Using an image processing based technique and predictive models for assessing lipid oxidation in rainbow trout fillet

Abstract

The potential of image processing techniques in combination with multivariate regression analysis was studied for prediction of the lipid oxidation of stored trout fillets at 4 °C. Fillet images were acquired using an imaging system during the time intervals of 0, 5, 10, 15 and 20 days and the means of colorimetric parameters (L*, a*, b*, C* and h*) were obtained from the images. The multivariate regression models were built using a stepwise regression algorithm for prediction of the oxidative attributes. The models did well predicting the contents of FFA, PV and TBARS with the coefficient of determination (R2adjusted) of 0.66, 0.78 and 0.94, respectively. Among all measured colorimetric parameters, hue angle (h*) and lightness (L*) showed more significant contributions in the modeling analyses. Overall, image processing techniques in combination with predictive models may possibly be useful as a precise and non-destructive method for assessment of oxidation in rainbow trout.