Over the last few years, artificial intelligence (AI) tools have become ubiquitous as users explore how they can use the technology to do everything from mimic the style of Shakespeare to solve complicated homework problems to book travel arrangements.
While some uses of AI are for fun or function, graduate students at the University of Wisconsin School of Veterinary Medicine (SVM) are leveraging the power of AI to advance both animal and human health. These graduate students also helped teach fellow DVM students to leverage the technology in their work and develop AI applications during the annual Artificial Intelligence for Veterinary Medicine and Life Sciences course in June. Their efforts come as the SVM hires two additional AI-focused faculty members as part of UW-Madison’s RISE initiative.
A ‘real-time’ decision-making tool
Marlee Henige (DVM’23), a PhD student in the Comparative Biomedical Sciences Graduate Program whose work is funded by a National Institutes of Health T32 training grant, fell in love with research during her fourth year as a veterinary medical student while enrolled in a pair of AI-focused classes taught by Doerte Doepfer (Department of Medical Sciences). The Middleton, Wisconsin, native aims to use the technology to allow veterinarians to make better-informed decisions in less time, conducting research to build a real-time tool to help address antimicrobial resistance, among other things.
Antimicrobials (antibiotics, antivirals, etc.) are “the cornerstone of modern medicine” and used to “treat infectious diseases in humans, animals and plants,” according to the World Health Organization (WHO). Over time, bacteria and viruses can mutate and become resistant to certain drugs, making the resulting infectious diseases more difficult to treat. A recent rise in antimicrobial resistance is one of the world’s top public health threats, the WHO says.
Henige’s research is focused on Digital dermatitis — also known as “hairy heel warts” — a contagious skin infection that is associated with lameness in beef and dairy cattle all over the world. The infections can cause painful lesions on the heels of animals’ feet.
Typically, clinicians prevent Digital dermatitis using disinfectants and treat individual cows’ lesions using, among other treatment agents, topical antibiotics. But they’re making an educated guess when deciding what drug to prescribe, Henige says. Ideally, clinicians will swab the infected skin and culture it to determine which type of bacteria is present. However, that takes time, and they often have to make a treatment decision before they have those results.
“I really want to improve decision making processes in real time,” Henige says. “Can you have any information to help you make a better-informed decision at the moment when the animal patient is seen?”
The veterinarian-turned-researcher thinks you can.
Doepfer’s lab has a collection of samples of Digital dermatitis-infected bovine skin that have been sequenced and turned into images to show their DNA. Traditional data analysis would compare the sequences to data banks of known sequences. The goal of Henige’s work is to shorten this process by using a computer vision model — an AI model that uses images to detect objects — that can analyze visualized DNA to show which antimicrobial resistant genes are present.
But it’s a little more complicated than just snapping a photo of a cow’s foot and inputting it into the model. Instead, veterinarians need to take a biopsy of the infected skin and then insert it into a hand-held, cellphone-sized DNA sequencer. Once the DNA of the microbes inhabiting the skin has been sequenced, you can feed the resulting sequences into the model and get a response. The model can help identify factors associated with the microbial resistance to certain drugs. The whole process should take less than 30 minutes, Henige says.
A bird’s eye look at cattle lameness
Fernando Valle (MSx’26) is using his time at the SVM to make a career pivot. Born in São Paulo, Brazil, Valle holds an undergraduate degree in computer science and previously worked with software companies building and designing web solutions.
“I was looking for something that challenged me,” he says, which led him to take summer courses at the Massachusetts Institute of Technology. He met Doepfer while on campus in Cambridge; she invited him to Madison to explore the academic world and develop AI-related projects.
After a year-long research internship working on AI models to help cows stay cool in warm weather without wasting water, Valle enrolled as a master’s student with the Comparative Biomedical Sciences Graduate Program in January 2025. His work in the Doepfer lab is funded by industry. Today, he’s working on a tool to help dairy farmers and cattle ranches identify animals with mobility issues, known as lameness, using AI-enhanced camera detection.
Using top view images of beef cattle walking, Valle is training an AI model to predict lameness based on hundreds of variables — including how fast an animal is walking, which direction it’s walking, and how certain body parts and its head move. To generate the variables, he assigns key landmarks to the bodies of the animals in images. The landmarks can then be tracked to help determine if an animal is experiencing lameness.
In total, the model evaluates more than 800 “features” (datapoints generated by the landmarks moving across a camera’s field of vision) to try to predict lameness. The tool Valle is building could be especially important for cattle feed yards and harvesting plants, he says. Early detection of lameness is important for animal welfare and for economic reasons. It also potentially enables farmers to address potential injury sources, initiate treatment, and prevent lameness in the future.
The potential health benefits for animals keeps Valle energized, he says.
“I come from business and software companies, where it is always about how to make more money,” Valle says. “Here, I get to do work that helps keep animals healthy and farmers happy.”
By Jack Kelly