An Ultra-Portable Vis-NIR Spectrometer with an Integrated Light Source for Quality Analysis of Food and Agricultural Produce

Abstract

There is a growing demand for on-site quality analysis of food and agricultural produce along with detection of adulterants for selecting quality produce. Traditional laboratory measurement techniques such as HPLC, GCMS and FTIR have been used as the gold standard for quality control through target chemical detection and quantification. However, such analyses are expensive and time consuming, requiring bulky instruments and usage of a plethora of chemicals, some even very toxic and hard to dispose. Recent developments in miniaturized optical-sensors have enabled the realization of ultra-portable devices that can incorporate chemometric based models for chemical identification and quantification. Chemometrics captures the signature chemical moieties of a compound through their spectrum, and together with the miniature optical sensors provides a potent platform for a rapid, non-destructive, and fairly accurate method for analyzing a wide variety of substances with portable point-of-use spectrometers. We design, fabricate and show one such handheld Vis-NIR spectrometer with an integrated light source, and present results of its performance with two widely used classes of chemometric techniques: classification and prediction. The proposed device (6.5cm x 2.5cm x 6.5cm) consists of a spectral sensor and an LED light source (wavelength 400-1000 nm) encased in a 3D printed thermoplastic holder. Figure 1 shows the complete spectrometer along with custom made sample holders containing powdered samples. Spectral data can be acquired from the device through either wired or wireless communication. Each spectrum can be viewed as a graphical image and also be exported in text format that is used for further data processing. Data analysis and model building were carried out using Mathematica (version 12) and Unscrambler X (version 10.4). Identification and classification of similar compounds is a key element in quality control. To demonstrate this, we analysed the reflectance spectra of four chemically different but optically similar culinary organic powders, namely, corn starch, calcium propionate (C6H10CaO4), icing sugar (C12H22O11) and white flour. All of them were white in colour with similar texture. The aim of this experiment was to classify these white powders from their reflectance spectra alone. Fifteen scans per sample were collected using the device and were subjected to multiplicative scatter correction as a pre-processing step. Figure 2(a) illustrates the results of the scores-plot using the principal component analysis (PCA) algorithm, wherein the four different colours represent the spectra (every point represents one spectrum) of the four powders. As it can be seen, all the spectra of a particular colour lie closely to each other forming a distinct cluster, implying they are similar to each other. Four different clusters represent the four powders, leading to the conclusion that the spectral measurement from this device along with PCA can successfully distinguish one type of substance from the other. Quantifying the active ingredient in medicinal plants is a direct way of analyzing their quality. Prediction with regression modelling has been widely used for such quantification. To demonstrate a prediction task, we have built a regression model to estimate the amount of pharmacologically active curcumin present in turmeric powders. The curcumin content was first accurately quantified by high performance liquid chromatography (HPLC), which served as the reference data. Reflectance spectra of five different varieties of turmeric powders (1-7% curcumin) were collected using the device. The correlation between the processed spectra and the curcumin content was examined by the partial least squared regression (PLSR) algorithm. The model gave a high coefficient of determination (r2 ) of 0.99 and a standard error of 0.17. This model was cross validated with test samples and the plot for the predicted versus actual values of curcumin content are shown in Figure 2(b). An ultra-portable spectrometer with an integrated light source for the acquisition of reflectance spectra and its subsequent analysis is thus demonstrated. The proposed device can help the user in taking faster and better decisions with on-site measurements, taking into account multiple parameters. Figure 1