STOIC2021: COVID-19 CT Scans with Severity Labels from France

France during the first wave of the pandemic, from March to April 2020, is the geographic and temporal scope of this collection. It contains Computed Tomography (CT) images from 10,735 individuals suspected of SARS-CoV-2 infection. The dataset, from the Radboud University Medical Center, provides binary labels for COVID-19 presence based on RT-PCR tests and severity defined as intubation or death within one month.

Use Cases

• Training image classification models to detect COVID-19 based on CT scan features.
• Developing prognostic models to predict severe outcomes (intubation/death) from imaging data.
• Benchmarking medical AI algorithms as part of the STOIC2021 challenge framework.
• Studying radiological manifestations of COVID-19 during the pandemic's initial phase.

Strengths

• Includes 10,735 patient CT scans, providing a substantial sample size for model training.
• Offers clinically validated binary labels for both disease presence and severe outcome.
• Data is temporally focused on a specific pandemic wave (March-April 2020), reducing confounding temporal factors.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the specific training sample is unknown, which may limit suitability assessment.
• Data may reflect geographic and temporal bias inherent to collection in France during a specific period.

Provenance

Source

Radboud University Medical Center

Collection Method

CT images collected from individuals suspected of SARS-CoV-2 infection during the first pandemic wave.

Time Range

March to April 2020

Freshness

Last update date is unknown; freshness unverified.

Geography

France