A Multimodel-Based Screening Framework for C-19 Using Deep Learning-Inspired Data Fusion.

Area of Science:

• Medical Informatics
• Artificial Intelligence
• Biomedical Engineering

Background:

• The rise of Internet of Medical Things (IoMT) and edge computing enhances remote healthcare monitoring.
• Existing COVID-19 screening models using RCNN, FTE, and wearable sensors are computationally intensive and unsuitable for lightweight environments.

Purpose of the Study:

• To propose a novel multimodal screening framework leveraging deep learning-inspired data fusion for improved remote COVID-19 monitoring.
• To enhance computational efficiency and suitability for lightweight environments in healthcare applications.

Main Methods:

• A Variation Encoder (VEN) measures skin temperature using YoLo-identified Regions of Interest (RoI).
• A multi-data fusion model integrates electronic health records with wearable sensor data.
• A data reduction mechanism optimizes computational efficiency, and a contingent probability method estimates feature weights.

Main Results:

• The proposed framework achieves a precision of 95.2% on a lab dataset.
• The model demonstrates superior performance compared to state-of-the-art methods.
• The design effectively fuses multimodal data, predicts feature weights, and selects relevant features.

Conclusions:

• The multimodal screening framework offers a computationally efficient and highly accurate solution for remote COVID-19 monitoring.
• The integration of deep learning, data fusion, and feature optimization enhances screening capabilities for IoMT applications.
• This approach provides a robust method for assessing abnormal COVID-19 instances using thermal and sensory data.