Raman spectroscopy, used trans-cutaneously and non-invasively from a finger, to predict COVID-19: A feasibility, proof-of-concept study.

BACKGROUNDA definitive COVID-19 infection typically is diagnosed by laboratory tests, including real-time, reverse-transcriptase Polymerase Chain Reaction (PCR)-based testing. These currently available COVID-19 tests require the patient to provide an extra-corporeal specimen and the results may not be immediate. Consequently, a variety of rapid antigen tests for COVID-19, all with a wide range of accuracy in terms of sensitivity and specificity, has proliferated (1,2). These rapid tests now represent a significantly larger proportion of all testing done for COVID-19, yet suffer from requiring a physical specimen from the nose or mouth and waiting 15 minutes for most. As a solution, we propose a non-invasive, trans-cutaneous, real-time viral detection device, based on the principles of Raman spectroscopy and machine learning. It does not require any extra-corporeal specimens and can be configured for self-administration. It can be easily used by non-experts and does not require medical training. Our approach suggests that our non-invasive, transcutaneous method may be broadly useful not only in COVID-19 diagnosis, but also in other diagnoses. METHODS160 COVID positive (+) patients and 316 COVID negative (-) patients prospectively underwent nasal PCR testing concurrently with testing using our non-invasive, transcutaneous, immediate viral detector. Both the PCR and our experimental viral detector tests were performed side-by-side on outpatients (N=389) as well as inpatients (N= 87) at Holy Name Medical Center in Teaneck, NJ between June 2021 and August, 2022. The spectroscopic data were generated using an 830nm Raman System with SpectraSoft (W2 Innovations)and then, using machine learning, processed to provide an immediate prediction. A unique patient-interface for finger insertion enabled the application of Raman spectroscopy to viral detection in humans. RESULTSThe data analysis algorithm demonstrates that there is an informative Raman spectrum output from the device, and that individual Raman peaks vary between cases and controls. Our proof-of-concept study yields encouraging results, with a specificity for COVID-19 of 0.75, and a sensitivity (including asymptomatic patients) of 0.80. CONCLUSIONSThe combination of Raman spectroscopy, artificial intelligence, and our unique patient-interface admitting only a patient finger achieved test results of 0.75 specificity and 0.80 sensitivity for COVID-19 testing in this first in human proof-of-concept study. More significantly, the predictability improved with increasing data.