Machine Learning Sensors for Diagnosis of COVID-19 Disease Using Routine Blood Values for Internet of Things...

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Biomarker Discovery

Background:

• Healthcare digitalization and the Internet of Things (IoT) necessitate efficient human sensor applications for real-time health monitoring.
• Prompt diagnosis of COVID-19 is crucial, but traditional methods involving tests and imaging are time-consuming and expensive.
• Machine learning (ML) sensors offer a potential solution for rapid and accessible COVID-19 detection within healthcare and ambient assisted living (AAL) contexts.

Purpose of the Study:

• To develop a fast, reliable, and cost-effective diagnostic tool for COVID-19 using routine blood values (RBVs).
• To evaluate the performance of various ML models in detecting COVID-19 based on RBVs.
• To identify key blood parameters and their combinations as effective biomarkers for ML-based COVID-19 diagnosis.

Main Methods:

• Utilized a dataset of 5296 patients (equal COVID-19 positive and negative) with 51 routine blood values.
• Examined 13 popular classifier ML models and one neural network model (LogNNet).
• Employed histogram-based gradient boosting (HGB) for classification and feature importance analysis.

Main Results:

• The HGB classifier achieved 100% accuracy in detecting COVID-19 in just 6.39 seconds.
• Identified 11 critical routine blood values (LDL, cholesterol, HDL-C, MCHC, triglyceride, amylase, UA, LDH, CK-MB, ALP, MCH) as key diagnostic features.
• Demonstrated the diagnostic utility of single, double, and triple combinations of these identified features.

Conclusions:

• The HGB model provides a highly accurate and efficient method for COVID-19 diagnosis using routine blood values.
• The 11 identified blood parameters and their combinations serve as valuable biomarkers for ML-based diagnostic sensors.
• Proposed integration of these biomarkers and ML models into edge computing (Arduino) and cloud IoT services for enhanced healthcare applications.