An Open-Source COVID-19 CT Dataset with Automatic Lung Tissue Classification for Radiomics

Area of Science:

• Medical Imaging
• Computerized Tomography
• Radiology

Background:

• The COVID-19 pandemic significantly impacts healthcare systems.
• Lung CT scans are crucial for COVID-19 prognosis.
• Existing datasets lack features for quantitative analysis.

Purpose of the Study:

• To present a novel, open-source lung CT dataset for COVID-19 research.
• To facilitate the development of quantitative analysis tools for medical imaging.
• To support clinicians in managing the COVID-19 pandemic.

Main Methods:

• Compiled a dataset of lung CT scans from 50 COVID-19 positive patients.
• Utilized Gaussian Mixture Model (GMM) for automatic threshold-based annotation.
• Included expert radiologist scoring correlated with GMM findings.

Main Results:

• The dataset includes CT volumes with GMM annotations and radiologist scores.
• Radiologist scores significantly correlated with ground glass opacities and consolidation identified by GMM.
• Open-source dataset and GMM fitting code are publicly available.

Conclusions:

• The released dataset offers a unique resource for medical image analysis researchers.
• This initiative aims to foster the development of algorithms supporting COVID-19 clinical management.
• The dataset and code promote advancements in quantitative lung CT analysis for infectious diseases.