Objective: Accurate blood pressure (BP) measurement is crucial for the diagnosis, risk assessment, treatment decision-making, and monitoring of cardiovascular diseases. Unfortunately, the gold-standard, cuff-based BP measurement suffer from inaccuracies and discomfort. As a result, continuous, cuff-less, and non-invasive BP assessment methods are merging as promising alternatives. This study is the first to comprehensively assess impedance cardiography (ICG) for Machine Learning enabled BP measurement. Methods: We analysed ICG data from 71 young and healthy adults. Nine different Machine Learning algorithms were evaluated for their BP estimation performance. Accuracy measures comprised mean difference, standard deviation, Limits of Agreement, mean absolute error, and root mean squared error. Additionally, the B-Score was calculated to evaluate the “true”, relative performance of the best-performing model. Results: The Multi-Linear Regression model showed the highest performance (mean difference = -0.01, r = 0.82) for systolic BP. The Support Vector Regressor model achieved the best results (mean difference = 0.15, r = 0.51) for diastolic BP. Both models exhibited positive B-Scores, indicative of their predictive capability. All tested models’ estimations correlated with both systolic and diastolic reference BP (r > 0.72 / r > 0.34, p < 0.001). Conclusion: The study highlights the potential of ICG based Machine Learning algorithms for accurately estimating arterial BP. Such algorithms can provide accurate and continuous BP data. This offers the possibility of enhancing patient care, timely diagnostics, and treatment decision-making. Our approach is especially promising for post-surgical and intensive care environments.
Read full abstract