Innovative analyses of milk components, sensor and registry data for improved breeding, health, and reproduction in dairy cows.

In Norway, there is a well-established tradition of storing extensive data on cow health, fertility, and production for breeding purposes and for the monitoring of animal welfare and health. Presently, approximately sixty percent of domestic milk production originate from automatic milking systems (AMS) often termed "robots". This leads to significant digitization of individual animal information, which presents opportunities to incorporate artificial intelligence (AI) into health and welfare practices in milk production. The primary objective of the iKu project is to utilize herd data, AMS- and sensor information to create innovative indicators of health and fertility of Norwegian dairy cows. These indicators will subsequently be assessed for animal breeding. Within dairy herds, routine milk samples are collected monthly and analyzed using near-infrared spectroscopy (FTIR). We will use these spectra, information from the AMS and auxilliary sensors to compute energy balance and investigate correlations with disease and impaired fertility in dairy cows. The iKu project comprises a collaborative, interdisciplinary team of recognized Norwegian researchers. By combining expertise in animal and veterinary sciences, informatics and mathematical data analysis from NMBU, the project endeavors to develop farm-specific machine learning algorithms for monitoring health and fertility. This project perpetuates the tradition of utilizing health data in livestock breeding and health monitoring, incorporating innovative data analysis and artificial intelligence (AI) methodologies.

The iKu project aims to enhance the use of available data on dairy cows’ health and reproduction to establish improved phenotypes for genetic selection and novel bio-indicators for health and reproduction. Monthly routine milk samples analyzed by Fourier Transform Infrared (FTIR) spectroscopy will be used to predict undesirable fertility and health outcomes related to negative energy balance such as delayed onset of estrus after calving, subclinical ketosis, clinical endometritis, and laminitis. Also, pathogen (group) specific cases of mastitis will be predicted by FTIR spectral data and the database on milk culture results. Altogether five novel phenotypes will be subjected to genomic breeding value estimation using Geno’s high-density SNP chip genotypes. The predictability of new phenotypes and bioindicators will be enhanced using information from automatic milking systems (milking robot), sensors such as online cell counters (OCC), activity measurements, body weight scales, and data from herd records on treatment, udder pathogens, somatic cell counts, insemination, culling, production and more. New mathematical approaches for high-dimensional and longitudinal data will be developed for the detection of unfavorable fertility and health events in the single animal, for breeding purposes and for the management of the dairy herd. We will use fatty acid profiles to establish thresholds (benchmarks) aiding management decisions at the single animal and herd level. The project also aim to establish farm specific machine learning algorithms for surveillance of health and fertility. In the iKu project, a multidisciplinary team of recognized Norwegian researchers have joined to establish a platform for collection, analysis, and utilization of big data to enhance breeding and health management of Norwegian dairy cows. We expect the platform to expand as research is progressing, and funding for an additional PhD students has already been granted by NMBU.