Rapid Quantitative Analysis of Active Ingredients in Radix Gentianae Macrophyllae Extracts Based on Near-Infrared Spectroscopy and Deep Learning

The content of active ingredients in traditional Chinese medicine (TCM) extracts greatly affects the quality and efficacy of TCM preparations. However, there is a lack of methods to rapidly analyze the active ingredients of large amounts of TCM extracts. In this study, based on near-infrared spectroscopy, the quantitative analytical model combining convolutional neural network (CNN) and gated recurrent unit (GRU) was constructed to predict the contents of loganic acid (LA) and gentiopicroside (GPS), which were the active ingredients in Radix Gentianae Macrophyllae (RGM) extracts. Meanwhile, the Bayesian algorithm was used to optimize the hyperparameters for improving the predictive performance of the CNN-GRU model. The results showed that the CNN-GRU model provided better prediction accuracy compared with the CNN, BP, and PLS models. In addition, the CNN-GRU model avoided complex spectral preprocessing and variable selection, which greatly improved the modeling facilitation. Finally, the feature extraction process of the CNN-GRU model was visualized to enhance the interpretability. This research innovatively integrated deep learning with near-infrared spectroscopy to achieve rapid and accurate analysis of the active ingredient content of TCM extracts, providing new ideas and methods for quality control of TCM preparations.