Machine learning-based blood pressure estimation using impedance cardiography data.

Area of Science:

• Biomedical Engineering
• Cardiovascular Physiology
• Machine Learning

Background:

• Accurate blood pressure (BP) measurement is vital for cardiovascular disease management.
• Traditional cuff-based BP measurement methods can be inaccurate and uncomfortable.
• Impedance cardiography (ICG) offers a potential non-invasive alternative for BP estimation.

Purpose of the Study:

• To assess the feasibility of machine learning (ML) algorithms for estimating BP using ICG data.
• To compare the performance of various ML algorithms in BP estimation against standard cuff-based measurements.
• To evaluate BP estimation accuracy during both mental and physical stress conditions.

Main Methods:

• Analyzed ICG data from 71 healthy adults.
• Evaluated nine ML algorithms for BP estimation against oscillometric BP measurements.
• Optimized models to minimize root mean squared error and assessed performance using accuracy and regression metrics.
• Tested models during the Trier social stress test and a bike ergometer exercise.

Main Results:

• Multi-linear regression achieved high accuracy for systolic BP (r=0.82, MAE=8.20 mmHg).
• Support vector regression demonstrated the highest accuracy for diastolic BP (r=0.51, MAE=6.05 mmHg).
• Both models showed feasibility for cuff-less BP estimation.

Conclusions:

• Machine learning algorithms utilizing ICG data are feasible for estimating reference blood pressure.
• Further research is required to address biases, enhance performance, and standardize validation for clinical application.
• ICG-based ML offers a promising avenue for non-invasive BP monitoring.