COVID-19 Detection using Hybrid CNN-RNN Architecture with Transfer Learning from X-Rays

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Deep Learning

Background:

• COVID-19's rapid global spread necessitates effective diagnostic tools.
• Accurate identification of infected individuals is crucial for controlling viral transmission.

Purpose of the Study:

• To develop a hybrid deep learning model combining CNNs and RNNs for enhanced COVID-19 detection from X-ray images.
• To leverage transfer learning to improve diagnostic accuracy.

Main Methods:

• Utilized four pre-trained CNNs (InceptionnetV3, Densenet121, Inception-ResNet V2, VGG19) for feature extraction.
• Employed an RNN to capture temporal dependencies in extracted features.
• Evaluated performance on a diverse dataset including COVID-19, pneumonia, and healthy X-rays.
• Applied Grad-CAM for visualizing infected areas in X-ray images.

Main Results:

• The hybrid CNN-RNN architecture demonstrated high accuracy, precision, recall, AUC, and F1-score.
• The VGG19-RNN model outperformed other state-of-the-art methods in COVID-19 detection.
• Achieved optimal training and validation accuracy of 99% and 97.70% respectively for VGG19-RNN.

Conclusions:

• The hybrid CNN-RNN model effectively captures spatial and temporal information for improved COVID-19 detection.
• This approach offers a robust and efficient solution for diagnosing COVID-19 from X-ray images.
• The model can support healthcare professionals in making timely and accurate diagnoses, aiding global pandemic control efforts.