A NEW study shows that AI stethoscope analysis can effectively detect valvular heart disease, identifying severe cases with greater accuracy than general practitioners. This technology could transform early screening for valvular heart disease in UK clinical settings.
Challenges in Valvular Heart Disease Detection
Valvular heart disease is a growing public health concern, yet over half of cases remain undiagnosed due to late symptom onset, limited public awareness, and low sensitivity of traditional stethoscope screening. Existing AI tools rely on murmur detection, which often misses common subtypes like mitral regurgitation and performs poorly on small datasets. Accurate early detection remains critical to prevent complications and guide timely intervention.
Methods and Results: AI-Enhanced Auscultation in Multi-Centre Study
Researchers developed a recurrent neural network (RNN) trained to directly predict clinically significant VHD from heart sound recordings, using echocardiographic labels as the reference standard rather than murmur labels. Data were collected from 1,767 patients across UK primary care and hospital settings, encompassing standard auscultation sites. The algorithm was evaluated against general practitioner (GP) performance using sensitivity, specificity, and AUROC metrics.
The AI achieved an AUROC of 0.83, showing exceptional sensitivity for severe aortic stenosis (98%) and severe mitral regurgitation (94%). General practitioners in the same study had a sensitivity of 62% and specificity of 64%. The VHD Detector significantly outperformed GP predictions (p = 0.01 for sensitivity; p = 0.002 for specificity), demonstrating its potential as a scalable, low-cost screening tool for moderate or severe valvular heart disease.
Clinical Implications and Future Considerations
This AI-augmented stethoscope could provide a rapid, non-invasive VHD screening solution, improving early diagnosis and guiding timely echocardiographic referral. With one of the world’s largest phonocardiogram datasets, the model is generalisable across multiple clinical settings. Future work should explore integration into routine primary care workflows and evaluate long-term patient outcomes. Wider adoption could reduce undiagnosed valvular heart disease and optimise healthcare resource allocation.
Reference
McDonald A et al. Development and validation of AI-Enhanced auscultation for valvular heart disease screening through a multi-centre study. npj Cardiovasc Health. 2026;DOI:10.1038/s44325-026-00103-y.
