TotalSegmentator CT dataset: 1,228 CT scans with segmentations of 117 anatomical structures

Provider
Jakob Wasserthal et al., University Hospital Basel (redistributed by Yongcheng Yao)
License
CC-BY-4.0 (source)
Size
5 files
Format
CSV (in sampled files)
Stored at
Hugging Face (https://huggingface.co/datasets/YongchengYAO/TotalSegmentator-CT-Lite)
Access
Free and public. No account with the data provider is needed.
DOI
10.5281/zenodo.10047292
Category
Health & Biomedical
Source
https://doi.org/10.1148/ryai.230024

About this dataset

The TotalSegmentator dataset is the training data behind the TotalSegmentator nnU-Net model from University Hospital Basel: 1,228 routine clinical CT examinations with dense voxel-level segmentation masks for 117 anatomical structures, covering abdominal organs (spleen, kidneys, liver, pancreas, stomach, bowel), the lung lobes, every vertebra from C1 to S1, ribs, the heart and great vessels, and major muscles. This repository is a derivative packaging by Yongcheng Yao for the MedVision benchmark: the per-structure masks are merged into one label map per case, CT volumes and masks are renamed by case ID, and an official train/test split is given in meta.csv. As measured 2026-11-14 the repository holds two archives, Images.zip (about 22.6 GB) and Masks.zip (about 0.8 GB), plus meta.csv; unzip them and read the NIfTI (.nii.gz) volumes with nibabel or SimpleITK. Typical uses are multi-organ CT segmentation, anatomy-aware foundation model pretraining, body composition and opportunistic screening research. The original data are released by the authors on Zenodo under CC BY 4.0; cite Wasserthal et al., Radiology: Artificial Intelligence 2023.

How to load this dataset in Python

import pandas as pd
df = pd.read_csv("hf://datasets/YongchengYAO/TotalSegmentator-CT-Lite/meta.csv")
df.head()

Data structure

File types in sampled files: CSV

Files

Example file: meta.csv

Analyze this dataset without downloading it

On Scigantic, TotalSegmentator CT dataset: 1,228 CT scans with segmentations of 117 anatomical structures 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