LMNet: Lightweight multi-scale convolutional neural network architecture for COVID-19 detection in IoMT environment

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Biology

Background:

• The COVID-19 pandemic, caused by SARS-CoV-2, necessitates rapid and accurate diagnostic tools.
• Current Reverse Transcription Polymerase Chain Reaction (RT-PCR) testing is manual, time-consuming, and poses risks to healthcare professionals.
• There is a critical need for automated, efficient diagnostic systems for COVID-19 detection.

Purpose of the Study:

• To develop a computationally efficient, lightweight Convolutional Neural Network (CNN) architecture for automated COVID-19 detection.
• To evaluate the performance of the proposed LMNet model, particularly when ensembled with other state-of-the-art models.
• To explore the integration of the developed model within the Internet of Medical Things (IoMT) environment.

Main Methods:

• A novel multi-scale lightweight CNN (LMNet) architecture was designed for COVID-19 detection.
• The LMNet model was ensembled with DenseNet169 and MobileNetV2 to enhance diagnostic performance.
• The computational cost and memory requirements of the proposed model were assessed.

Main Results:

• The proposed LMNet architecture demonstrated reduced computational expense and memory footprint compared to existing models.
• The ensemble of LMNet with DenseNet169 and MobileNetV2 achieved superior performance over other state-of-the-art models.
• The ensemble model's suitability for backend integration in smart devices for IoMT applications was confirmed.

Conclusions:

• The developed LMNet-based ensemble model offers a promising automated solution for rapid and efficient COVID-19 diagnosis.
• This approach addresses the limitations of traditional RT-PCR testing, including speed and risk to personnel.
• The model's integration potential within IoMT environments highlights its utility for future healthcare applications.