Microbial volatile organic compound fingerprints for non-contact and real-time infection monitoring using electronic nose in infant incubator

Abstract

In infant incubators, neonatal infections pose critical morbidity and mortality risks, often requiring empirical antibiotic treatments due to time-consuming diagnostic methods, which can lead to potential misuse and antibiotic resistance. To address this challenge, we developed a non-invasive electronic nose system for real-time infection monitoring by identifying pathogens through microbial volatile organic compounds (mVOCs) as an alerting system. The sensor array, optimized using nine specific mVOC chemicals produced by sepsis-causing pathogens, demonstrated effective discrimination among seven pathogens: S. aureus and S. epidermidis (Gram-positive bacteria), E. coli and K. pneumoniae (Gram-negative bacteria), and C. albicans, C. glabrata, and C. parapsilosis (fungi). The electronic nose, employing an LSTM model, achieved 97 % classification accuracy under laboratory conditions. Furthermore, the system's quantification ability was validated with R² values exceeding 0.80 for all seven pathogens. When tested in a real-size incubator (155 liters) simulating practical applications, the system achieved an overall accuracy of 85.4 % in microbial discrimination. These findings suggest that integrating the electronic nose into a smart incubator could facilitate evidence-based antibiotic prescriptions through real-time, non-invasive infection monitoring.