COV-SNET: A deep learning model for X-ray-based COVID-19 classification.

Area of Science:

• Artificial Intelligence
• Medical Imaging
• Pulmonary Diseases

Background:

• Diagnosing COVID-19 from X-rays is challenging due to similar imaging characteristics with other pneumonias.
• Limited availability of public COVID-19 X-ray datasets hinders deep learning model development.
• Transfer learning is a common technique to address data scarcity in medical imaging.

Purpose of the Study:

• To develop and evaluate deep learning models for COVID-19 diagnosis using chest X-ray images.
• To assess the robustness and sensitivity of the proposed COV-SNET models.

Main Methods:

• Developed two COV-SNET models, a deep neural network pretrained on over 100,000 X-ray images.
• Applied transfer learning to address the limited number of available COVID-19 X-ray images.
• Evaluated model performance using sensitivity metrics for three-class and two-class classification.

Main Results:

• Both COV-SNET models demonstrated robustness in diagnosing COVID-19.
• Achieved a sensitivity of 95% for both the three-class and two-class diagnostic models.
• Highlighted the limitations of public X-ray datasets for current COVID-19 deep learning models.

Conclusions:

• Deep learning models like COV-SNET show promise for COVID-19 diagnosis from X-rays.
• The study underscores the need for larger, more diverse public datasets to improve model generalizability.
• Future research should explore advanced techniques to overcome data limitations and enhance diagnostic accuracy.