CODENET: A deep learning model for COVID-19 detection.

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Healthcare
• Deep Learning for Diagnostics

Background:

• Conventional COVID-19 testing is costly and slow.
• Chest X-ray (CXR) analysis offers a potential alternative but lacks automated, interpretable solutions.
• Artificial intelligence (AI) can enhance medical radiography for disease detection and reduce healthcare burdens.

Purpose of the Study:

• To develop an accurate, interpretable, and automated diagnostic framework for COVID-19 using CXR images.
• To leverage deep neural networks (DNNs) and contrastive learning for improved feature extraction and generalization.
• To establish a practical AI-driven tool for COVID-19 detection via medical radiography.

Main Methods:

• Development of a novel Convolutional Neural Network (CNN) named CodeNet for COVID-19 diagnosis.
• Application of contrastive learning to maximize the utility of latent image data for feature enhancement.
• Utilizing CXR images as the primary input for the AI model.

Main Results:

• The proposed CodeNet method achieved a high accuracy of 94.20% on the evaluation dataset.
• CodeNet outperformed several existing comparative methods in COVID-19 detection.
• Ablation studies confirmed the model's efficacy, and interpretability analysis demonstrated its clinical utility.

Conclusions:

• The CNN-based CodeNet model with contrastive learning shows superior performance in detecting COVID-19 from CXR images.
• This AI approach offers a practical and interpretable solution, potentially reducing healthcare burdens.
• The study highlights the potential of AI and computer vision in diagnosing diseases using medical imaging data.