Using Deep Learning Model for Adapting and Managing COVID-19 Pandemic Crisis

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Epidemiology

Background:

• The COVID-19 pandemic highlighted the need for advanced telemedicine tools for early disease detection and management.
• Oman sought effective solutions to manage the pandemic crisis and prepare for future outbreaks.

Purpose of the Study:

• To create a smart and effective telemedicine tool for early COVID-19 detection and diagnosis.
• To support pandemic crisis management in Oman and future pandemic containment.
• To develop robust, real-time diagnostic models for telemedicine applications.

Main Methods:

• Utilized Machine Learning (ML) and Deep Learning (DL) techniques, including Convolutional Neural Networks using Tensorflow (CNN-TF).
• Developed an Automated Medical Immediate Diagnosis service (AMID).
• Analyzed Chest X-rays (CXRs) for disease classification and severity assessment.

Main Results:

• Random Forest Regression was the best among five regression models.
• Random Forest classification outperformed eight classification models and Recurrent Neural Network using Tensorflow (RNNTF).
• K-Means++ was the best clustering model; CNN-TF successfully discriminated between positive and negative COVID-19 cases.

Conclusions:

• The developed ML and DL models are effective for early COVID-19 detection and diagnosis via telemedicine.
• These tools can significantly improve medical diagnostics, assist medical staff, and aid in pandemic management.
• The study provides a robust framework for future pandemic preparedness and response systems.