Metal oxide-based gas sensor array for VOCs determination in complex mixtures using machine learning.

Area of Science:

• Materials Science and Engineering
• Analytical Chemistry
• Biomedical Engineering

Background:

• Non-invasive disease detection using breath volatile organic compounds (VOCs) is a promising diagnostic approach.
• Accurate identification and quantification of multiple VOCs in complex mixtures remain a significant challenge.

Purpose of the Study:

• To develop a metal oxide sensor array capable of detecting and quantifying multiple VOCs in breath.
• To apply machine learning algorithms for accurate analysis of sensor responses to VOC mixtures.

Main Methods:

• Fabrication of a three-component metal oxide sensor array (NiO-Au, CuO-Au, ZnO-Au) using DC reactive sputtering.
• Exposure of the sensor array to individual and mixed concentrations of ethanol, acetone, toluene, and chloroform.
• Analysis of sensor response data using various machine learning algorithms, including Random Forest (RF) and K-Nearest Neighbor (KNN).

Main Results:

• The sensor array demonstrated high cross-sensitivity to different VOCs.
• KNN and RF algorithms achieved over 99% accuracy in classifying VOCs within mixtures.
• KNN regression analysis yielded an R² value > 0.99, with low limits of detection (LOD) for individual VOCs.

Conclusions:

• The developed metal oxide sensor array, coupled with machine learning, can accurately classify and quantify multiple VOCs simultaneously.
• This technology holds significant potential for non-invasive disease diagnosis and treatment monitoring.