Xiao Li, Liu Jinxin, Hua Marti Z, Lu Xiaonan
Department of Food Science and Agricultural Chemistry, Faculty of Agricultural and Environmental Sciences, McGill University, Sainte-Anne-de-Bellevue, QC H9X 3V9, Canada.
Department of Food Science and Agricultural Chemistry, Faculty of Agricultural and Environmental Sciences, McGill University, Sainte-Anne-de-Bellevue, QC H9X 3V9, Canada.
Food Chem. 2025 Jan 15;463(Pt 2):141289. doi: 10.1016/j.foodchem.2024.141289. Epub 2024 Sep 16.
Total phenolic content (TPC) and antioxidant capacity of maple syrup were determined using Raman spectroscopy and deep learning. TPC was determined by Folin-Ciocalteu assay, while the antioxidant capacity was measured by 2,2-diphenyl-1picrylhydrazyl (DPPH) assay, oxygen radical absorbance capacity (ORAC) assay, and ferric reducing antioxidant power (FRAP) assay. A total of 360 spectra were collected from 36 maple syrup samples of different colours (dark, amber, light) by both benchtop and portable Raman spectrometers. These spectra were used to establish predictive models for assessing the antioxidant profiles of maple syrup. Deep learning models developed along with portable Raman spectroscopy exhibited comparable predictive performance to those developed along with benchtop Raman spectroscopy. Base on the spectral dataset collected using portable Raman spectroscopy, the developed deep learning models exhibited low RMSEs (root mean square errors, 7.2-17.9 % of mean reference values), low MAEs (mean absolute errors, 5.2-13.1 % of mean reference values) and high R values (>0.88). The results showed a great goodness of fit and accuracy for predicting the antioxidant profiles of maple syrup, indicating the potential of using portable Raman spectrometer for on-site analysis of antioxidant profiles of maple syrup.
使用拉曼光谱和深度学习测定了枫糖浆的总酚含量（TPC）和抗氧化能力。TPC通过福林-西奥尔特法测定，而抗氧化能力通过2,2-二苯基-1-苦基肼（DPPH）法、氧自由基吸收能力（ORAC）法和铁还原抗氧化能力（FRAP）法测定。通过台式和便携式拉曼光谱仪从36个不同颜色（深色、琥珀色、浅色）的枫糖浆样品中总共收集了360个光谱。这些光谱用于建立评估枫糖浆抗氧化特性的预测模型。与便携式拉曼光谱一起开发的深度学习模型表现出与与台式拉曼光谱一起开发的模型相当的预测性能。基于使用便携式拉曼光谱收集的光谱数据集，开发的深度学习模型表现出低均方根误差（RMSE，平均参考值的7.2 - 17.9%）、低平均绝对误差（MAE，平均参考值的5.2 - 13.1%）和高R值（>0.88）。结果表明，对于预测枫糖浆的抗氧化特性具有很好的拟合优度和准确性，表明使用便携式拉曼光谱仪对枫糖浆抗氧化特性进行现场分析的潜力。
