Computer vision techniques on magnetic resonance images for the non-destructive classification and quality prediction of chicken breasts affected by the White-Striping myopathy

Abstract

Listen

Translate article

This study was designed to assess the capability of MRI-computer vision algorithms, as a non-destructive technique, to classify and predict quality characteristics of chicken breast affected by White-Striping (WS) myopathy. Samples showing moderate and severe degrees of the myopathy were analyzed together with normal samples (no WS symptoms). The influence of the computational algorithms to analyze the MRI images and the techniques of data analysis on the classification and prediction results was aimed. Computational features from both texture (GLCM) and fractal (OPFTA) algorithms were useful to i) classify WS chicken breast by means of different classification technique, Principal Component Analysis and Decision Tree, and ii) predict physico-chemical characteristics of these chicken breast with high accuracy, applying Multiple Linear Regression. The results show the feasibility of objectively classifying chicken breasts without sample destruction into two degrees of severity. This is of remarkable relevance in large processing plants where WS incidence is high and a quick decision-making is required for the fate of affected samples. • MRI-computer vision enabled classification of chicken breasts affected by myopathy. • Both texture (GLCM) and fractal (OPFTA) algorithms enabled such classification. • Multiple Linear Regression on MRI computational features allowed quality prediction.