Author ORCID Identifier

Date Available


Year of Publication


Degree Name

Master of Science in Biosystems and Agricultural Engineering (MSBiosyAgE)

Document Type

Master's Thesis


Agriculture; Engineering


Biosystems and Agricultural Engineering

First Advisor

Dr. Akinbode Adedeji


Cross-contamination between food grains during harvesting, transportation, and/or food processing is still a major issue in the food industry. Due to cross-contact with gluten-rich grains (wheat, barley, and rye grains), gluten can get into food that’s naturally free from gluten and thus may not be safe for consumption for people susceptible to gluten-related disorders such as celiac disease, wheat allergy, gluten intolerance or sensitivity. The conventional method of gluten detection is cumbersome, time-consuming, and requires well-trained personnel. Therefore, there is a need for a rapid and equally effective technique to authenticate gluten contamination in foods. This research work explored the use of a Fourier transform infrared (FTIR) spectroscopy coupled with machine learning approaches to detect and quantify gluten contamination in grain-based foods. The research was divided into three different phases including the use of FTIR with supervised machine learning (ML) approaches to authenticate cross-contact between non-gluten and gluten flours, the use of FTIR with ML approaches to detect and quantify wheat flour contamination in gluten-free bread (cornbread), and finally, the use of Enzyme-linked immunosorbent assay (ELISA) as a complementary test to estimate and establish a gluten-free threshold of ≤ 20 ppm for the amount of gluten in wheat contaminated flour and cornbread.

Different machine learning algorithms such as linear discriminant analysis (LDA), partial least square regression (PLSR), k-nearest neighbor (KNN), support vector machine, decision tree, and ensemble learning method were used for the development of ML models. The results obtained for the first phase of the research show that FTIR with LDA and PLSR has the potential to detect and quantify cross-contact between a non-gluten (corn flour, CF) and gluten-rich (wheat flour, WF, barley flour, BF, and rye flour, RF) flours, at contamination levels of 0.5% - 10% (w/w), with 0.5% increments. For the second phase, a majority voting-based ensemble learning (stack of random forest, k-nearest neighbor (KNN) and support vector classifier) model was able to detect WF contamination in a cornbread at the true-positive rate and false-negative rate of 1.0, respectively. The ELISA tests for both phases (the raw flour samples and the baked bread) showed a threshold limit of ≤ 0.5% contamination level for CF contaminated with WF to be labeled gluten-free and ≤ 3.5% for the cornbread contaminated with the WF to be gluten-free. This research is still in its development stage and has the potential to contribute towards artificial intelligence applications in ensuring food safety, and to food quality inspection.

Digital Object Identifier (DOI)