Discrimination of authenticity of Fritillariae Cirrhosae Bulbus based on terahertz spectroscopy and chemometric analysis

A rapid, simple and non-destructive analysis methodology for discriminating the authenticity of Fritillariae Cirrhosae Bulbus (FCB) was established based on terahertz spectroscopy and chemometric analysis. 150 authentic FCB and 100 counterfeit FCB samples (Non-FCB) were analyzed. Then, different modeling algorithms including orthogonal projection to latent structure discriminant analysis (OPLS-DA), stepwise linear discriminant analysis (SLDA), support vector machine (SVM), back-propagation neural network (BPNN) and random forest (RF) were applied and were compared to obtain the best classification and model. Meanwhile, variable importance analysis (VIP) and 5 kinds of data pre-treatment technologies were investigated for improving the discriminant ability of the model. The results showed FCB and Non-FCB samples could be successfully discriminated by combining chemometric tools, and the OPLS-DA had the best identification effect than others. Meanwhile, it could be found that the VIP analysis could significantly increase the discriminant ability of the model and the discrimination accuracy of models based on UV-scaling was significantly better than others, which proved VIP was a powerful way to select useful variables and proper data pre-treatment technologies were very important for improving the discriminant effect. In addition, according to the best OPLS-DA model (vip-uv-OPLS-DA), the test samples could be successfully predicted with 100% accuracy, and the validation parameters R2X, R2Y and Q2 were 0.97, 1.0 and 1.0, respectively, which proved the model had powerful discriminant and predictive ability. In conclusion, terahertz spectroscopy combined with chemometric analysis is an effective method for the authenticity of FCB.