An Efficient Deep Learning Method for Detection of COVID-19 Infection Using Chest X-ray Images.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Science

Background:

• Deep learning, particularly Convolutional Neural Networks (CNNs), shows promise for detecting COVID-19 from chest radiography.
• Current state-of-the-art CNN models require substantial parameters and memory, hindering real-time diagnostic applications.
• There is an urgent need for efficient and lightweight CNN models for rapid COVID-19 detection.

Purpose of the Study:

• To propose a lightweight CNN model, LW-CORONet, for efficient and accurate COVID-19 detection using chest X-ray (CXR) images.
• To design a model with minimal learnable layers suitable for real-time applications.
• To evaluate the performance of LW-CORONet against existing models and analyze hyperparameter effects.

Main Methods:

• Development of LW-CORONet, a lightweight CNN model with five learnable layers (convolution, ReLU, pooling, fully connected).
• Evaluation of the model on two large CXR datasets (2250 and 15,999 images) for multi-class and binary classification.
• Comparative analysis with four pre-trained CNN models and state-of-the-art methods, including hyperparameter tuning.

Main Results:

• LW-CORONet achieved high classification accuracies: 98.67% and 99.00% on Dataset-1, and 95.67% and 96.25% on Dataset-2 for multi-class and binary classification, respectively.
• The model demonstrated superior efficiency in terms of parameters and memory usage compared to contemporary models.
• Hyperparameter analysis confirmed the model's robustness and effectiveness.

Conclusions:

• LW-CORONet is an effective lightweight CNN for rapid and accurate COVID-19 detection from CXR images.
• The model's reduced computational demands make it suitable for real-time diagnostic systems.
• LW-CORONet can serve as a valuable supplementary tool for radiologists in diagnosing COVID-19.