Multivariate Classification of UHT Milk as to the Presence of Lactose Using Benchtop and Portable NIR Spectrometers

Lactose hydrolyzed milk was developed in the 1970s to serve individuals with lactose intolerance. This demand for lactose-free products by lactose-intolerant consumers has created a market segment for this food whose quality control has to be guaranteed. In order to assess milk samples for lactose content, this work proposes an analytical methodology to classify regular and lactose-free ultra high temperature (UHT) milks using multivariate classification methods and NIR spectra obtained in FT-NIR and ultra-compact NIR spectrometers, aiming at field analysis. For this, 71 samples were purchased; 41 were lactose-free UHT milk and 30 regular UHT milk. Diffuse transflectance spectra were obtained by FT-NIR (833 to 2500 nm, 4 cm^?1 of resolution and mean of 16 scans), and by ultra-compact NIR (908 to 1676 nm, with 12.5 nm of resolution and mean of 50 scans). The classification models were obtained by PLS-DA and LDA techniques with robust variables selection by SPA and GA, evaluating different spectral pre-processing (MSC, SNV, and derivatives). The three models developed (PLS-DA, GA-LDA, and SPA-LDA) with benchtop equipment spectra correctly classified all samples with sensitivity and specificity of 100%. For the portable equipment spectra, PLS-DA and GA-LDA models obtained sensitivity and specificity of 100%. The SPA-LDA model, however, presented sensitivity and specificity of 80 and 100%, respectively. These results indicate that methodologies using NIR equipment, especially the ultra-compact NIR, with multivariate classification techniques are feasible in discrimination between regular and lactose-free milk in the field, thus enabling a quick and precise analysis.