Automated Grading of Salvia miltiorrhiza Roots Using Neural Network

With the development of artificial intelligence technology at present, neural networks have been widely applied to different types of Chinese herbal medicine classification tasks and have achieved very good results. However, there is currently a lack of sufficient experiments to demonstrate that neural networks have the same ability in different quality grading tasks for a single type of medicinal material. The subject of our research, Salvia miltiorrhiza, has been a very valuable traditional Chinese medicine since ancient times and has been widely used in the treatment of cardiovascular diseases. At present, the screening of high-quality S.miltiorrhiza heavily relies on experienced pharmacists for manual screening, which is not only time-consuming and labor-intensive, but also unable to be widely popularized. Therefore, achieving the automation of S.miltiorrhiza quality grading tasks has important economic value. In this article, we constructed a small dataset on S.miltiorrhiza, which includes a total of 91 S.miltiorrhiza samples, classified by traditional Chinese medicine experts into two categories: A and B. Due to the difficulty in obtaining samples of S.miltiorrhiza, we used methods such as data expansion and parameter optimization to improve the quality grading effect of S.miltiorrhiza the absence of sufficient data. These experiments demonstrate that neural networks are also effective in quality grading tasks for individual species. Finally, we also provided evidence for neural network prediction through thermodynamic diagrams. And emphasized the future work direction and potential applications.