Machine Learning Approaches for Blood Pressure Classification from Photoplethysmogram: A Comparative Analysis.
Journal:
Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual International Conference
PMID:
40040211
Abstract
The cuffless estimation of blood pressure (BP) has become a prominent area of research in recent years fueled by its potential clinical implications and the growing interest from the wearable device industry. It has been accelerated by the emergence of learning-based models. Most use multiple pulse signals such as ECG, PPG, etc. rely on pulse transit time estimates and related features for estimating BP. However, the accuracy and reliability of such methods are questionable and hence there has been limited clinical uptake. This study focuses on evaluating the accuracy obtained from different machine learning models in classifying blood pressure using PPG signals. Specifically, Logistic Regression, Support Vector Classifier, Bagging Classifier, and Random Forest Classifier were studied. These models were trained and tested using data from the PulseDB dataset. Three blood pressure classes Hypotension, Normal, and Hypertension were considered for classification. Results indicate varying performance across the machine learning models, with overall accuracy ranked in the following order, 61%, 65%, 72%, and 87%. This research contributes valuable insights into the efficiency of different machine-learning approaches in cuffless blood pressure classification thereby informing the development of wearable devices for cardiovascular health monitoring.