Deep neural network to detect COVID-19: one architecture for both CT Scans and Chest X-rays

Area of Science:

• Artificial Intelligence
• Medical Imaging
• Infectious Disease Diagnostics

Background:

• The exponential spread of COVID-19 necessitates advanced diagnostic tools.
• Current AI tools for COVID-19 detection often rely on single imaging modalities (CT or CXR).
• Integrating multiple data types may enhance the detection of COVID-19 related anomalies.

Purpose of the Study:

• To develop and evaluate a novel AI model capable of processing both CT scans and Chest X-rays.
• To improve the accuracy and reliability of AI-driven COVID-19 detection by leveraging multimodal imaging data.
• To assess the performance of the proposed model against existing deep neural networks (DNNs).

Main Methods:

• Engineered a Convolutional Neural Network (CNN)-tailored Deep Neural Network (DNN).
• The DNN was designed to collectively train and test both CT scans and Chest X-rays.
• Experimental validation was performed to assess the model's diagnostic capabilities.

Main Results:

• The proposed AI model achieved an overall accuracy of 96.28%.
• The model demonstrated a high Area Under the Curve (AUC) of 0.9808.
• A low false negative rate of 0.0208 was recorded, indicating high sensitivity.
• Existing DNNs showed coherent results when integrating both CT scans and CXRs.

Conclusions:

• The multimodal AI approach integrating CT scans and CXRs significantly improves COVID-19 detection accuracy.
• The developed CNN-tailored DNN offers a promising tool for mass triage and diagnosis of COVID-19.
• Multimodal data integration represents a key advancement in AI-powered medical diagnostics for infectious diseases.