Surface-enhanced Raman spectroscopy (SERS) has long been recognized for its rapid and sensitive detection capabilities; however, challenges persist in practical fabrication of the substrates and interpreting complex data. Herein, we propose a deep learning (DL) assisted SERS approach to enable rapid and sensitive detection of analytes on a practical yet highly effective substrates prepared by direct spray-coating of a nanoparticle-free true solution of a reactive Ag ink and on-site thermal annealing mediated generation of nanostructures. This design ensured homogeneous distribution of Ag nanostructures throughout the entire substrate, significantly increasing the number of hotspots and enhancing the Raman signals, thereby achieving an impressive analytical enhancement factor of ~1010 in a reproducible and consistent manner. The diagnostic utility of this platform was demonstrated by detecting the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) spike (S) protein in both buffer and saliva, with detection limits of 74.3pg/mL and 7.43ng/mL, respectively. The DL-assisted SERS not only accurately identified the presence or absence of viral antigen, but also automatically quantified the viral load. This automatic identification achieved an outstanding accuracy of ~99.9%, highlighting the exceptional performance of the proposed platform. This simple, cost-effective, scalable, and ultra-sensitive DL-assisted SERS platform offers significant opportunities for early and precise detection in a range of analytical scenarios.
Read full abstract