Acoustic Analysis of Heart Sounds to Estimate Blood Pressure

Kabir, Muammar (1,2), Cao, Xiaoshu (1,2), Saha, Shumit (1,2), Gavrilovic, Bojan (2), Yadollahi, Azadeh (1,2)

1: Institute of Biomaterials and Biomedical Engineering, University of Toronto

2: KITE, Toronto Rehabilitation Institute, University Health Network

Background: Cardiovascular disorders, such as hypertension, are the second leading causes of death in Canada, claiming more than 33,600 lives per year. Diagnosis of cardiovascular disorders requires accurate monitoring of blood pressure (BP). Even a small drop of 5mmHg in BP can decrease the risk of cardiovascular diseases such as stroke by 41%. The gold standard method to assess BP requires intra-arterial catheters, which is invasive. Non-invasive BP measurement using upper arm cuff is inaccurate in patients with increased upper arm size. Analysis of heart sounds, which are generated by the closure of heart valves, can be a promising method to estimate BP. Accordingly, the aim of this study was to develop an accurate and convenient technology for monitoring BP.
Methods: Healthy participants were recruited for this pilot study conducted at SleepdB laboratory of Toronto Rehabilitation Institute. The participants were seated for 15 minutes while devices including electrocardiogram electrodes and Finapres® (to continuously monitor BP) were attached. In addition, a small wearable device (The Patch) developed by our team was attached over the suprasternal notch to record tracheal respiratory and heart sounds. The participants were asked to perform 15 Mueller Maneuvers (breathing against an occluded airway) of 10 second duration each. All data were collected simultaneously using Biopac acquisition system. An automated algorithm was developed to localize heart sounds. Subsequently the intensity of heart sounds (HSI) was calculated as the area under the curve of the localized heart sounds’ envelope. For each subject, Pearson’s or Spearman’s correlation was used to investigate the correlation between relative changes in systolic/diastolic BP and HSI.
Results: Data from 6 subjects (3 females), age of 33.5±5.1 years and BMI of 21.6±3.2 kg/m2 were investigated. Significant positive correlation between relative changes in systolic BP and HSI was observed in all subjects (r>0.6, p<0.001, for all). In addition, the changes in HSI was significantly correlated with the diastolic BP changes (r>0.45, p<0.001, for all).
Conclusion: Our results indicate that changes in HSI are significantly correlated with the relative changes in BP. After developing appropriate model with additional heart sound features and validation in a larger dataset, heart sound analysis can be used for continuous monitoring of BP and detection of early signs of hypertension.