Feasibility of the direct application of near-infrared reflectance spectroscopy on intact chicken breasts to predict meat color and physical traits

Physical and color characteristics of chicken meat were investigated on 193 animals by directly applying a fiberoptic probe to the breast muscle and using the visible-near-infrared (NIR) spectral range from 350 to 1,800 nm. Data on pH was recorded 48 h postmortem (pH); lightness (L*), redness (a*), and yellowness (b*) 48 h postmortem; thawing and cooking losses and shear force after freezing. Partial least squares regressions were performed using untreated data, raw absorbance data (log(1/R)), and multiplicative scatter correction plus first or second derivative spectra. Models were validated using full cross-validation, and their predictive ability was determined by root mean square error of cross-validation (RMSECV) and correlation coefficient of cross-validation (r(cv)). Means (+/- SD) of pH, L*, a*, b*, thawing loss, cooking loss, and shear force were 5.83 +/- 0.13, 44.54 +/- 2.42, -1.90 +/- 0.62, 3.21 +/- 3.28, 4.84 +/- 2.44%, 19.39 +/- 2.95%, and 16.08 +/- 3.83 N, respectively. The best prediction models were developed using log(1/R) spectra for b* (r(cv) = 0.93; RMSECV = 1.16) and a* (r(cv) = 0.88; RMSECV = 0.29), while a medium predictive ability of NIR was obtained for pH, L*, and thawing and cooking losses (r(cv) from 0.69 to 0.76; RMSECV from 0.01 to 1.73). Finally, predicted model for shear force (r(cv) = 0.41; RMSECV = 3.18) was unsatisfactory. Results suggest that NIR is a feasible technique for the assessment of several quality traits of intact breast muscle.