When a veterinarian at North Carolina State University placed a digital stethoscope on a quiet cat named Mango last fall, the screen flashed a reading that seemed almost too simple: no murmur detected. But something told Dr. Jake Johnson, a cardiology resident and the study's first author, to look closer. "This device essentially couldn't find a murmur in cats," he later said. Mango, it turned out, had a real heart problem that the machine had completely missed.
Mango's story captures the key finding of a new study published in the Journal of the American Veterinary Medical Association: human veterinarians, trained over years to listen carefully to animal hearts, still outperform the AI inside modern digital stethoscopes when it comes to diagnosing heart disease in dogs and cats. The research, conducted at North Carolina State University's College of Veterinary Medicine between August and December 2025, examined 105 companion animals — 54 dogs and 51 cats.
Kursten Pierce, an assistant professor of clinical sciences and board-certified cardiologist at NC State, said the idea for the study grew from conversations with practicing veterinarians. "We were hearing from veterinarians who were second-guessing themselves based on the stethoscope's findings," she said. The reason? Those AI-enabled stethoscopes are built using data from human patients, not dogs or cats. "Veterinarians use the same stethoscope that a human physician uses," Pierce explained. "The diagnostic AI for these stethoscopes is trained on human data, not dog or cat data."
The results were striking. In dogs, the AI correctly identified 33 out of 38 murmurs — about 87 percent accuracy. Fourth-year veterinary students matched that score exactly. But where the AI truly struggled was in detecting irregular heartbeats, called arrhythmias. The device never once classified a single dog as having a normal heart rhythm, and while it correctly caught all six cases of atrial fibrillation, it falsely flagged 22 additional dogs as having the condition. That means three out of every four arrhythmia alerts in dogs were wrong.
Cats proved even more challenging. The AI diagnosed a murmur in only 2 of the 51 cats studied, missing 20 cats who actually had one. Fourth-year students, by contrast, correctly identified murmurs in nearly two-thirds of affected cats.
Johnson emphasized that the study is not a rejection of the technology itself. The stethoscope's built-in electrocardiogram — a device that tracks the heart's electrical activity — produced genuinely good quality tracings. "This tool works best as an adjunct — a quick ECG or a flag worth a second look — not a stand-alone diagnosis," Johnson said.
The researchers hope their findings will help veterinarians and veterinary students become more aware of what these devices can and cannot do, and to trust the expertise they have built over years of training. In a world increasingly guided by artificial intelligence, the beating heart of veterinary medicine remains distinctly human.