Standardization of FT-MIR instruments for milk analysis

  • Dehareng, F. , Grelet, C. , Holroyd, S. , Vand, D.B.H. , Warnecke, S. , Fernndez Pierna, J.A. , Broutin, P. & Dardenne, P. (2017). Standardization of FT-MIR instruments for milk analysis. Bulletin of the IDF, 490-2017: 39-69.
Type Journal Article
Year 2017
Title Standardization of FT-MIR instruments for milk analysis
Journal Bulletin of the IDF
Volume 490-2017
Pages 39-69
Abstract Fourier-transform mid-infrared (FT-MIR) spectrometry is the most used method worldwide for compositional analysis and quality checks during routine liquid milk testing. The technique allows fast, non-destructive quantification of the (physico-)chemical properties of raw milk as an alternative to reference test methods, which are usually slow, difficult, expensive and time consuming. In 1961, a patent application was made for an FT-MIR method determining fat, protein and lactose in milk [1]. FT-MIR supplies information that is complementary to chemical information and provides high throughput with high sensitivity in a short response time from small sample volumes [2]. In 1993, the first purpose-built FT-MIR instrument was marketed, the Anadis MI-200 [3]. With the introduction of FT-MIR, equations based on the full range of spectral data have been developed to predict fat, protein, urea and lactose content of milks worldwide. The main components of milk can be linked with characteristic bands in the FT-MIR spectrum of milk because their chemical composition and chemical bonds result in absorbance of light at specific wavenumbers. However, MIR instruments are not stable over time because of physical wear and tear, maintenance operations and replacement of some subcomponents. Therefore, calibration to allow prediction of milk components requires adjustment in routine use for each instrument by slope and intercept correction. Recently, MIR prediction equations have been developed to predict new parameters related to milk composition (fatty acids, detailed protein composition, minerals), technological properties of the milk (ability to coagulate), metabolism of dairy cows (methane emissions, energy balance, energy intake, efficiency, ketosis) and the detection of adulteration. Some of these new parameters are indirect parameters. This means that the prediction equations are not directly related to a component present in the milk, but are the result of interrelation(s) with other known or unknown components, as marked by changes in the MIR spectra. To develop these new indirect equations, it is necessary to build a large spectral database with reference values for each of the properties to be correlated. Reference values should cover a large time period, wide geographical range, number of animals the milk sample originated from and other variables that could influence milk composition. The wider the variability that is captured in the spectral database, the greater will be the robustness of the equation. For a wider applicability of results, a collection of spectra originating from different laboratories, using different models or brands of FTMIR apparatus, in different countries and from different time periods should preferably be used. However, physical use, technical differences, maintenance operations and subcomponent replacement mean that even a single MIR instrument may not be stable over time. This results in spectra obtained from the same milk sample by two different instruments, or from the same instrument at two different times, having noticeable differences. A standardization protocol would help to reduce these differences. Standardization procedures are needed to correct spectra from different spectrometers on a reference basis and to bring all spectra towards a common spectral response in order to reach the following objectives: • Obtain stable predictions over time and correct deviations resulting from hardware modifications • Transfer equations between instruments of the same brand • Transfer equations between instruments of different brands • Create equations for parameters that are difficult or costly to measure by pooling reference analysis results and spectra (creation of a large spectral database) • Benefit qualitative approaches (untargeted approach, detection of abnormal spectra) The purpose of this paper is to give an overview of some existing approaches for spectral standardization.
Authors Dehareng, F., Grelet, C., Holroyd, S., Vand, D.B.H., Warnecke, S., Fernndez Pierna, J.A., Broutin, P., Dardenne, P.