Raman hyperspectral imaging and spectral similarity analysis for quantitative detection of multiple adulterants in wheat flour

Authors: LOHUMI, SANTOSH - Chungnam National University; LEE, HOONSOO - Us Forest Service (FS); Kim, Moon; Qin, Jianwei - Tony Qin; CHO, BYOUNG-KWAN - Chungnam National University

Submitted to: Biosystems Engineering
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 3/19/2019
Publication Date: 3/28/2019
Citation: Lohumi, S., Lee, H., Kim, M.S., Qin, J., Cho, B. 2019. Raman hyperspectral imaging and spectral similarity analysis for quantitative detection of multiple adulterants in wheat flour. Biosystems Engineering. 181:103-113.
DOI: https://doi.org/10.1016/j.biosystemseng.2019.03.006

Interpretive Summary: Food safety incidents and public health concerns about food adulteration are driving the need to develop fast, sensitive, and reliable detection methods for food hazards and adulteration. Although previous studies have investigated the use of Raman microscopy imaging for quality and authenticity analysis of food products, line-scan Raman imaging has emerged only recently as a possible means of high sensitivity food analysis capable of screening samples of greater size and at higher speed. Here, we assess the application of line-scan Raman hyperspectral imaging (RHI) for simultaneous detection of three chemical adulterants mixed into wheat flour at concentration between 0.05% and 1.5 % w/w: benzoyl peroxide, alloxan monohydrate, and L-cysteine. A method called spectral angle mapping was applied to Raman hyperspectral images (first preprocessed to remove fluorescence baseline signals and mitigate signal noise) to visualize and identify adulterant pixels among the background wheat pixels, and pixel-based calculations of the adulterants’ presence in the images were found to agree with the target concentrations used in preparing the samples. The results show that Raman hyperspectral imaging with the application of spectral angle mapping can be an elegant tool for rapid and noninvasive quality and authenticity analysis of powdered foods to help producers and processors ensure the safety and quality of their products.

Technical Abstract: Recent food safety incidents and public health concerns related to food adulteration drive the need for fast, sensitive, and reliable methods for the detection of food hazards and adulteration. Although Raman microscopy imaging has been used for quality and authenticity analysis of food products previously, the application of line-scan Raman imaging has emerged only recently. Here, we assess the applicability of line-scan Raman hyperspectral imaging (RHI) for simultaneous detection of three potential chemical adulterants in wheat flour (0.05–1.5% w/w). RHI of wheat samples are collected (0.2-mm step size, 1 s exposure time) in an aluminum sample holder using a 785-nm line laser to generate Raman scattering. Reference spectra are obtained for all adulterants (benzoyl peroxide, alloxan monohydrate, and L-cysteine). Collected Raman data are preprocessed by adaptive iteratively reweighted penalized least square and median filter methods for fluorescence baseline removal and to mitigate high frequency noise, respectively. Spectral angle mapping (SAM) is applied to the preprocessed data to distinguish adulterants’ pixels from the flour background using the pure endmember as input extracted by independent component analysis. SAM images for each adulterant are converted to binary images to effectively visualize and quantitatively detect the adulterant pixels in wheat flour. The pixel-based calculated proportions of adulterants in wheat flour agree with the concentrations added. The applicability and reproducibility of the developed technique are assessed and the results demonstrate that RHI in combination with SAM provides a novel, elegant tool for rapid and noninvasive quality and authenticity analyses of powdered foods.