Plasmonic artificial inspector for herbal medicines via surface-enhanced Raman spectroscopy and deep learning

For the safety monitoring of herbal medicines (HMs), HM inspectors carry out an organoleptic examination before acceptance for market supply. The organoleptic test processes are often regarded as labor-intensive, thus calling for efficient and reliable aids. Here, we propose a plasmonic artificial HM inspector based on a collaboration between surface-enhanced Raman spectroscopy (SERS) and deep learning (DL). Inherently, a SERS spectrum of an HM specimen contains several peaks that match bioactive compounds in the sample, yielding so-called fingerprint information of HM. Besides, its rapid, few-second data-acquisition speed lends the SERS-DL analysis adaptability to a complementary inspection method for organoleptic examination. Regarding the accuracy and reliability of this new method, the synergistic integration of SERS with DL attains ~95% accuracy in labor-saving differentiation of 35 HM species with similar appearances or of the same genus. Our SERS-DL-based analysis can potentially aid the organoleptic HM inspection and help upgrade the HM database, along with images and other analytical chemistry data.