COVIDetectioNet: COVID-19 diagnosis system based on X-ray images using features selected from pre-learned deep...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Infectious Diseases

Background:

• The COVID-19 pandemic necessitates rapid and reliable diagnostic tools.
• Traditional COVID-19 testing methods can be time-consuming and have limited availability.
• Radiological imaging, specifically chest X-rays, presents a faster and potentially more accessible diagnostic alternative.

Purpose of the Study:

• To develop and validate an expert-designed artificial intelligence system, COVIDetectioNet, for diagnosing COVID-19 using chest X-ray images.
• To leverage deep learning and feature selection techniques for enhanced diagnostic accuracy.
• To compare the performance of the proposed model against existing diagnostic methods.

Main Methods:

• Utilized a pretrained AlexNet Convolutional Neural Network (CNN) architecture with transfer learning.
• Employed the Relief feature selection algorithm to identify effective features from deep layers.
• Classified selected features using a Support Vector Machine (SVM) classifier.
• Trained and validated the model on a dataset of 6092 chest X-ray images (Normal, COVID-19, Pneumonia).

Main Results:

• The COVIDetectioNet model achieved a diagnostic accuracy of 99.18%.
• The model demonstrated superior performance compared to previously reported studies for COVID-19 detection from X-rays.
• Feature selection and SVM classification effectively identified COVID-19 indicators in chest X-rays.

Conclusions:

• The proposed COVIDetectioNet model offers a highly accurate and efficient AI-driven solution for COVID-19 diagnosis via chest X-rays.
• This approach shows significant potential to supplement or expedite traditional diagnostic workflows.
• The study highlights the efficacy of combining deep learning, feature selection, and machine learning for medical image analysis in infectious disease detection.