Rapid compositional analysis of Atlantic salmon (Salmo salar) using visible-near infrared reflectance spectroscopy

Rapid measurement of salmon flesh quality parameters (> 400 samples day^-1) was demonstrated in the laboratory and remotely at industrial sites. Visible-near infrared spectroscopy (VNIRS) was applied to predict astaxanthin (AX) and fat content in farmed Atlantic salmon. Fish were sampled from thirteen batches (1-6 kg whole weight, containing 2.3-16.3% fat and 1.2-12.5 μg g^-1 AX), and models validated on small (average ± SD: 1.4 ± 0.4 kg) and large fish (4.2 ± 0.9 kg). Both constituents were well predicted in minced Norwegian Quality Cutlet (NQC) samples (r² ≥ 0.86; standard error of prediction (SEP) ≤ 0.7% for fat and ≤ 0.7 μg g^-1 for AX). Comparable metrics were observed for AX prediction in whole NQCs (r² = 0.80-0.88; SEP 0.7 μg g^-1). Fat was better predicted in small fish than large fish for whole NQCs (r² = 0.82, SEP 1.0% cf r² = 0.59, SEP = 0.59%) and non-destructive scanning through the skin of whole, gutted fish (r² = 0.77, SEP = 1.2% cf r² = 0.49, SEP = 1.5%). Models were also developed for screening polyunsaturated fatty acid (PUFA) concentrations, e.g. in NQCs for docosahexaenoic acid (r² = 0.73) and n-3:n-6 PUFA (r² = 0.89).