Cytometric fingerprinting and machine learning (CFML): A novel label-free, objective method for routine mastitis screening

Abstract Bovine mastitis costs the US dairy industry $2 billion, an average of $200 per cow annually. Mastitis is currently diagnosed based on macroscopic alteration of milk or with a somatic cells count (SCC), which are non-specific markers of infection. Cows that have milk samples with no macroscopic alteration (i.e. clots) with more than 200,000 SCC per mL are classified as experiencing subclinical mastitis. Here, we demonstrate a novel cytometric fingerprinting and machine learning (CFML) toolchain as a label-free, objective, high-throughput microbiological milk quality evaluation method for routine mastitis screening. Milk samples were collected from each quarter of the udder from paired 20 milking Holstein cows. Cytometric fingerprints were immediately obtained along with simultaneous pathological analysis. Cytometric fingerprints largely resembled SCC and unique somatic cytometric fingerprints were observed in response to bacterial pathogens distinct from algal and fungal. To demonstrate applications of machine learning in reducing human intervention in future on-farm automated mastitis screening systems, we trained multiple machine learning models on cytometric fingerprints. Tested classifiers were found to be efficient, scalable and robust in classifying specific pathogen, identifying the lactation stage and pathogen intensity with 99.27%, 100%, and 100% accuracies respectively. Our findings indicate that CFML is sensitive to milk samples from cows experiencing subclinical mastitis spanning distinct types and levels of infections. The use of CFML is hence recommended for rapid, high-throughput mastitis typing. This would assist in the use of data-driven monitoring approaches leading to proper and judicious use of antibiotics in animal agriculture.