Session: 06-02-01: Vibration and Acoustics in Biomedical Applications

Paper Number: 145486

145486 - Feasibility of Left Ventricular Function Assessment via Precordial Vibrations in Heart Failure Patients 

Heart failure (HF) is a prevalent and potentially life-threatening condition characterized by the heart's inability to pump blood effectively. Managing HF requires careful monitoring of left ventricular function (LVF), as changes in LVF can indicate worsening heart function or acute exacerbations, necessitating timely interventions to prevent adverse outcomes. Traditionally, assessing LVF has relied on methods such as echocardiography, cardiac MRI, or invasive procedures, which are costly, time-consuming, and typically confined to clinical settings. This limitation poses challenges for continuous monitoring and early detection of HF progression, particularly in remote or home-based settings. A promising alternative lies in analyzing vibrations transmitted from the heart valves and myocardium through the thorax to the precordial region of the chest. These vibrations carry valuable information about cardiac function and can be captured using non-invasive, cost-effective sensors, offering the potential for continuous monitoring of LVF outside of clinical environments. In this study, we aimed to evaluate the feasibility of estimating left ventricular ejection fraction (LVEF), a key metric of LVF, from precordial vibration signals. To achieve this, data from 70 HF patients were collected and analyzed. The vibration signals were processed to extract various features within different frequency ranges, capturing unique signatures associated with cardiac function. Machine learning regression models were then trained using these features and cross validated to predict LVEF values. The performance of the models was assessed by comparing the predicted LVEF values with those obtained from echocardiography, considered the gold standard for LVF assessment. The accuracies were tested on 50% of data left aside from training the models. Among trained machine learning models, Gaussian process regression revealed a strong correlation coefficient of 0.93 between the predicted and echocardiography derived LVEF values, indicating the effectiveness of the approach in estimating LVF from precordial vibrations. Furthermore, the mean squared error of 1.82 demonstrated the accuracy of the predictions, suggesting the potential clinical utility of this method for monitoring LVF in HF patients. Importantly, variations in the precordial vibrations were observed to correlate with the severity of valvular dysfunction, highlighting the sensitivity of this approach to detect not only changes in LVF but also valvular abnormalities. This finding underscores the potential for comprehensive clinical assessments using precordial vibration analysis, offering valuable insights into both myocardial and valvular function in HF patients. In conclusion, our study demonstrates the promise of precordial vibration analysis as a non-invasive, cost-effective method for continuous monitoring of LVF in HF patients. Further research and validation are warranted to explore its clinical applicability and potential integration into routine HF management protocols, with the ultimate goal of improving patient outcomes and quality of life.

Presenting Author: Mohammad Ahmed Florida Institute of Technology

Presenting Author Biography: Mr. Ahmed is a PhD candidate from Florida Institute of Technology. His research focus is in the areas of physiological signal analysis, signal processing, wearable sensing and application of machine learning for developing innovative tools for disease diagnostic and monitoring.

Authors:

Mohammad Ahmed Florida Institute of Technology
Yasith Weerasinghe Florida Institute of Technology
Michael Grillo Florida Institute of Technology
Amirtaha Taebi Mississippi State University
Mehmet Kaya Florida Institute of Technology
Peshala Thibbotuwawa Gamage Florida Institute of Technology