Automated COVID-19 diagnosis and classification using convolutional neural network with fusion based feature...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Biology

Background:

• The COVID-19 pandemic necessitates rapid and accurate diagnostic tools.
• Traditional reverse transcription-polymerase chain reaction (rt-qPCR) methods for COVID-19 diagnosis are time-consuming and expensive.
• Automated diagnostic models using deep learning (DL) are crucial for efficient COVID-19 detection.

Purpose of the Study:

• To design an effective deep learning model for automated COVID-19 diagnosis and classification.
• To develop a fusion-based Convolutional Neural Network (CNN) model, termed FM-CNN, for enhanced diagnostic performance.
• To improve upon the limitations of traditional COVID-19 diagnostic methods.

Main Methods:

• The proposed FM-CNN model incorporates Wiener Filtering (WF) for noise reduction in chest X-ray (CXR) images.
• Feature extraction is performed using a fusion of Gray Level Co-occurrence Matrix (GLCM), Gray Level Run Length Matrix (GLRM), and Local Binary Patterns (LBP).
• Particle Swarm Optimization (PSO) is utilized for optimal feature subset selection, followed by CNN classification.

Main Results:

• The FM-CNN model demonstrated high diagnostic performance on a CXR dataset.
• The model achieved a maximum sensitivity of 97.22%, specificity of 98.29%, accuracy of 98.06%, and F-measure of 97.93% for multi-class classification.
• These results indicate the model's proficiency in accurately identifying COVID-19 from chest X-ray images.

Conclusions:

• The FM-CNN model provides an efficient and automated approach for COVID-19 diagnosis using chest X-rays.
• The fusion-based feature extraction and CNN classification offer a robust solution for early and accurate detection.
• This deep learning approach presents a promising advancement over conventional diagnostic techniques.