Department of Applied chemistry, UPV/EHU, Paseo Manuel Lardizabal 3, 20018 San Sebastián, Spain.
Department of Applied chemistry, UPV/EHU, Paseo Manuel Lardizabal 3, 20018 San Sebastián, Spain.
Talanta. 2018 Jul 1;184:58-64. doi: 10.1016/j.talanta.2018.02.111. Epub 2018 Mar 1.
A method based on digital image is described to quantify tartrazine (E102), yellow, and allura red (E129) colorants in food samples. HPLC is the habitual method of reference used for colorant separation and quantification, but it is expensive, time-consuming and it uses solvents, sometimes toxic. By a flatbed scanner, which can be found in most laboratories, images of mixtures of colorants can be taken in microtitration plates. Only 400 µL of sample are necessary and up to 92 samples can be measured together in the same image acquisition. A simple-to-obtain color fingerprint is obtained by converting the original RGB image into other color spaces and individual PLS models are built for each colorant. In this study, root mean square errors of 3.3 and 3.0 for tartrazine and 1.1 and 1.2 for allura red have been obtained for cross-validation and external validation respectively. Results for repeatability and reproducibility are under 12%. These results are slightly worse but comparable to the ones obtained by HPLC. The applicability of both methodologies to real food samples has proven to give the same result, even in the presence of a high concentration of an interfering species, provided that this interference is included in the image analysis calibration model. Considering the colorant content found in most samples this should not be a problem though and, in consequence, the method could be extended to different food products. Values of LODs of 1.8 mg L and 0.6 mg L for tartrazine and allura red have been obtained by image analysis.
本文描述了一种基于数字图像的方法，用于定量检测食品样品中的食用色素柠檬黄（E102）、黄色和诱惑红（E129）。高效液相色谱法（HPLC）是常用的参考方法，用于色素分离和定量，但该方法昂贵、耗时且使用溶剂，有时具有毒性。通过大多数实验室都配备的平板扫描仪，可以对微滴定板中的色素混合物进行成像。只需要 400µL 的样品，并且可以在同一次图像采集中共测量 92 个样品。通过将原始 RGB 图像转换为其他颜色空间，可以获得简单易获取的颜色指纹，然后为每种色素分别建立 PLS 模型。在本研究中，交叉验证和外部验证分别得到了柠檬黄的均方根误差为 3.3 和 3.0，诱惑红的均方根误差为 1.1 和 1.2。重复性和再现性的结果低于 12%。这些结果略差于 HPLC 得到的结果，但具有可比性。两种方法在实际食品样品中的适用性证明可以得到相同的结果，即使存在高浓度的干扰物质，只要将这种干扰包含在图像分析校准模型中即可。考虑到大多数样品中的色素含量，这应该不是问题，因此，该方法可以扩展到不同的食品产品。通过图像分析，柠檬黄和诱惑红的检出限（LOD）分别为 1.8mg/L 和 0.6mg/L。
