A method of two-dimensional correlation spectroscopy combined with residual neural network for comparison and differentiation of medicinal plants raw materials superior to traditional machine learning: a case study on Eucommia ulmoides leaves

Abstract

Background: Eucommia ulmoides leaf (EUL), as a medicine and food homology plant, is a high-quality industrial raw material with great development potential for a valuable economic crop. There are many factors affecting the quality of EULs, such as different drying methods and regions. Therefore, quality and safety have received worldwide attention, and there is a trend to identify medicinal plants with artificial intelligence technology. In this study, we attempted to evaluate the comparison and differentiation for different drying methods and geographical traceability of EULs. As a superior strategy, the two-dimensional correlation spectroscopy (2DCOS) was used to directly combined with residual neural network (ResNet) based on Fourier transform near-infrared spectroscopy. Results: (1) Each category samples from different regions could be clustered together better than different drying methods through exploratory analysis and hierarchical clustering analysis; (2) A total of 3204 2DCOS images were obtained, synchronous 2DCOS was more suitable for the identification and analysis of EULs compared with asynchronous 2DCOS and integrated 2DCOS; (3) The superior ResNet model about synchronous 2DCOS used to identify different drying method and regions of EULs than the partial least squares discriminant model that the accuracy of train set, test set, and external verification was 100%; (4) The Xinjiang samples was significant differences than others with correlation analysis of 19 climate data and different regions. Conclusions: This study verifies the superiority of the ResNet model to identify through this example, which provides a practical reference for related research on other medicinal plants or fungus.