Comparison between two statistical models for prediction of turkey breast meat colour

The aims of the present study were: (1) to determine the relevant objective measurements which could express visual assessment of turkey meat colour; and (2) to use these variables for the early prediction of the colour development of turkey breast meat. The colour of the meat was assessed subjectively by an expert at a processing plant at 24 hr post mortem, using a four-category scale (score a: light-pale meat, score b: light pink meat or normal meat, score c: dark meat, score d: very dark meat). Objective measurements included meat pH, temperature, dielectric loss factor, pigment concentration, L(∗) (lightness), a(∗) (redness) and b(∗) (yellowness) colour coordinates determined at different times post mortem. Colour coordinates and pH were chosen as relevant variables when measured at 1 and 4 hr post mortem and were used in prediction models. Linear analysis (canonical discriminant analysis) showed that the efficiency of prediction was 15%. A non-linear analysis (neural network) gave better prediction; the colour of the meat being correctly predicted for 70% of the muscles.