Human clinicians outperform AI stethoscope in diagnosing pet heart problems

Human clinicians outperformed an AI-enabled digital stethoscope in diagnosing heart problems in pets, according to a study conducted at North Carolina State University’s College of Veterinary Medicine. The research, which ran from August 1, 2025, to December 31, 2025, evaluated the effectiveness of the Core 500 stethoscope, developed by EKO Health Inc. This device connects to a smartphone app that provides real-time AI-assisted interpretations of heart sounds, including the presence of murmurs and irregular rhythms. The study included 105 companion animals, 54 dogs and 51 cats, examined by three types of professionals: a board-certified veterinary cardiologist, a cardiology resident, and a fourth-year veterinary student. Each participant underwent cardiac auscultation using the AI-enabled stethoscope. The results showed that human clinicians consistently outperformed the device in both detecting murmurs and identifying arrhythmias. In dogs, the cardiologist identified 38 (70%) with murmurs and 24 (44%) with arrhythmias. The AI stethoscope correctly detected murmurs in 33 of the 38 dogs, achieving an accuracy rate of 87%. However, its arrhythmia detection was problematic, failing to identify any dogs as free of arrhythmia and incorrectly labeling 22 dogs as having atrial fibrillation when they did not. Fourth-year veterinary students matched the stethoscope’s performance in detecting murmurs but performed similarly to the device in identifying arrhythmias. In cats, the situation was even more pronounced. The cardiologist found that 22 (43%) had murmurs, while only one (2%) exhibited an arrhythmia. The AI stethoscope failed to detect murmurs in 20 of the 22 affected cats, identifying murmurs in just two. Veterinary students, however, demonstrated significantly better performance, correctly identifying murmurs in nearly two-thirds of the affected cats. The study highlighted the limitations of AI systems trained primarily on human data, which may not be suitable for diagnosing conditions in non-human species. Kursten Pierce, an assistant professor of clinical sciences and board-certified cardiologist at NC State, emphasized that stethoscopes are universal tools used by both human physicians and veterinarians. While AI-enabled devices offer valuable features such as recording heart murmurs or generating electrocardiograms, their diagnostic accuracy can be compromised when applied to animals. She noted that some veterinarians expressed doubts about the reliability of the AI’s findings, prompting the need for this study. Jake Johnson, a cardiology resident at NC State and lead author of the research, described the results involving cats as “clinically concerning.” He pointed out that the AI stethoscope’s inability to detect murmurs in cats could lead to missed diagnoses, potentially allowing serious health issues to go unnoticed. The findings underscore the critical role of human expertise in interpreting AI-generated data, especially when the underlying training data does not align with the target population. The study raises broader questions about the generalizability of AI technologies designed for human use in veterinary medicine. As more veterinarians adopt AI-assisted tools, there is a growing need for validation studies tailored specifically to animal patients. Researchers suggest that future developments should focus on refining AI models to account for physiological differences between species, ensuring that these tools enhance rather than hinder clinical decision-making. The research has been published in the Journal of the American Veterinary Medical Association.