Noninvasive SARS-CoV-2 detection using a low-cost electronic nose.

The COVID-19 pandemic highlighted the urgent need for rapid and accurate SARS-CoV-2 detection. Current diagnostic methods often suffer from discomfort, slow results, and limited accuracy in early infection stages. This study proposes a solution: a noninvasive, rapid, and accurate detection approach for point-of-care settings using a metal-oxide-sensor-based electronic nose. This innovative electronic nose uses an array of off-the-shelf gas sensors. These sensors detect and analyze the volatile organic compounds present in saliva and exhaled breath, which change based on the presence of the SARS-CoV-2 virus. We evaluated the discriminatory power of the electronic nose using a suite of machine learning algorithms, specifically K-Nearest Neighbors (KNN), Support Vector Machine (SVM), Neural Networks (NN), and Random Forest, to differentiate between SARS-CoV-2 infected and non-infected samples. Accuracy metrics ranged from 76% to 89% for exhaled breath samples and from 75% to 86% for saliva samples. Optimal accuracy was achieved with the KNN algorithm, yielding an Area Under the Curve (AUC) of 0.861 (95% CI 0.825‒0.897) for saliva and 0.895 (95% CI 0.850‒0.940) for exhaled breath. These results support the feasibility and proof-of-concept performance of a low-cost electronic nose for SARS-CoV-2 detection in a real-world hospital cohort.