A groundbreaking study has revealed that a machine-learning algorithm, originally trained to detect heart murmurs in humans, can be successfully adapted for dogs to diagnose heart murmurs and stage preclinical myxomatous mitral valve disease (MMVD). This could be a game-changer in veterinary medicine, offering a reliable, cost-effective tool for early detection of cardiac issues in dogs without requiring a cardiologist’s expertise.

The Problem: Heart Murmur Detection in Dogs Heart murmurs are a crucial indicator of cardiac disease in dogs, particularly MMVD, which is one of the most common heart conditions in aging dogs. However, accurately grading these murmurs requires significant clinical experience, making early diagnosis difficult and often delayed in primary care settings.

The Study: AI-Powered Solution A team of researchers, led by Andrew McDonald and Jose Novo Matos, trained a recurrent neural network algorithm to detect and grade heart murmurs in dogs based on electronic stethoscope recordings. The study involved 756 dogs, both with and without cardiac disease, attending referral centers in the UK. The algorithm, initially designed for humans, was fine-tuned to predict the cardiologist’s assessment of the murmur grade in dogs.

Key Results: High Accuracy and Specificity The algorithm showed impressive results:

• Sensitivity: 87.9% in detecting heart murmurs of any grade.

• Specificity: 81.7% in ruling out murmurs.

• Grade Matching: It exactly matched the cardiologist's murmur grade in 57% of cases.

• MMVD Staging: The AI tool effectively differentiated between stage B1 and B2 of preclinical MMVD with an area under the curve (AUC) of 0.861.

This study demonstrates that AI can play a pivotal role in improving the detection of cardiac disease in dogs, particularly in primary care, where access to specialized cardiologists may be limited.

Conclusion: A New Frontier in Veterinary Diagnostics The successful adaptation of human-based AI models for veterinary use opens up new possibilities for early, accurate, and affordable screening of heart disease in dogs. With further development, this technology could soon become a vital tool in everyday veterinary practice, improving the health and longevity of our beloved canine companions.

This breakthrough in machine learning could revolutionize veterinary diagnostics, allowing vets to catch heart issues before they become critical and provide more timely, effective treatments.