DiabeticSense: A Non-Invasive, Multi-Sensor, IoT-Based Pre-Diagnostic System for Diabetes Detection Using Breath

Area of Science:

• Biomedical Engineering
• Medical Diagnostics
• Digital Health

Background:

• Diabetes mellitus is a chronic metabolic disorder requiring frequent blood glucose monitoring.
• Current invasive methods like finger-prick tests cause discomfort and lead to poor patient adherence.
• There is a need for convenient, non-invasive alternatives for early diabetes detection.

Purpose of the Study:

• To develop DiabeticSense, a novel, portable, non-invasive system for diabetes detection using breath samples.
• To create an affordable digital health device for early diabetes detection and intervention.
• To assess the efficacy of breath volatile organic compound (VOC) analysis for identifying diabetic conditions.

Main Methods:

• Designed a portable device utilizing electrochemical sensors to analyze VOCs in breath.
• Integrated vital signs with sensor voltage data for diabetes prediction.
• Collected breath samples from 100 patients at a hospital for dataset creation.
• Employed a gradient boosting classifier model for data processing and cross-validation.

Main Results:

• Achieved a diagnostic accuracy of 86.6% for diabetes detection.
• Demonstrated a 20.72% improvement in accuracy compared to existing regression techniques.
• Identified distinct VOC concentration patterns between diabetic and non-diabetic individuals.

Conclusions:

• The DiabeticSense system offers a non-invasive, cost-effective, and user-friendly approach to preliminary diabetes detection.
• This technology has the potential to significantly enhance patient adherence to regular diabetes monitoring.
• Breath analysis presents a promising avenue for developing advanced, patient-centric diagnostic tools for metabolic disorders.