An AI-enabled digital stethoscope designed for human use performed poorly at detecting heart murmurs in cats, according to research published in the Journal of the American Veterinary Medical Association (JAVMA).
The device detected only two of 22 feline murmurs identified by veterinary clinicians, giving it a sensitivity of 9.1%. The researchers say the findings highlight the importance of validating AI-assisted diagnostic tools specifically for the species in which they are being used.
What did the study find?
Researchers at North Carolina State University prospectively enrolled 54 dogs and 51 cats presenting to a university teaching hospital between August and December 2025. Each animal was assessed at four thoracic sites using a Core 500 AI-enabled digital stethoscope, alongside examination by a board-certified veterinary cardiologist, cardiology resident and fourth-year veterinary student. Six-lead ECG was also performed, with echocardiography carried out when clinically indicated.
The device and its accompanying application are designed for human use. Its AI provides real-time interpretations of whether a murmur is present and of the patient's cardiac rhythm.
Performance differed markedly between dogs and cats.
Veterinary clinicians identified murmurs in 22 of the 51 cats. The AI detected just two, missing the remaining 20 and producing a sensitivity of 9.1%. Fourth-year veterinary students performed substantially better, identifying murmurs in nearly two-thirds of affected cats.
The two feline murmurs that the device did detect were particularly notable. They occurred in the cats with the lowest heart rates in the feline cohort, at 120 and 155 beats per minute, and were high-grade murmurs graded 4 and 5. The researchers said this suggests the device may be more likely to identify the most acoustically prominent feline murmurs.
Based on these findings, the authors caution against using the AI interpretation as a sole screening tool for murmurs in cats without further feline-specific validation. They also noted that the larger footprint of the digital stethoscope compared with a traditional stethoscope may make positioning more difficult in smaller patients.
The results were different in dogs
The device performed considerably better at detecting canine murmurs.
Clinicians identified murmurs in 38 of the 54 dogs, of which the AI correctly detected 33. This produced a sensitivity of 86.8%, although specificity was lower at 56.3%. Its performance was identical to that of the fourth-year veterinary students in the study.
Murmur grade was the only significant predictor of whether the AI would identify a canine murmur. Dogs with murmurs graded 3 or above had around 15 times greater odds of having their murmur detected by the device than dogs with lower-grade murmurs.
The difference between the canine and feline results is particularly relevant because, although the physical stethoscope can be used across species, the diagnostic AI was trained using human rather than canine or feline data.
Arrhythmia classification also raised concerns
The researchers also found limitations in the device's AI-assisted rhythm interpretation.
In dogs, the stethoscope did not classify a single patient as having no arrhythmia. It reported atrial fibrillation in 28 dogs, an unclassified arrhythmia in 21 and a poor signal in five. Although it correctly identified all six genuine cases of atrial fibrillation, it also classified dogs with other rhythms as having the condition.
Sinus arrhythmia was a particular source of error: nine of the 13 dogs with sinus arrhythmia were classified by the device as having atrial fibrillation. Overall agreement between the AI rhythm classification and the clinicians' diagnosis was low.
There was only one cat with an arrhythmia in the study, limiting the conclusions that can be drawn about rhythm classification in feline patients.
Importantly, the researchers distinguished between the device's ability to collect information and its AI interpretation of that information. First author and cardiology resident Jake Johnson said the ECG recording itself was of good quality, adding: “This tool works best as an adjunct – a quick ECG or a flag worth a second look.”
Why does this matter for vets and nurses?
Digital stethoscopes can offer useful functions including recording heart sounds and ECG traces. But this study shows why the AI interpretation attached to a device should not automatically be assumed to perform equally well across species.
For veterinary teams, the key points are:
- Do not rely on this AI system to rule out a feline heart murmur. It missed 20 of 22 murmurs in the study.
- Separate the device from its interpretation. Useful digital recording and ECG capabilities do not necessarily mean the diagnostic algorithm is reliable.
- Consider what an AI system was trained on. Algorithms developed from human data may need species-specific validation before being used to guide veterinary decisions.
- Treat AI as an adjunct to clinical judgement. The findings support using these tools alongside, rather than instead of, trained veterinary assessment.
The authors also stress that the findings apply specifically to the technology tested in this study and should not be generalised to every AI-enabled stethoscope. Other systems may use different algorithms and have different diagnostic performance.
Source: Johnson JH, Stern JA, DeFrancesco TC, Pierce KV. An artificial intelligence–enabled digital stethoscope demonstrates moderate murmur detection in dogs but not cats and unreliable arrhythmia classification in both species. Journal of the American Veterinary Medical Association. Published online 5 August 2026.
