AI-Enabled Stethoscope Misdiagnoses Heart Conditions in Pets
A study found that an AI-enabled stethoscope designed for universal use performed poorly at diagnosing heart conditions in pets because its algorithms were trained on human cardiac data. The device missed 91% of heart murmurs in cats and produced numerous false positives for arrhythmias in dogs, underperforming human veterinarians.
What happened
A study by North Carolina State University compared an AI-enabled digital stethoscope to human veterinary experts in diagnosing heart conditions in 105 companion animals (54 dogs and 51 cats).
The AI stethoscope's diagnostic algorithms were trained on human cardiac data rather than animal data. In dogs, the AI correctly identified 87% of murmurs but falsely labeled 22 additional dogs as having atrial fibrillation when they did not. In cats, the AI diagnosed murmurs in only 2 of 51 animals, missing 20 true murmurs, while a board-certified cardiologist and veterinary students significantly outperformed it.
Reported impact
Veterinarians have been adopting the AI-enabled stethoscopes for their technical features but were second-guessing themselves based on the stethoscope's findings. The poor performance in cats is described as clinically concerning because veterinarians rely on the tool for reassurance, yet it failed to detect real disease.
Index analysis
The AI stethoscope performed poorly at detecting arrhythmias in dogs, never correctly identifying any dog as free of abnormal heartbeat and producing 22 false positives for atrial fibrillation out of 30 dogs without the condition. The system showed especially poor performance in cats, where it missed 20 of 22 true murmurs, a 91% false negative rate for that species.
Governance controls
The study emphasizes the importance of human expertise when using AI-assisted diagnostic technology, particularly for instruments designed for universal use across different species.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.