基于人工智能-多元多息指纹图谱探索中药一法通识品种鉴定新方法

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

该文建立了一种基于中药多元多息指纹图谱联合人工智能识别的中药一法通识品种鉴定新方法,方法先通过对同一味药材采用不同处理手段得到具有不同性质的化学成分信息,并构建一种集反相色谱法、亲水色谱法以及分子排阻色谱法为一体的具有普适性的多元多息指纹图谱采集方式,实现了药材中小极性小分子、大极性小分子及大分子类化合物的全面表征,再对采集的多元多息指纹图谱进行数据标准化处理,采用卷积神经网络识别不同品种中药材,获得了准确率达92%的识别模型。该方法能够对中药品种进行快速、准确、高效地鉴定,克服了传统中药品种鉴别中的主观色彩,能够更加客观精准地给出鉴定结果。

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.