Deep learning for diagnosis of COVID-19 using 3D CT scans

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Infectious Diseases

Background:

• COVID-19, a novel coronavirus pneumonia, has caused a global health crisis.
• Computed Tomography (CT) scans are crucial for diagnosing COVID-19 pneumonia.
• Manual analysis of extensive CT scan data can cause diagnostic delays.

Purpose of the Study:

• To develop and evaluate an AI approach for classifying COVID-19 from normal CT volumes.
• To leverage deep learning for rapid and accurate detection of COVID-19 pneumonia.

Main Methods:

• Utilized the ResNet-50 deep learning model for image-level COVID-19 prediction on individual CT slices.
• Implemented a fusion strategy to aggregate image-level predictions for 3D CT volume diagnosis.
• Evaluated the AI model's performance in classifying COVID-19 versus normal CT scans.

Main Results:

• The proposed AI method demonstrated high efficacy in detecting COVID-19 on CT scans.
• Achieved an Area Under the Curve (AUC) value of 96% for COVID-19 detection.
• Indicated the potential for AI to significantly reduce diagnostic time.

Conclusions:

• AI, specifically deep learning with ResNet-50, can accurately diagnose COVID-19 pneumonia from CT scans.
• This AI approach offers a promising solution to overcome diagnostic bottlenecks in hospitals.
• The findings support the integration of AI tools in radiological workflows for infectious disease management.