Department of Food Engineering, School of Food Engineering, University of Campinas, Campinas, SP, Brazil.
Department of Food Science, School of Food Engineering, University of Campinas, Campinas, SP, Brazil.
Spectrochim Acta A Mol Biomol Spectrosc. 2023 Mar 15;289:122226. doi: 10.1016/j.saa.2022.122226. Epub 2022 Dec 9.
Cinnamon is a valuable aromatic spice widely used in pharmaceutical and food industry. Commonly, two-cinnamon species are available in the market, Cinnamomum verum (true cinnamon), cropped only in Sri Lanka, and Cinnamomum cassia (false cinnamon), cropped in different geographical origins. Thus, this work aimed to develop classification models based on NIR-hyperspectral imaging (NIR-HSI) coupled to chemometrics to classify C. verum and C. cassia sticks. First, principal component analysis (PCA) was applied to explore hyperspectral images. Scores surface displayed the high similarity between species supported by comparable macronutrient concentration. PC3 allowed better class differentiation compared to PC1 and PC2, with loadings exhibiting peaks related to phenolics/aromatics compounds, such as coumarin (C. cassia) or catechin (C. verum). Partial least square discriminant analysis (PLS-DA) and Support vector machine (SVM) reached similar performance to classify samples according to origin, with error = 3.3 % and accuracy = 96.7 %. A permutation test with p < 0.05 validated PLS-DA predictions have real spectral data dependency, and they are not result of chance. Pixel-wise (approach A) and sample-wise (approach B, C and D) classification maps reached a correct classification rate (CCR) of 98.3 % for C. verum and 100 % for C. cassia. NIR-HSI supported by classification chemometrics tools can be used as reliable analytical method for cinnamon authentication.
肉桂是一种有价值的芳香香料，广泛应用于制药和食品工业。市场上通常有两种肉桂，即锡兰肉桂（Cinnamomum verum），仅在斯里兰卡种植，以及中国肉桂（Cinnamomum cassia），在不同的地理起源地种植。因此，本工作旨在开发基于近红外高光谱成像（NIR-HSI）结合化学计量学的分类模型，以对 C. verum 和 C. cassia 棒进行分类。首先，应用主成分分析（PCA）对高光谱图像进行探索。得分表面显示了物种之间的高度相似性，这得到了类似的大量营养素浓度的支持。与 PC1 和 PC2 相比，PC3 允许更好的分类区分，其载荷显示出与酚类/芳香族化合物相关的峰，如香豆素（C. cassia）或儿茶素（C. verum）。偏最小二乘判别分析（PLS-DA）和支持向量机（SVM）达到了类似的性能，根据产地对样品进行分类，误差=3.3%，准确率=96.7%。置换检验（p<0.05）验证了 PLS-DA 预测具有真实的光谱数据依赖性，而不是偶然的结果。像素级（方法 A）和样本级（方法 B、C 和 D）分类图对 C. verum 的正确分类率（CCR）达到 98.3%，对 C. cassia 的正确分类率达到 100%。分类化学计量学工具支持的 NIR-HSI 可以用作肉桂鉴定的可靠分析方法。
