Infrasound-based stethoscope for early detection of mitral regurgitation

Introduction: Accurate diagnosis and treatment of mitral regurgitation (MR) depend on the physician’s clinical expertise and timely referral for advanced imaging modalities. We hypothesized that a novel electronic stethoscope (Voqx™, Sanolla) with subsonic capabilities and an acoustic range of 0-2,000 Hz embedded with a machine-learning algorithm would improve physician’s ability to accurately diagnose mitral regurgitation.

Material and method: Using the Voqx™ stethoscope we recorded heart sounds from 87 patients referred for echocardiography, using four standard heart auscultation points (2nd right intercostal space, 2nd left intercostal space, left lower sternal border, and apex). Of them, 19 had moderate or severe MR, while 68 had no valvular disease. A machine-learning model was developed to construct a diagnostic algorithm. The data was divided into 70% learning data and 30% test data. This process was performed six times.

Results and discussion: The infrasound-based algorithm derived the best performance using the left lower border auscultation point. Infrasound-based machine learning algorithm had a sensitivity of 94%, specificity of 81%, and accuracy of 84% for the diagnosis of MR. A significant difference was found in the frequency domain between MR and a non-valvular disease (for 0-9 Hz P<0.05, for 40-75 Hz P<0.01). The width of the first heart sound, which was generated from 0-100 Hz, was significantly lower for the MR group (P<0.05).

Conclusion: We show preliminary but promising results for the detection of MR by using heart auscultation relying on infrasound information. This tool can be used for fast and easily available initial diagnosis of MR.