Rapid Detection of SARS-CoV-2 Antigen Utilizing Machine Learning-Enabled Graphene-Based Smart Gas Sensors

Since its outbreak in December 2019, COVID-19 has rapidly spread around the world causing more than 759 million cumulative cases and more than 6.8 million cumulative deaths. To contain the pandemic, a highly sensitive, rapid, cost-efficient, simple-to-use, and noninvasive diagnosis for ruling out infection of COVID-19 at its early stage is crucial. Breath analysis is a promising approach for the detection of SARS-CoV-2 since it is noninvasive and easy to use. Most of current works on COVID-19 diagnosis using breath analysis are based on the analysis of the variation of volatile organic compounds (VOCs) biomarkers in the exhaled breath due to COVID-19 infection, while there are few efforts reported on the direct detection of SARS-CoV-2 in exhaled aerosols. In this work, a novel approach for the rapid detection of aerosols containing SARS-CoV-2 antigen using a single-channel functionalized graphene-based chemiresistive nanosensor is presented. Multiple transient features are extracted from the acquired sensing response signal and utilized as fingerprints of antigen-containing aerosols. With the supervised machine learning model, a high prediction performance for effective detection is achieved, such as accuracy-97.2%, sensitivity-92.3%, and specificity-100%. This proof-of-concept work proposes the first steps toward an efficient and effective scheme to detect the SARS-CoV-2 antigen by breath analysis, which may facilitate a noninvasive, rapid, highly sensitive approach to COVID-19 diagnosis.