D-Amino Acids and Computer Vision Image Analysis: A New Tool to Monitor Hazelnuts Roasting?

Abstract

The roasting process allows complex chemical reactions in foods leading the formation of the Maillard-related compounds, which are of fundamental importance for colour and aroma development in hazelnuts (Corylus avellana L.). Colour is a charming characteristic of roasted foods; despite of their fundamental role in organoleptic properties, the roasting allows accumulation of some potentially toxic technological contaminants, as showed in other matrices (e.g. acrylamide, furan and D-amino acids). Free and protein-bound D-amino acids (D-aa) in foods can origin by technological processing (their formation is pH, time, and temperature-dependent), and also from bacterial racemases in fermented foods, as recently showed in cocoa ( FRIEDMAN 1999; CALIGIANI et al. 2007) So, these compounds can be used as thermal and bacterial markers. Racemisation impairs digestibility/nutritional quality of foods; some D-aa acids are both beneficial and deleterious (DA-WEN SUN 2008). The L*a*b* (international standard for colour measurements adopted by the Commission Internationale d'Eclairage, CIE, 1976) and Red Green Blue (RGB) systems are the most employed methods for colour analysis. Computer Vision Image Analysis (CVIA) coupled with Artificial Intelligence (AI) and related tools (Principal Component Analysis, Artificial Neural Networks, Self Organising Maps) should be considered a powerful approach for evaluating colour development in foods (PATZOLD & BRUCKNER 2006). First aim of our study was to monitor via GC-MS the D-aa formation in hazelnuts during different conditions of roasting. Second aim was to develop a CVIA/AI method to study the colour development in roasted hazelnut, considering the correlation between colour and D-aa too. The racemisation of L-aa (particularly D-aspartic acid plus D-asparagine, D-asx) in hazelnut was time-temperature dependent. Protein-bound D-asx (expressed as D/(L+D); 1.49%) and free D-ala (5.87%) were predominant in Infra-Red roasted samples. Hot-air flow process leads the formation of free-D-aa (D-ala: 6.77%; D-asx: 3.47%; D-glu: 4.50%). Using Back Propagation Neural Network on the Discrete Fourier Transform output (whole surface colour), the identification of hazelnuts samples roasted in different condition was achieved. Red colour decreased according to the correspondent increase of D-amino acids content. Concluding, CVIA/AI is a promising complex tool useful to simplify the comprehension of a complex model system, as colour development in roasted hazelnuts. Keywor

Keywords

• haselnut
• roasting
• computer vision image analysis
• d-amino acids