Contribution to pulmonary diseases diagnostic from X-ray images using innovative deep learning models

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Pulmonology

Background:

• Accurate pulmonary disease diagnosis is crucial but challenging, with radiographic interpretation posing difficulties.
• Artificial intelligence (AI), particularly machine learning (ML) and deep learning (DL), shows promise in medical image analysis.
• Existing methods for lung disorder identification using radiography require enhancement for accuracy and efficiency.

Purpose of the Study:

• To develop and evaluate deep learning models for identifying specific pulmonary diseases from thoracic radiography.
• To create AI systems capable of distinguishing between normal and diseased lung conditions, including COVID-19, pneumonia, and pulmonary opacity.
• To compare the performance of novel deep learning models against current state-of-the-art methods in lung disease detection.

Main Methods:

• Proposed three distinct deep learning models: CovCXR-Net, MDCXR3-Net, and MDCXR4-Net.
• Trained models on thoracic radiography datasets to classify various lung conditions.
• Evaluated model performance using established benchmarks and accuracy metrics.

Main Results:

• CovCXR-Net achieved 99.09% accuracy in identifying COVID-19 or normal cases.
• MDCXR3-Net reached 97.74% accuracy for differentiating COVID-19, pneumonia, or normal.
• MDCXR4-Net attained 90.37% accuracy in classifying COVID-19, pneumonia, pulmonary opacity, or normal conditions.
• All proposed models demonstrated superior performance compared to existing state-of-the-art approaches.

Conclusions:

• Deep learning models offer a highly accurate and effective approach to diagnosing pulmonary diseases from chest X-rays.
• The developed AI systems, CovCXR-Net, MDCXR3-Net, and MDCXR4-Net, show significant potential in clinical settings for rapid and reliable disease identification.
• These findings highlight the transformative impact of AI in improving the accuracy and efficiency of lung disease diagnosis.