E-DiCoNet: Extreme learning machine based classifier for diagnosis of COVID-19 using deep convolutional network.

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Healthcare
• Infectious Disease Diagnostics

Background:

• The rapid global spread of COVID-19 necessitates rapid and accurate diagnostic tools.
• Conventional COVID-19 testing methods often require specialized equipment and can have limited sensitivity.
• Chest X-rays offer a potential alternative for preliminary diagnosis of respiratory illnesses.

Purpose of the Study:

• To evaluate the performance of state-of-the-art Convolutional Neural Network (CNN) models for automated COVID-19 diagnosis using chest X-ray images.
• To propose and assess a modified pre-trained CNN-ResNet50 based Extreme Learning Machine (ELM) classifier for multi-class classification of COVID-19, bacterial pneumonia, and normal cases.
• To introduce a computationally efficient and highly accurate model for automatic detection of respiratory infections.

Main Methods:

• Utilized a dataset comprising chest X-ray images of patients with COVID-19, bacterial pneumonia, and healthy individuals.
• Employed a modified pre-trained ResNet CNN architecture integrated with an Extreme Learning Machine (ELM) classifier.
• Trained and validated the proposed CNN model on publicly available datasets.

Main Results:

• The proposed CNN-ResNet50-ELM model achieved high performance metrics, including 94.07% accuracy, 98.15% sensitivity, and 91.48% specificity.
• The model demonstrated superior classification performance compared to other state-of-the-art methods.
• Achieved precision of 98.15% and an F1 score of 91.22% for multi-class classification.

Conclusions:

• The developed CNN model offers a computationally efficient and highly accurate solution for the automatic multi-class diagnosis of COVID-19, pneumonia, and normal cases from chest X-rays.
• This automated diagnostic approach can significantly reduce the burden on healthcare systems.
• The model's high sensitivity and specificity support its potential clinical utility in early disease detection.