Exploration of a Novel Universal Method for Identification of Chinese Medicines Varieties Based on Artificial Intelligence with Fingerprints on Multi-elements and Multi-data

KeyWord:a method of general  convolutional neural network  identification of traditional Chinese medicines  fingerprints on multi elements and multi data

Abstract:

      A novel universal method based on artificial intelligence with fingerprints on multi elements and multi data was established for the identification of varieties of Chinese medicines.Firstly,the same medicinal material was treated by different methods to acquire the information of chemical substances with different properties.A universal liquid phase method,including reversed phase chromatography,hydrophilic chromatography and molecular exclusion chromatography was used to collect the chemical information.The three liquid phase methods were complementary to each other to fully characterize the small polar molecules,large polar molecules and macromolecular compounds in the medicinal material.Then,a model with the fingerprints on multi elements and multi data was established by convolutional neural network to identify the varieties of Chinese medicines.In addition,all the fingerprints were treated by a standard process.Finally,a model with a test set accuracy of 92% was obtained.This method could be applied to the rapid,accurate and efficient identification of Chinese medicines,overcoming the subjective consciousness in the variety identification of traditional Chinese medicines,and giving therefore the identification results more objective and accurate.