Predicting blood pressure from physiological index data using the SVR algorithm.

Area of Science:

• Biomedical Engineering
• Machine Learning in Healthcare

Background:

• Hypertension and related diseases pose significant health risks.
• Traditional blood pressure monitoring methods are cumbersome and uncomfortable for continuous use.
• Wearable devices offer potential for continuous blood pressure measurement, but current machine learning prediction accuracy is insufficient.

Purpose of the Study:

• To develop an efficient blood pressure prediction method for continuous monitoring.
• To address the limitations of traditional blood pressure measurement techniques.
• To improve the accuracy of machine learning-based blood pressure prediction.

Main Methods:

• Proposed an efficient blood pressure prediction method utilizing the support vector machine regression (SVR) algorithm.
• Compared SVR performance against linear regression (LinearR) and back propagation neural network (BP).
• Evaluated models using accuracy, pass rate, mean absolute percentage error (MAPE), mean absolute error (MAE), R-squared (R²), and Spearman's rank correlation coefficient.

Main Results:

• The SVR model demonstrated accurate and effective blood pressure prediction capabilities.
• SVR outperformed classical machine learning algorithms in predicting blood pressure based on multiple evaluation metrics.
• Experimental results confirmed the efficacy of the proposed SVR method.

Conclusions:

• Machine learning techniques, particularly multi-feature joint training, can significantly enhance blood pressure measurement accuracy.
• Improved accuracy facilitates better disease classification and more precise clinical decision-making.
• The SVR approach offers a viable solution for continuous, accurate blood pressure monitoring.