Toward enhancement in prediction of Pseudomonas counts distribution in salmon fillets using NIR hyperspectral imaging

Pseudomonas spp. counts (PC) is a good indicator for spoilage evaluation, but its detection with traditional techniques is time-consuming, destructive and inefficient. The aim of this study was to evaluate the potential of near-infrared (NIR) hyperspectral imaging to predict PC distribution in salmon fillets. Full wavelength data were related to PC values by partial least square regression (PLSR), resulting in coefficient of determination (RP2) of 0.90 and root mean square error of prediction (RMSEP) of 0.52. Most effective wavelengths (MEW) were selected by regression coefficients (RC), successive projections algorithm (SPA) and competitive adaptive reweighted sampling (CARS) algorithm, respectively, to optimise the PLSR model. CARS-PLSR model built with ten MEWs of 941, 1105, 1161, 1178, 1222, 1242, 1359, 1366, 1628 and 1652 nm performed better with RP2 of 0.91 and RMSEP of 0.49. Colour maps were finally generated and PC distribution was visualised. NIR-HIS shows a great promise for evaluating PC distribution of salmon flesh during cold storage.