A Stratified Model for Enhanced Cuffless Blood Pressure Estimation Using Hypertension History

Area of Science:

• Biomedical Engineering
• Cardiovascular Research
• Wearable Technology

Background:

• Clinical characteristics influence blood pressure (BP) estimation accuracy in wearable devices.
• Long-term cardiovascular changes affect BP estimation models.
• Existing models may not account for patient heterogeneity.

Purpose of the Study:

• To investigate the impact of clinical characteristics on BP estimation using ECG and PPG signals.
• To develop and evaluate stratified BP estimation models for improved accuracy.
• To enhance the reliability of cuffless BP monitoring systems.

Main Methods:

• Applied four clustering algorithms to segment the dataset based on clinical characteristics.
• Identified hypertension history as a key factor for patient stratification.
• Developed and compared stratified BP estimation models against baseline models.

Main Results:

• Hypertension history was the primary determinant for patient grouping.
• Stratified models significantly improved systolic BP (SBP) and diastolic BP (DBP) estimation by 36% and 33%, respectively.
• Enhanced accuracy was achieved even with simpler machine learning models.

Conclusions:

• Patient stratification based on clinical characteristics, particularly hypertension history, is crucial for accurate BP estimation.
• Stratified models offer a significant improvement over baseline models for cuffless BP monitoring.
• This approach advances the development of reliable wearable BP monitoring solutions.