Robust Feature Selection for Continuous BP Estimation in Multiple Populations: Towards Cuffless Ambulatory BP Monitoring

• DOI: 10.36227/techrxiv.24112650.v1
• Publication Date: 2023-09-11T16:07:50Z
• Authors: Ana Cisnal; Yanke Li; Bertram Fuchs; Mehdi Ejtehadi; Robert Riener; Diego Paez Granados
• Keywords: Photoplethysmogram

<p>Current blood pressure (BP) estimation methods have not achieved an accurate and adaptable approach for application in populations at risk of cardiovascular disease, with generally limited sample sizes. Here, we introduce algorithm BP solely reliant on photoplethysmography (PPG) signals demographic features. Our automatically obtains signal features employs the Markov Blanket (MB) feature selection to discern informative transmissible features, achieving a robust space population shift.</p> <p>We validated our Aurora-BP database, compromising ambulatory wearable cuffless measurements over 500 individuals. By evaluating several machine-learning regression methods, Gradient Boosting emerged as most effective. The comparative assessment encompassed both generic model (trained unclassified data) specialized models (tailored each distinct population), former demonstrating consistent superiority MAE 10.2 mmHg (0.28) systolic 6.7 (0.18) diastolic whole dataset.</p> <p>Moreover, comparison in-clinic performance showed significant decrease accuracy latter 2.85 (p < 0.0001, F-value = 32764.76) 2.82 65675.36) errors.</p> <p>Our work contributes resilient from PPG signals, underscoring advantages causal quantifying disparities between measurements.</p>