Comparison of different optimizers implemented on the deep learning architectures for COVID-19 classification

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Infectious Diseases

Background:

• The global COVID-19 pandemic necessitates effective diagnostic tools, especially for asymptomatic cases.
• Distinguishing COVID-19 from similar respiratory illnesses like SARS-D is challenging due to overlapping symptoms.
• Breaking the transmission chain requires early detection, including identifying individuals without apparent symptoms.

Purpose of the Study:

• To evaluate the accuracy of various deep learning architectures and optimizers for COVID-19 image classification.
• To propose an effective deep learning model for accurate and rapid COVID-19 detection.
• To address overfitting issues in deep learning models through learning rate experimentation.

Main Methods:

• Comparison of deep learning architectures with different optimizers and learning rates.
• Implementation of an ensemble of a 2-layered Convolutional Neural Network (CNN).
• Application of Transfer Learning techniques for enhanced classification performance.

Main Results:

• Experimentation with various learning rates to mitigate overfitting in deep learning models.
• The proposed ensemble CNN with Transfer Learning achieved a classification accuracy of approximately 90.45%.
• The developed model demonstrated efficient classification with reduced computational time.

Conclusions:

• Deep learning, particularly ensemble CNNs with Transfer Learning, shows significant promise for COVID-19 detection.
• Accurate and rapid classification of COVID-19 from medical images is crucial for pandemic control.
• The study highlights the potential of AI in improving diagnostic capabilities for infectious diseases.