30/09/2026
Reading bovine mastitis pathogens off an agar plate has always leaned on years of hands-on experience — the kind that takes a career to build and doesn't transfer easily between vets, between clinics, or between generations of staff. That experience gap is becoming a bigger deal than it used to be. Bovine mastitis is one of the largest sources of production loss in dairy farming, and getting the causative pathogen right, quickly, matters directly for how antibiotics get used on farm — which makes it a real piece of the "One Health" approach to antimicrobial resistance (AMR). A 2015 G7/OECD-cited review warned that, left unaddressed, AMR could contribute to 10 million deaths a year globally by 2050 — a big part of why appropriate, evidence-based antibiotic use in livestock is being treated as a public-health issue, not just a farming one.
Azabu University and CarbConnect® have started a joint pilot to see whether AI can help close that experience gap. The idea is simple to describe and harder to execute well: a vet or technician photographs bacterial colonies growing on a culture plate with an ordinary smartphone, sends the image to CarbConnect®'s life-science AI platform, and gets back a candidate identification spanning 16 bacterial genera associated with mastitis — without needing the pattern-recognition instincts that normally take years on the bench to build.
The division of labor is clear: Azabu University is leading study design, strain and image collection across the 16 target genera, and the clinical-vet evaluation in the field; CarbConnect® is handling the imaging-condition standardization, training-data curation, and the identification model itself. The project runs across fiscal years 2026 through 2028, under a JRA (Japan Racing Association) livestock-promotion program aimed at appropriate mastitis diagnosis and prudent antibiotic use, with Professor Kazuhiro Kawai of Azabu University's Faculty of Veterinary Medicine serving as lead investigator.
Worth being precise about where this stands today: this is a research collaboration and clinical implementation study, not a released CarbConnect® product feature. Any identification produced during the pilot is a candidate suggestion, not a diagnosis — a qualified veterinarian's judgment remains the deciding factor. If the pilot holds up through its three-year run, the stated long-term goal is broader than specialist microbiologists alone: reaching veterinary technicians and field staff at clinics and testing labs who don't have decades of plate-reading experience behind them, and strengthening testing capacity at the point of care.
It's also a useful example of the direction CarbConnect® is heading as a platform: rather than one model built for one narrow task, the goal is domain-specific research collaborations — like this one — plugging into shared, reusable AI infrastructure for life-science imaging. The full announcement, including the public-health framing around AMR and the study's scope, is linked below.
Read the full announcement: https://buff.ly/733PelD
