COVID-19-20: chest CT with radiologist-corrected COVID-19 lung lesion masks, 199 training volumes (MICCAI 2020 challenge)

Provider
Holger R. Roth, Ziyue Xu et al. (NIH, NVIDIA, Children's National); mirrored by MedOtter
License
CC-BY-4.0 (source)
Size
402 files
Format
NIfTI, Parquet (in sampled files)
Stored at
Hugging Face (https://huggingface.co/datasets/MedOtter/COVID-19-20)
Access
Free and public. No account with the data provider is needed.
Category
Health & Biomedical
Source
https://doi.org/10.1016/j.media.2022.102605

About this dataset

Training data of the MICCAI 2020 COVID-19 Lung CT Lesion Segmentation Challenge (COVID-19-20): non-contrast chest CT of patients with SARS-CoV-2 pneumonia and binary masks of the lung lesions (ground-glass opacities and consolidation). There are 199 CT volumes from the multinational NIH consortium, stored as int16 Hounsfield-unit NIfTI (volume-covid19-A-XXXX_ct.nii.gz) with an aligned mask (_seg.nii.gz) under Train/; masks were initialized by an NVIDIA and NIH deep learning model and corrected by board-certified radiologists in ITK-SNAP. The official train and validation file lists come as an .xlsx, and a preview Parquet shows one slice per case. About 3.9 GB (measured 2026-11-21); read with nibabel or SimpleITK. The 50 validation and 46 test cases had no public masks and are not included. Use for COVID-19 pneumonia lesion segmentation and lung infection quantification. Images are from the TCIA collection CT Images in COVID-19 (doi 10.7937/TCIA.2020.GQRY-NC81, CC BY 4.0); the challenge data page states the annotations are CC0. Cite Roth et al., Medical Image Analysis 82:102605 (2022).

How to load this dataset in Python

from datasets import load_dataset
ds = load_dataset("MedOtter/COVID-19-20")
print(ds)

Data structure

File types in sampled files: NIfTI, Parquet

Files

Example file: data/train-00000-of-00001.parquet

Analyze this dataset without downloading it

On Scigantic, COVID-19-20: chest CT with radiologist-corrected COVID-19 lung lesion masks, 199 training volumes (MICCAI 2020 challenge) 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