Assessing the Efficacy of Various Machine Learning Algorithms in Predicting Blood Pressure Using Pulse Transit Time.

Area of Science:

• Biomedical Engineering
• Cardiovascular Physiology
• Machine Learning in Healthcare

Background:

• Non-invasive and cuffless blood pressure monitoring is crucial for continuous health assessment.
• Traditional methods are often cumbersome, limiting real-time application.
• Pulse Transit Time (PTT) offers a potential solution using physiological signals.

Purpose of the Study:

• To investigate the efficacy of PTT derived from IPG, PPG, and ECG for non-invasive blood pressure monitoring.
• To evaluate machine learning models for predicting blood pressure categories using PTT and cardiovascular features.
• To assess the feasibility of integrating these techniques into wearable health devices.

Main Methods:

• Collected synchronized data from 100 healthy participants using custom IPG, ECG, PPG, and blood pressure devices.
• Applied machine learning models (Random Forest, Logistic Regression, SVC, K-Neighbors) to predict blood pressure categories.
• Utilized PTT and other cardiovascular features as predictors in the machine learning models.

Main Results:

• Random Forest model achieved 90% overall accuracy in blood pressure classification.
• The model demonstrated robustness with a 95% CI for accuracy (80%-95%), handling class imbalance effectively.
• PTT derived from PPG was identified as a critical predictive feature.

Conclusions:

• PTT measurements from PPG, IPG, and ECG are effective predictors for non-invasive blood pressure monitoring.
• These findings support the integration of PTT-based techniques into wearable devices for continuous monitoring.
• This advancement offers a significant step towards cuffless, non-invasive blood pressure assessment.