Near infrared spectroscopy using short wavelengths and leave-one-cow-out cross-validation for quantification of somatic cells in milk

摘要

A short wavelength near infrared (NIR) analysis method for somatic cell count (SCC) determination in unhomogenised raw milk was developed. NIR spectra (700-1100nm) of unhomogenised milk samples from 14 cows during days 7 to 36 of their lactation were measured for two consecutive years. Measurements from animals with different physiological conditions (healthy and mastitic) were included. A "leave-one-cow-out" cross-validation partial least square (PLS) regression model for SCC in the raw milk was developed. Good correlation (R~2 = 0.76) was observed between estimated and reference SCC levels in milk samples when the differential spectra and SCC reference values were used for developing the PLS model. The NIR estimation achieved a standard error of 0.3 LogSCC. The results were well within the range needed for real-time monitoring of milk on a daily basis. Thus, the NIR method may prove to be a valuable tool for monitoring somatic cells in routine milk samples analysed for dairy cow health management.