Researchers at the University of Cambridge have developed an advanced machine learning algorithm that can accurately detect heart murmurs in dogs.
This breakthrough technology could help veterinarians quickly identify heart disease in dogs, particularly in small breeds like King Charles Spaniels, which are more prone to heart conditions.
Heart murmurs are one of the main signs of mitral valve disease, a common heart problem in adult dogs.
In fact, around one in 30 dogs seen by a vet has a heart murmur, and this number is even higher for smaller breeds and older dogs. Early detection is crucial, as timely medication can extend a dog’s life and improve its quality of life.
The research team, led by Dr. Andrew McDonald from Cambridge’s Department of Engineering, adapted an algorithm initially created to detect heart murmurs in humans.
They used audio recordings from digital stethoscopes and found that the algorithm could detect heart murmurs in dogs with 90% accuracy—comparable to experienced cardiologists.
“Mammalian hearts are quite similar, and when things go wrong, they do so in similar ways,” explained Professor Anurag Agarwal, a specialist in acoustics and bioengineering. The team began with heart sounds from around 1,000 human patients and then modified the algorithm to analyze heart sounds from dogs.
To train and test the algorithm, the researchers collected data from nearly 800 dogs undergoing heart exams at four veterinary centers in the UK.
The dogs had their heart sounds recorded using an electronic stethoscope, and all of them received full physical exams and heart scans (echocardiograms) by a cardiologist. This dataset is now the largest collection of dog heart sounds ever created.
The algorithm was fine-tuned to not only detect heart murmurs but also to grade their severity. This is crucial because mild murmurs might not need immediate treatment, while more severe ones could indicate advanced heart disease that requires further care.
“Grading a heart murmur and deciding on treatment takes a lot of skill and experience,” said Professor Jose Novo Matos from Cambridge’s Department of Veterinary Medicine. “We aim to give general vets a tool to detect heart disease and assess how serious it is, helping dog owners make the best choices for their pets.”
In tests, the algorithm’s assessments closely matched those of experienced cardiologists in more than half of the cases. In 90% of cases, its grading was within one grade of the cardiologist’s evaluation, a promising result given the variability in how different vets rate heart murmurs.
Dr. McDonald emphasized that this AI tool could benefit all veterinarians, regardless of their stethoscope skills. “We’ve automated the process of grading heart murmurs, which is vital for deciding the next steps and treatments,” he said.
For dogs with valve disease, effective medications are available, unlike in humans where surgery is often the only option. Professor Agarwal highlighted the importance of early medication to provide dogs with the best quality of life.
Dr. Novo Matos added, “AI isn’t a threat—it’s a tool to make us better at what we do. There aren’t enough specialists to scan every dog, but with tools like this, we can quickly identify the dogs that need help the most.”
