Automatic COVID-19 Detection from Chest X-ray using Deep MobileNet Convolutional Neural Network

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Biology

Background:

• The COVID-19 pandemic necessitates rapid and accurate diagnostic tools.
• Chest X-rays (CXRs) are crucial for detecting lung abnormalities associated with COVID-19.
• Automated detection systems can alleviate healthcare system strain.

Purpose of the Study:

• To develop a computational tool for automated COVID-19 detection using CXR images.
• To enhance diagnostic accuracy through deep learning and machine learning techniques.
• To differentiate between COVID-19, non-COVID lung opacity, and normal CXR findings.

Main Methods:

• Utilized a deep learning transfer learning approach with a pre-trained MobileNetV2 Convolutional Neural Network (CNN).
• Employed a Support Vector Machine (SVM) classifier on features extracted from the MobileNetV2 model.
• Evaluated the model's performance on CXR data for COVID-19, non-COVID lung opacity, and normal cases.

Main Results:

• The baseline model achieved 92.28% accuracy.
• The transfer learning model combining MobileNetV2 and SVM achieved an improved accuracy of 93.2%.
• This hybrid approach demonstrated high effectiveness in classifying COVID-19, non-COVID lung opacity, and normal lung conditions.

Conclusions:

• The integration of MobileNetV2 features with SVM classifiers offers a robust method for COVID-19 detection.
• Deep learning algorithms show significant promise in improving the discrimination of various lung conditions from CXR.
• This computational tool can aid in the rapid and accurate diagnosis of COVID-19, supporting clinical decision-making.