Breathomics by Cavity-enhanced Comb Spectroscopy

Abstract: Breathomics aims to address the current unmet clinical needs by utilizing exhaled breath contents for non-invasive and real-time medical diagnostics. We demonstrate a frequency comb breathalyzer powered by machine learning for detecting COVID-19, finding 85 % accuracy among a 170-subject cohort. To enhance diagnostic power, we introduce Modulated Ringdown Comb Interferometry, a new technique enabling the quantification of “odor” of arbitrarily complex and unknown contents at new record sensing performance and requiring only simple instruments. For better interpretation, we present a data-driven framework to allow identification of concentrations and cross sections of species whether known or unknown to established databases.