IoMT-fog-cloud based architecture for Covid-19 detection

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Science

Background:

• Current COVID-19 detection methods often suffer from small datasets, limited validation, and a focus on accuracy over model complexity.
• Existing cloud-based healthcare systems face privacy, security, latency, and performance issues due to centralized data transmission.
• The need for efficient, accurate, and less complex diagnostic tools for COVID-19 is critical.

Purpose of the Study:

• To develop a computationally efficient and accurate system for detecting COVID-19, pneumonia, and normal cases from medical images.
• To address the limitations of existing methods by reducing model complexity and improving data privacy and transmission performance.
• To enhance the Quality of Service (QoS) in cloud-based medical diagnostics.

Main Methods:

• Image preprocessing using Discrete Wavelet Transform (DWT) for feature decomposition and Principal Component Analysis (PCA) for dimensionality reduction.
• Feature energy tracking with Teager Kaiser Energy Operator (TKEO), Shannon Wavelet Entropy Energy (SWEE), and Log Energy Entropy (LEE) to select optimal features.
• Implementation of a Convolutional Neural Network (CNN) with deep neurons and small kernel windows to minimize model complexity and improve feature learning.

Main Results:

• The DWT-PCA and TKEO techniques demonstrated effective noise reduction, with PSNR of 3.14 dB and SNR of 1.48 (original) and 1.47 (preprocessed).
• The CNN model achieved high performance in classification: 97% accuracy, 100% precision, 97% recall, 99% F1-score, and 98% AUC.
• Integration of fog computing as an intermediate layer improved QoS by addressing latency and computational costs, enhancing data security and privacy.

Conclusions:

• The proposed DWT-PCA and TKEO-enhanced CNN system effectively reduces model complexity while maintaining high diagnostic accuracy for COVID-19 detection.
• The incorporation of fog computing significantly improves the performance, security, and privacy of the diagnostic system.
• The developed IFC-Covid system shows potential as a user-friendly, cost-effective, and real-time application for COVID-19 screening.