Machine Learning for COVID-19 Determination Using Surface-Enhanced Raman Spectroscopy

Area of Science:

• Biomedical Engineering
• Analytical Chemistry
• Infectious Disease Diagnostics

Background:

• Rapid and cost-effective detection of SARS-CoV-2 is crucial for clinical management.
• Current diagnostic methods face challenges in speed and accessibility.
• Spectroscopic techniques coupled with chemometrics present a promising alternative.

Purpose of the Study:

• To evaluate the efficacy of surface-enhanced Raman spectroscopy (SERS) combined with machine learning (ML) for SARS-CoV-2 detection.
• To assess the performance of SERS-ML in analyzing clinical samples such as saliva and nasopharyngeal swabs.
• To determine the feasibility of this method for practical clinical application.

Main Methods:

• Surface-enhanced Raman spectroscopy (SERS) was employed to analyze saliva and nasopharyngeal swab samples.
• Machine learning algorithms, including random forest (RF), were utilized for spectral data analysis.
• A cohort of 175 saliva and 114 nasopharyngeal swab samples were investigated.

Main Results:

• The random forest classifier demonstrated high performance in analyzing SERS spectra.
• For saliva samples, the random forest model achieved a precision of 94.0% and a recall of 88.9%.
• The SERS-ML approach showed effectiveness even with a limited number of clinical samples.

Conclusions:

• The integration of SERS and shallow machine learning provides a viable strategy for SARS-CoV-2 detection.
• This method holds potential for rapid, low-cost, and efficient identification of SARS-CoV-2 in clinical settings.
• Further development could enhance its role in infectious disease diagnostics.