NONINVASIVE SHORT-WAVELENGTH NEAR-INFRARED SPECTROSCOPIC METHOD TO ESTIMATE THE CRUDE LIPID-CONTENT IN THE MUSCLE OF INTACT RAINBOW-TROUT

Short-wavelength near-infrared (SW-NIR) spectroscopy (700-1100 nm) was used to predict the crude lipid content of intact, whole rainbow trout (weight range 66.5-883 g). The measurements were performed noninvasively and in the diffuse reflectance mode by conveying the light to the intact, whole fish over an optical fiber bundle. The backscattered light was collected by a second fiber bundle, concentric to the first, and focused onto the entrance slit of the monochromator. Two multivariate calibration models, a multiple linear regression (MLR) model and a partial least-squares (PLS) model, were used to correlate chemical with spectral data. The best correlation between the laboratory values and predicted values from spectral measurements of crude lipid content was obtained at a site midway between the dorsal fin and the adipose fin above the lateral line. The PLS cross-validation model using three latent variables yielded a standard error of prediction of cross-validation [SEPCV = 2.27% (w/w), R = 0.81]. Comparable results were achieved with a MLR cross-validation model using three wavelengths (934, 957, and 845 nm) [SEPCV = 2.48% (w/w), R = 0.76]. These results suggest that either a full spectrum or a discrete wavelength SW-NIR method could be used to predict the crude lipid content of fish muscle in whole trout without scale damage or other injuries to the animal.