[Paper Review] MosMedData: Chest CT Scans With COVID-19 Related Findings Dataset

S. P. Morozov, A. E. Andreychenko+8 more|arXiv (Cornell University)|May 13, 2020

COVID-19 diagnosis using AI35 citations

TL;DR

A dataset of anonymised chest CT scans with COVID-19 findings (and a small subset with binary masks) from Moscow hospitals, released under CC BY-NC-ND 3.0.

ABSTRACT

This dataset contains anonymised human lung computed tomography (CT) scans with COVID-19 related findings, as well as without such findings. A small subset of studies has been annotated with binary pixel masks depicting regions of interests (ground-glass opacifications and consolidations). CT scans were obtained between 1st of March, 2020 and 25th of April, 2020, and provided by municipal hospitals in Moscow, Russia. Permanent link: https://mosmed.ai/datasets/covid19_1110. This dataset is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported (CC BY-NC-ND 3.0) License. Key words: artificial intelligence, COVID-19, machine learning, dataset, CT, chest, imaging

Motivation & Objective

• Provide a publicly available collection of chest CT scans with and without COVID-19 related findings.
• Include a subset of studies annotated with pixel-level masks for ground-glass opacifications and consolidations.
• Offer data suitable for AI and machine learning research in COVID-19 imaging.
• Document data provenance, licensing, and collection period for reproducibility and reuse.

Proposed method

• Collect anonymised chest CT scans from municipal hospitals in Moscow.
• Label a small subset with binary pixel masks for ground-glass opacifications and consolidations.
• Provide scans with and without COVID-19 related findings to support supervised and unsupervised learning.
• Specify acquisition period (March 1, 2020 to April 25, 2020) and licensing terms.
• Ensure data availability under CC BY-NC-ND 3.0 Unported license.

Experimental results

Research questions

• RQ1 What is the composition and distribution of chest CT scans with and without COVID-19 related findings in the MosMedData collection?
• RQ2 How many studies are annotated with pixel-level masks for ROIs, and what regions do these masks cover?
• RQ3 How can the dataset support machine learning model development for COVID-19 chest imaging tasks (e.g., detection, segmentation) while respecting licensing constraints?

Key findings

• Dataset comprises anonymised chest CT scans with COVID-19 related findings and without such findings.
• A small subset of studies is annotated with binary pixel masks depicting ROIs (ground-glass opacifications and consolidations).
• Data were collected between March 1 and April 25, 2020, from Moscow municipal hospitals.
• Licensed under Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported (CC BY-NC-ND 3.0).
• Dataset aims to support AI and machine learning research in COVID-19 imaging.