Gluten Analysis Composition Using Nir Spectroscopy and Artificial Intelligence Techniques

Main Article Content

Oscar Jossa-Bastidas
Unai Sainz Lugarezaresti
Ainhoa Osa Sanchez
Guillermo Yedra Doria
Begonya Garcia-Zapirain

Abstract

Nowadays, food allergies are increasing among the population, affecting several health
problems. In Europe, up to 20% of people have allergies which are regarded as chronic
diseases. In this research, the food allergy and intolerance of gluten is analyzed. It is naturally
contained in grain seeds and frequently included in a variety of foods, including cereals like
wheat, barley, and other items. The NIR spectroscopy is a widely used approach in food
analysis, employed for measuring and testing different parameters depending on the sample
analyzed. In this research, artificial intelligence feature selection techniques are employed to
explore the most important wavelengths in the range of 900m - 1,700 nm in three types of flour
samples (rye, corn, and oats), with the objective of predicting the presence of gluten. A total of
12,053 flour samples were collected using the DLP NIR scan Nano EVM sensor and stored for
selecting the best features using the correlation Matrix and two feature selection methods:
Select K best, and random forest. The best accuracy obtained was 84.42% by selecting the best
range of wavelengths around 1,089 - 1,325nm and using the Random Forest classifier. The
proposed prototype system showed different advantages: ease of use, non-invasiveness, and
speed of prediction. The findings achieved are the first steps to developing an accurate real
time gluten prediction device for improving the quality of life of persons suffering from gluten
allergies.

Article Details

Section
Articles