| چکیده انگلیسی مقاله |
The combination of chemometrics and near-infrared spectroscopy (NIRS) has been previously used for the detection of food adulteration in many cases [1]. The major advantage of NIRS is that usually no sample preparation is needed; hence the analysis is simple, cheap, and very quick. In addition, it can be carried out online employing handheld devices [2]. However, the knowledge of multivariate data analysis and chemometrics is necessary to magnify the relevant information and lessen the undesirable information in the spectra obtained from NIRS. Since handheld NIRS enables rapid screening of adulteration in food samples, this study was designed to determine the feasibility of handheld NIRS in combination with the chemometrics approach in the detection of citric acid adulterated industrial lime juice samples. In the current study, NIR spectra for 24 genuine and 168 adulterated industrial lime juices were recorded in triplicate in the reflectance mode using a handheld NIRS (Tellspec, 700 to 1900 nm) connected to a smartphone. Adulterated samples were prepared by means of dilution of genuine samples with water at different levels (10, 15, 20, 25, 30, 35, and 40%) and subsequently, adjusting the Brix number by the addition of exogenous citric acid. Sample outliers were removed using principal component analysis (PCA). Evaluation of the PCA scores plot revealed that PCA analysis can provide an estimation of the nature of lime juice samples based on their NIR spectra. In addition, one class modeling approach was employed for the detection of adulterated samples using MATLAB software [3]. On this matter, soft-independent modelling of class analogy (SIMCA) model by multiplicative scatter correction (MSC) preprocessing resulted in 89% sensitivity in the prediction of genuine samples and 99% specificity in the prediction of adulterated samples with an overall accuracy of 94%. Our findings demonstrate the potential of chemometrics in combination with portable NIRS for monitoring lime juice quality in terms of genuine or adulterated nature. However, further investigations using classification models (e.g., partial least squares-discriminat analysis, PLS-DA) and regression models (e.g., partial least squares regression, PLSR) are required to confirm the promising results of the current study. |
