Multi-model fusion of classifiers for blood pressure estimation

Area of Science:

• Biomedical Engineering
• Cardiovascular Health
• Signal Processing

Background:

• Prehypertension is a significant risk factor for cardiovascular disease.
• Early detection of prehypertension is crucial for timely intervention and prevention.
• Current methods for blood pressure monitoring can be invasive or inconvenient.

Purpose of the Study:

• To develop a non-invasive method for classifying blood pressure into healthy and prehypertension categories.
• To continuously estimate systolic and diastolic blood pressure values using photoplethysmograms.
• To enhance classification and regression accuracy through a multi-model fusion approach.

Main Methods:

• Photoplethysmogram (PPG) signal denoising using Discrete Cosine Transform (DCT).
• Extraction of time and frequency domain features from denoised PPG signals.
• Multi-model fusion of Support Vector Machine (SVM), Random Forest (RF), and K-Nearest Neighbors (KNN) classifiers for categorization.
• Support Vector Regression (SVR) for estimating systolic and diastolic blood pressure values.

Main Results:

• The proposed method accurately classifies blood pressure into healthy and prehypertension categories.
• Continuous estimation of both systolic and diastolic blood pressure values was achieved.
• The multi-model fusion approach demonstrated superior classification and regression accuracy compared to individual classifiers.

Conclusions:

• Photoplethysmogram-based analysis offers a promising non-invasive approach for prehypertension detection and blood pressure estimation.
• Multi-model fusion significantly improves the accuracy of blood pressure classification and regression.
• This method holds potential for convenient and early screening of prehypertension.