Prediction of pregnancy state from milk mid-infrared (MIR) spectroscopy in dairy cows

• Published in: Acta fytotechnica et zootechnica Vol. 23; no. Monothematic Issue; pp. 224 - 232
• Main Author: Rienesl, Lisa
• Format: Journal Article Web Resource
• Language: English
• Published: Slovak University of Agriculture in Nitra 01-12-2020
• Subjects: Animal production & animal husbandry; dairy cow; Genetics & genetic processes; Génétique & processus génétiques; Life sciences; mir spectroscopy; pls; pregnancy prediction; Productions animales & zootechnie; Sciences du vivant
• ISSN: 1336-9245; 1335-258X; 1336-9245
• DOI: 10.15414/afz.2020.23.mi-fpap.224-232

Pregnancy assessment is a very important tool for the reproductive management in efficient and profitable dairy farms. Nowadays, mid-infrared (MIR) spectroscopy is the method of choice in the routine milk recording system for quality control and to determine standard milk components. Since it is well known that there are changes in milk yield and composition during pregnancy, the aim of this study was to develop a discriminant model to predict the pregnancy state from routinely recorded MIR spectral data. The data for this study was from the Austrian milk recording system. Test day records of Fleckvieh, Brown Swiss and Holstein Friesian cows between 3 and 305 days of lactation were included in the study. As predictor variables, the first derivative of 212 selected MIR spectral wavenumbers were used. The data set contained roughly 400,000 records from around 40,000 cows and was randomly split into calibration and validation set by farm. Prediction was done with Partial Least Square Discriminant Analysis. Indicators of model fit were sensitivity, specificity, balanced accuracy and Area Under Receiver Operating Characteristic Curve (AUC). In a first approach, one discriminant model for all cows across the whole lactation and gestation lengths was applied. The sensitivity and specificity of this model in validation were 0.856 and 0.836, respectively. Splitting up the results for different lactation stages showed that the model was not able to predict pregnant cases before the third month of lactation and vice versa not able to predict non-pregnancy after the third month of lactation. Consequently, in the second approach a prediction model for each different (expected) pregnancy stage and lactation stage was developed. Balanced accuracies ranged from 0.523 to 0.918. Whether prediction accuracies from this study are sufficient to provide farmers with an additional tool for fertility management, it needs to be explored in discussions with farmers and breeding organizations.