基于电子舌的白及及其近似饮片的快速辨识研究＊

目的 探讨电子舌方法用于白及及其近似饮片快速辨识的可行性。方法 收集45批白及饮片及其近似品天麻饮片30批、玉竹饮片30批、黄花白及饮片29批,分别进行药典与地方标准辨识（M₁法）、HPLC指纹图谱辨识（M₂法）,并结合原始采购信息获取最终饮片种类的标杆信息（Y）,再采集电子舌味觉感官数据（X）并利用化学计量学方法分别建立主成分分析-判别分析（PCA-DA）、偏最小二乘-判别分析（PLS-DA）的45批白及饮片与剩余89批饮片的二分类辨识模型和45批白及饮片、30批天麻饮片、30批玉竹饮片、29批黄花白及饮片的四分类辨识模型（Y=F（X）,M₃法）。结果 经留一法交互验证,基于PCA-DA、PLS-DA二分类辨识模型的正判率分别为98.51%、100.00%,基于PCA-DA、PLS-DA四分类辨识模型的正判率分别为100.00%（无未分类样本）、100.00%（有4个未分类样本）,模型判别良好,结合正判率与模型未分类样本数两项指标,最终选择二分类辨识以PLS-DA为最终辨识模型、四分类辨识以PCA-DA为最终辨识模型,两种模型正判率均为最高,且均未出现未分类样本。结论 电子舌可快速准确辨识白及及其近似饮片,为未来研发智能化中药饮片快速辨识设备提供了思路。

A Fast Identification of Bletillae Rhizoma and Similar Decoction Pieces with Electronic Tongue Technology

Objective To discuss the feasibility of fast identification of Bletillae Rhizoma and similar decoction pieces with electronic tongue technology.Methods Collected 45 batches of Bletillae Rhizoma pieces and 30 batches of Gastrodia elata pieces, 30 batches of Polygonatum odoratum pieces, 29 batches of Bletilla ochracea pieces, respectively conducted the pharmacopoeia and local standards identification (M₁ method), HPLC fingerprint identification (M₂ method), and combined the original purchase information to obtain the benchmark information (Y) of the final type of decoction pieces. Then electronic tongue taste sensory data (X) was collected and chemometric methods was used to establish two-class identification model of 45 batches of Bletillae Rhizoma and remaining 89 batches of decoction pieces and four-class identification model of 45 batches of Bletillae Rhizoma and 30 batches of Gastrodia elata, 30 batches of Polygonatum odoratum, 29 batches of Bletilla ochracea pieces with methods of principal component analysis-discriminant analysis (PCA-DA) and partial least squares-discriminant analysis (PLS-DA) (Y = F(X), M₃ method).Results With leave-one-out cross validation method, the positive judgment rates of two-class identification models based on the PCA-DA and PLS-DA were 98.51% and 100.00%, and the positive judgment rates of four-class identification models based on the PCA-DA and PLS-DA were 100.00 % (No unclassified samples) and 100.00% (there are 4 unclassified samples). The model discriminated well. And finally, the two-class identification with PLS-DA was chosen as the final identification model. Four-class identification with PCA-DA as the final identification model through combining the two indicators of positive judgment rate and the unclassified samples of the model. Two kinds of models had the highest positive judgment rate, and no unclassified samples appeared.Conclusion The electronic tongue can quickly and accurately identify Bletillae Rhizoma and similar decoction pieces, providing new ideas for the future development and research of intelligent equipment for fast identification of traditional Chinese medicine decoction pieces.