COVID-19 prediction through X-ray images using transfer learning-based hybrid deep learning approach

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Science

Background:

• Accurate and rapid COVID-19 detection is crucial for pandemic control.
• While RT-PCR is effective, chest X-rays offer a beneficial alternative for detecting viral effects.
• Classifying chest X-ray reports using transfer learning is now feasible.

Purpose of the Study:

• To introduce a novel Hybrid Deep Convolutional Neural Network (HDCNN) for COVID-19 detection using chest X-rays.
• To evaluate the performance of HDCNN against established Convolutional Neural Network (CNN) models.
• To utilize gradient-weighted class activation mapping (Grad-CAMs) for decision interpretability.

Main Methods:

• Development of a Hybrid Deep Convolutional Neural Network (HDCNN) integrating CNN and Recurrent Neural Network (RNN) architectures.
• Application of transfer learning with Grad-CAMs for visualizing decision-making processes.
• Comparative analysis of HDCNN against Inception-v3, ShuffleNet, SqueezeNet, VGG-19, and DenseNet models.

Main Results:

• HDCNN achieved a high accuracy of 98.20%.
• The model demonstrated strong performance with a precision of 97.31%, recall of 97.1%, and an F1 score of 0.97.
• HDCNN outperformed other contemporary deep learning models in COVID-19 detection from chest X-rays.

Conclusions:

• The proposed HDCNN model shows significant promise for the diagnosis of COVID-19 using chest X-rays.
• HDCNN's superior performance suggests its potential as a valuable tool in clinical settings, pending necessary approvals.
• This deep learning approach offers an effective method for enhancing diagnostic capabilities during infectious disease outbreaks.