A Fine-tuned deep convolutional neural network for chest radiography image classification on COVID-19 cases

Area of Science:

• Medical Imaging and Diagnostics
• Artificial Intelligence in Healthcare
• Infectious Disease Detection

Background:

• The global impact of coronavirus disease 2019 (COVID-19) necessitates effective diagnostic tools.
• Chest radiography abnormalities are indicative of COVID-19 infection.
• Accurate and rapid screening of COVID-19 patients is crucial for containment.

Purpose of the Study:

• To propose a unique framework for diagnosing COVID-19 infection using chest radiography.
• To leverage deep learning models for enhanced COVID-19 detection accuracy.

Main Methods:

• A modified VGG-16 deep convolutional neural network was developed.
• Fully connected layers were replaced and fine-tuned for COVID-19 classification.
• A two-stage training process involving network freezing and unfreezing was employed.

Main Results:

• The proposed classification model achieved 97.12% accuracy for COVID-19 identification.
• High sensitivity (99.2%) and specificity (99.6%) were recorded.
• The model demonstrated superior performance compared to other classification models.

Conclusions:

• The developed framework offers a highly accurate and efficient method for COVID-19 diagnosis from chest X-rays.
• This AI-driven approach can significantly support early detection and patient screening efforts.
• The fine-tuned VGG-16 model shows promise in combating the COVID-19 pandemic.