Rapid and noncontact identification of soybean flour in edible insect using NIR spectral imager: A case study in Protaetia brevitarsis seulensis powder

Authors: HERNANDA, REZA - Chungbuk National University; KIM, JUNTAE - Chungnam National University; FAQEERZADA, MOHAMMAD - Chungnam National University; CHO, BYOUNG-KWAN - Chungnam National University; Kim, Moon; Baek, Insuck; LEE, HOONSOO - Chungbuk National University

Submitted to: Food Control
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/7/2024
Publication Date: 11/8/2024
Citation: Hernanda, R., Kim, J., Faqeerzada, M., Cho, B., Kim, M.S., Baek, I., Lee, H. 2024. Rapid and noncontact identification of soybean flour in edible insect using NIR spectral imager: A case study in Protaetia brevitarsis seulensis powder. Food Control. 169. Article e111019. https://doi.org/10.1016/j.foodcont.2024.111019.
DOI: https://doi.org/10.1016/j.foodcont.2024.111019

Interpretive Summary: Adulteration, the practice of intentionally mixing one substance with another to increase profits, is a significant food safety concern. Currently, there is no efficient and rapid method to inspect adulteration in edible insects such as Protaetia brevitarsis seulensis (PBS) larvae powder. In this research, partial least squares regression (PLSR) and one-dimensional convolutional neural network (1D-CNN) models were developed using hyperspectral reflectance images in short-wave infrared (SWIR) region to evaluate adulteration levels by measuring the quantitative difference of added soybean powder to PBS larvae powder. The 1D-CNN and PLSR models achieved adulterant detection accuracies of 0.994 and 0.991, respectively. These results demonstrate that the SWIR imaging method has a potential for use in food processing facilities to quantitatively detect adulteration levels in edible insects.

Technical Abstract: A high concern about food adulteration has continuously risen due to the negative contribution to human health, which may cause fatal issues. Some advanced chemical analyses, such as chromatography and polymerase chain reaction (PCR) have been intensively implemented for food authentication. Yet, these techniques have a few disadvantages, including destructive, time-consuming, and labor intensive. For this reason, a nondestructive and rapid measurement should be considered. In this study, adulterant quantitative analysis has been performed, in the case of Protaetia brevitarsis seulensis (PBS) larvae powder. To realize the adulteration, a set of artificial contaminated PBS powder was created by adding soybean powder from 0% to 50% concentration with 5% intervals. For the nondestructive measurement, a shortwave infrared hyperspectral imaging (SWIR-HSI, 1000–2100 nm) was utilized and optimized through regression analyses. Two types of regression analyses, namely partial least squares regression (PLSR) and one-dimensional convolutional neural network (1D-CNN) regression were implemented for quantitative measurements. This in-depth analysis revealed the superior performance of 1D-CNN indicated by an R2P of 0.994, RMSEP of 1.153%, and RPD of 12.916 than PLSR (R2P = 0.991, RMSEP = 1.435%, and RPD = 10.385) in predicting the concentration of PBS powder mixture. Thereto, an optimum regression model achieved by 1D-CNN demonstrates a good potential of SWIR-HSI in predicting soybean powder in PBS mixture nondestructively.