferior vena cava, pancreas, both adrenal glands, duodenum, bladder and prostate/uterus. This repository, prepared by Yongcheng Yao for the MedVision benchmark, contains every labelled case of the public train and validation splits, 300 CT and 60 MRI volumes with masks, in AMOS22-CT and AMOS22-MRI folders; the unlabelled test images are left out. Image voxel data are unmodified apart from reorientation to RAS+. As measured 2026-11-14 the data ship as three zip parts totalling about 22.7 GB; unzip and read the NIfTI (.nii.gz) files with nibabel, SimpleITK or MONAI. Used for training and benchmarking CT and MRI organ segmentation, cross-modality generalisation and medical foundation models. The AMOS organisers release the data on Zenodo under CC BY 4.0; cite Ji et al., NeurIPS 2022 Datasets and Benchmarks.

How to load this dataset in Python

from huggingface_hub import hf_hub_download
path = hf_hub_download(repo_id="YongchengYAO/AMOS22-Lite", filename="data-part001.zip", repo_type="dataset")
print(path)

Data structure

Files

Example file: data-part001.zip

Analyze this dataset without downloading it

On Scigantic, AMOS 2022: abdominal multi-organ segmentation benchmark (300 CT + 60 MRI) opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.

This page is also available as Markdown and as JSON.

Related Health & Biomedical datasets