Automatic distinction between COVID-19 and common pneumonia using multi-scale convolutional neural network on chest...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Infectious Diseases

Background:

• COVID-19 pneumonia emerged as a global health threat in late 2019.
• Rapid and accurate diagnosis is crucial for managing the pandemic.
• Chest computed tomography (CT) is a key imaging modality for pneumonia detection.

Purpose of the Study:

• To develop and evaluate an artificial intelligence (AI) system for rapid COVID-19 pneumonia diagnosis.
• To assist radiologists and clinicians in differentiating COVID-19 from other pneumonias.
• To mitigate the workload on healthcare professionals during the pandemic.

Main Methods:

• Retrospective collection of 206 COVID-19 positive RT-PCR patients with 416 chest CT scans.
• Inclusion of 412 non-COVID-19 pneumonia patients with 412 chest CT scans.
• Design and evaluation of an AI system utilizing a multi-scale convolutional neural network (MSCNN) at slice and scan levels.

Main Results:

• The AI system demonstrated promising diagnostic performance in detecting COVID-19.
• The system effectively differentiated COVID-19 from other common pneumonias.
• Satisfactory performance was achieved even with limited training data.

Conclusions:

• The developed AI system shows significant potential for assisting in the quick diagnosis of COVID-19 pneumonia.
• The AI tool can help alleviate the burden on healthcare systems during health crises.
• Publicly available data facilitates further research and development in AI-driven medical diagnostics.