A strategy of fast evaluation for the raw material of Tiepi Fengdou using FT-NIR and ATR-FTIR spectroscopy coupled with chemometrics tools

Abstract

Tiepi Fengdou, as a precious traditional Chinese medicinal material in China, is a dried product of Dendrobium officinale that holds unique efficacy of nourishing Yin and clearing heat. However, there are many similar species named Fengdou for trade in the herbal market, leading to confusion about the currently commercially available Tiepi Fengdou medicinal materials, which brings great difficulties to the identification and evaluation of raw materials quality of Dendrobium. Therefore, it is necessary to establish a rapid and effective method for D. officinale and other species. In this study, deep learning (DL) models directly combined the two-dimensional correlation spectroscopy (2DCOS) images based on full bands and four characteristic bands of Fourier transform near-infrared (FT-NIR) and attenuated total reflectance Fourier transform infrared (ATR-FTIR) spectroscopy from D. officinale and 9 species of Dendrobium were established, and these identification effect of DL models were optimized and compared. The results show that the separation effect based on the two spectra with second derivative (SD) preprocessing is the best according to different categories via principal component analysis. Then, compared with ATR-FTIR, the DL models of SD full band, 9000–5500 cm−1 and 5250–4100 cm−1 band had absolute advantages to discriminate D. officinale and 9 species of Dendrobium based on FT-NIR. Based on this, the DL model with parameters of 16 bate size and 60 epochs combined with synchronous 2DCOS images is well based on FT-NIR to identify D. officinale and other species of Dendrobium. This method can not only quickly and accurately identify the raw materials (D. officinale) of Tiepi Fengdou, but also provide a theoretical basis for extended further research on other fields of medicinal plants or fungi.