Researchers from the University of Cambridge have created an innovative machine learning algorithm aimed at improving the detection of heart murmurs in dogs.
This development is particularly significant as heart murmurs can be early signs of cardiac disease, which is especially common in smaller dog breeds like King Charles Spaniels.
Algorithm Adaptation and Performance
By adapting an algorithm originally intended for human medical use, the team has successfully enabled it to identify and categorize heart murmurs in dogs by analyzing audio recordings obtained from digital stethoscopes.
When put to the test, the algorithm achieved an impressive sensitivity rate of 90%, closely matching the proficiency of experienced cardiologists.
Heart murmurs are vital indicators of mitral valve disease—the most prevalent heart condition found in adult dogs.
Presently, approximately one in every 30 dogs examined by vets is diagnosed with a heart murmur, with the likelihood increasing among older dogs and smaller breeds.
Research Methodology and Data Collection
Due to the widespread nature of mitral valve disease and related heart conditions, early detection is critical for effective treatment that can significantly enhance a dog’s quality of life.
The new algorithm from the Cambridge team offers a promising, cost-effective screening tool for veterinarians, potentially improving canine health outcomes across practices.
Their research has been documented in the Journal of Veterinary Internal Medicine.
Dr. Andrew McDonald, the lead researcher and part of the Cambridge engineering department, pointed out that heart disease in dogs represents a significant public health issue—one that may even overshadow similar concerns in humans.
Smaller breeds are particularly susceptible to heart conditions as they age, and since dogs cannot articulate their ailments, veterinarians play a crucial role in catching these diseases early.
Professor Anurag Agarwal, a leader in acoustics and bioengineering, noted that there is a noticeable absence of databases on canine heart sounds.
To fill this gap, the study initially utilized a human heart sound database, supporting the idea that animals share physiological characteristics and adverse heart functions may present similarly across species.
Future Implications in Veterinary Care
The research encompassed nearly 800 dogs undergoing standard heart assessments at four specialized veterinary centers throughout the UK.
Each dog received detailed physical examinations and echocardiograms from cardiologists, who then rated the heart murmurs and diagnosed any underlying cardiac issues.
This endeavor culminated in the largest collection of canine heart sound data ever assembled.
Professor Jose Novo Matos, a co-author specializing in small animal cardiology, emphasized the importance of using a diverse group of dogs—spanning various ages, sizes, and breeds—to enhance the algorithm’s performance.
A broader dataset will not only improve the algorithm’s accuracy but also increase its practical application for veterinarians and dog owners alike.
The team has fine-tuned the algorithm to not only detect heart murmurs but also to assess their severity based on audio recordings.
This allows for a differentiation between mild murmurs and advanced heart conditions that may require treatment.
Accurate grading of heart murmurs and determining the urgency of intervention typically necessitates considerable experience, along with specialist referrals and expensive scans.
The aim of this development is to equip general practitioners with the capability to identify and evaluate heart issues effectively, empowering them to make informed decisions for their clients.
Analysis of the algorithm showed that it agreed with cardiologists’ evaluations in over half of the assessed cases, and it matched expert assessments within one grading level in 90% of instances.
This is especially encouraging, as inconsistencies in grading heart murmurs among veterinarians are not uncommon.
Dr. McDonald explained that establishing the grade of a heart murmur is crucial for deciding on subsequent treatment options, a process that the new algorithm aims to simplify.
He believes this tool has tremendous potential to support veterinarians of all experience levels, making them more effective in their roles.
While humans often require surgical interventions for valve disease, dogs can benefit from effective medications.
Professor Agarwal underlined the importance of timing in initiating treatment, as it is essential for ensuring dogs maintain the best possible quality of life for as long as feasible.
In conclusion, Professor Novo Matos expressed enthusiasm for the role of artificial intelligence in veterinary care, viewing it as an ally rather than a competitor.
He acknowledged the challenges of screening every dog with a heart murmur due to time constraints and limited specialists, but he believes tools like this algorithm can significantly help veterinarians and pet owners pinpoint dogs that need immediate attention.
Source: ScienceDaily