Applying Machine Learning with Localized Surface Plasmon Resonance Sensors to Detect SARS-CoV-2 Particles.

Jiawei Liang,Yu Qin,Wei Zhang,Wenjun Hu,Ying Li,Gang Logan Liu

doi:10.3390/bios12030173

Abstract

The sudden outbreak of COVID-19 rapidly developed into a global pandemic, which caused tens of millions of infections and millions of deaths. Although SARS-CoV-2 is known to cause COVID-19, effective approaches to detect SARS-CoV-2 using a convenient, rapid, accurate, and low-cost method are lacking. To date, most of the diagnostic methods for patients with early infections are limited to the detection of viral nucleic acids via polymerase chain reaction (PCR), or antigens, using an enzyme-linked immunosorbent assay or a chemiluminescence immunoassay. This study developed a novel method that uses localized surface plasmon resonance (LSPR) sensors, optical imaging, and artificial intelligence methods to directly detect the SARS-CoV-2 virus particles without any sample preparation. The virus concentration can be qualitatively and quantitatively detected in the range of 125.28 to 106 vp/mL through a few steps within 12 min with a limit of detection (LOD) of 100 vp/mL. The accuracy of the SARS-CoV-2 positive or negative assessment was found to be greater than 97%, and this was demonstrated by establishing a regression machine learning model for the virus concentration prediction (R2 > 0.95).

Highlights

Achieving a rapid diagnosis of SARS-CoV-2 is important to prevent the spread of the epidemic and enables early intervention in the disease [1]
Virus detection based on surface plasmon resonance (SPR) usually requires sophisticated and expensive optomechanical systems to monitor the changes in the refraction angle produced by molecular interactions, which limits its large-scale application [15]
We introduce microscopic color imaging based on nanostructured local surface plasmon resonance (LSPR)

Summary

Introduction

Achieving a rapid diagnosis of SARS-CoV-2 is important to prevent the spread of the epidemic and enables early intervention in the disease [1]. Some studies have applied surface plasmon resonance (SPR) and low-cost local surface plasmon resonance (LSPR) to the rapid detection of SARS-CoV-2 particles [10–15] Conventional LSPR devices only extract the global changes in the entire or partial surface optical signal, since most of them only extract simple spectral signals, such as changes in the phase or intensity of spectral peaks, which conceals and ignores substantial rich information regarding the temporal and spatial distribution of the signal changes during detection [10–14]. These limitations will affect their sensitivity, specificity and accuracy in clinical detection

Methods

Results

Conclusion