Rapid and Accurate Authentication of Porcini Mushroom Species Using Fourier Transform Near-Infrared Spectra Combined with Machine Learning and Chemometrics

Hong Liu,Honggao Liu,Jieqing Li,Yuanzhong Wang
DOI: https://doi.org/10.1021/acsomega.3c01229
IF: 4.1
2023-05-23
ACS Omega
Abstract:<p> </p> <p>Porcini mushrooms have high nutritional value and greatpotential,but different species are easily confused, so it is essential to identifythem rapidly and precisely. The diversity of nutrients in stipe andcap will lead to differences in spectral information. In this research,Fourier transform near-infrared (FT-NIR) spectral information aboutimparity species of porcini mushroom stipe and cap was collected andcombined into four data matrices. FT-NIR spectra of four data setswere combined with chemometric methods and machine learning for accurateevaluation and identification of different porcini mushroom species.From the results: (1) improved visualization level of t-distributedstochastic neighbor embedding (t-SNE) results after the second derivativepreprocessing compared with raw spectra; (2) after using multiplepretreatment combinations to process the four data matrices, the modelaccuracies based on support vector machine and partial least-squarediscriminant analysis (PLS-DA) under the best preprocessing methodwere 98.73–99.04% and 98.73–99.68%, respectively; (3)by comparing the modeling results of FT-NIR spectra with differentdata matrices, it was found that the PLS-DA model based on low-leveldata fusion has the highest accuracy (99.68%), but residual neuralnetwork (ResNet) model based on the stipe, cap, and average spectraldata matrix worked better (100% accuracy). The above results suggestthat distinct models should be selected for dissimilar spectral datamatrices of porcini mushrooms. Additionally, FT-NIR spectra have theadvantages of being nondevastate and fast; this method is expectedto be a promising analytical tool in food safety control.</p>
chemistry, multidisciplinary
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to quickly and accurately identify different species of Porcini mushrooms. Porcini mushrooms have high nutritional value and potential economic value, but different species of Porcini mushrooms are easily confused. Therefore, a fast and accurate method is required to identify them. Specifically, by combining Fourier - transform near - infrared spectroscopy (FT - NIR) with machine learning and chemometrics methods, the paper aims to develop an efficient, non - destructive identification technique to ensure that high - quality Porcini mushrooms are not misidentified, while preventing toxic or low - quality Porcini mushrooms from entering the market and safeguarding the health and rights of consumers. ### Research Background Wild edible mushrooms are widely popular for their unique flavor and texture, especially Porcini mushrooms, which are one of the most widely consumed and valuable wild edible mushrooms worldwide. However, different species of Porcini mushrooms are very similar in appearance, and it is difficult for even experienced experts to quickly and accurately identify them. In addition, the shelf life of fresh Porcini mushrooms is usually only 1 - 3 days, so they are mostly sold in the form of dried slices in the market, which further increases the difficulty of identification. Commercial fraud is common in the mushroom supply chain, so there is an urgent need to find a fast, economical and accurate Porcini mushroom variety identification technique. ### Solutions The paper uses the following methods to solve the problem: 1. **Data Collection**: Collected FT - NIR spectral information of the stipes and caps of different species of Porcini mushrooms and combined them into four data matrices. 2. **Pre - processing**: Used multiple pre - processing methods (such as first - order derivative, second - order derivative, standard normal variate transformation, etc.) to process the spectral data to remove noise and scattering information. 3. **Modeling**: Used partial least squares discriminant analysis (PLS - DA), support vector machines (SVM) and residual neural network (ResNet) models to model different data matrices respectively, exploring the feasibility of near - infrared spectroscopy combined with chemometrics and machine learning in non - destructive identification of Porcini mushrooms. 4. **Performance Evaluation**: By comparing the performance of different models, evaluated the impact of different data matrix types on the classification performance of the models. ### Main Results - **t - SNE Visualization**: After second - order derivative pre - processing, the visualization level of t - SNE results has been significantly improved. - **Model Accuracy**: Under the best pre - processing method, the accuracy rates of PLS - DA and SVM models are 98.73% - 99.04% and 98.73% - 99.68% respectively. - **Data Fusion**: The PLS - DA model based on low - level data fusion has the highest accuracy rate (99.68%), but the ResNet model based on stipe, cap and average spectral data matrices performs better (100% accuracy rate). ### Conclusion The research proves that FT - NIR spectroscopy combined with chemometrics and machine learning is an effective method that can be used for the rapid and accurate identification of Porcini mushrooms. Different models perform differently on different data matrices, among which the ResNet model performs best on stipe, cap and average spectral data matrices. Future research can expand the range of sample points and species types, create a more powerful FT - NIR spectral database, improve the adaptability of the model, and provide a reliable and effective means for online spectral monitoring in the Porcini mushroom supply chain.