Open Access

ARTICLE

Blood Pressure Estimation with Phonocardiogram on CNN-Based Approach

by Kasidit Kokkhunthod¹, Khomdet Phapatanaburi², Wongsathon Pathonsuwan¹, Talit Jumphoo¹, Patikorn Anchuen³, Porntip Nimkuntod⁴, Monthippa Uthansakul¹, Peerapong Uthansakul^1,*

1 School of Telecommunication Engineering, Suranaree University of Technology, Nakhon Ratchasima, 30000, Thailand
2 Telecommunications Engineering, Rajamangala University of Technology Isan, Nakhon Ratchasima, 30000, Thailand
3 Navaminda Kasatriyadhiraj Royal Air Force Academy, Saraburi, 18180, Thailand
4 Institute of Medicine, Suranaree University of Technology Hospital, Nakhon Ratchasima, 30000, Thailand

* Corresponding Author: Peerapong Uthansakul. Email:

Computers, Materials & Continua 2024, 79(2), 1775-1794. https://doi.org/10.32604/cmc.2024.049276

Received 02 January 2024; Accepted 21 March 2024; Issue published 15 May 2024

Abstract

Monitoring blood pressure is a critical aspect of safeguarding an individual’s health, as early detection of abnormal blood pressure levels facilitates timely medical intervention, ultimately leading to a reduction in mortality rates associated with cardiovascular diseases. Consequently, the development of a robust and continuous blood pressure monitoring system holds paramount significance. In the context of this research paper, we introduce an innovative deep learning regression model that harnesses phonocardiogram (PCG) data to achieve precise blood pressure estimation. Our novel approach incorporates a convolutional neural network (CNN)-based regression model, which not only enhances its adaptability to spatial variations but also empowers it to capture intricate patterns within the PCG signals. These advancements contribute significantly to the overall accuracy of blood pressure estimation. To substantiate the effectiveness of our proposed method, we meticulously gathered PCG signal data from 78 volunteers, adhering to the ethical guidelines of Suranaree University of Technology (Human Research Ethics number EC-65-78). Subsequently, we rigorously preprocessed the dataset to ensure its integrity. We further employed a K-fold cross-validation procedure for data division and alignment, combining the resulting datasets with a CNN for blood pressure estimation. The experimental results are highly promising, yielding a Mean Absolute Error (MAE) and standard deviation (STD) of approximately 10.69 ± 7.23 mmHg for systolic pressure and 6.89 ± 5.22 mmHg for diastolic pressure. Our study underscores the potential for precise blood pressure estimation, particularly using PCG signals, paving the way for a practical, non-invasive method with broad applicability in the healthcare domain. Early detection of abnormal blood pressure levels can facilitate timely medical interventions, ultimately reducing cardiovascular disease-related mortality rates.

---

Keywords

---