Blood Pressure Estimation using Arterial Diameter: Exploring Different Machine Learning Methods

Blood pressure (BP) variation have been shown to result in health complications and are continuously related to the risk of stroke and coronary heart disease. Measurement of BP is one of the most useful parameters for early diagnosis. In this paper, we present a unique approach for extensive BP measurement that involves Machine Learning (ML) techniques which utilizes the extracted inherent features that are characteristic of arterial pressure and pulse transit relationships. Concurrent measurement of luminal diameter waveform were obtained from the arterial segments using our developed dualchannel A-mode ultrasound system and a probe with a pair of single-element ultrasound transducers. The data recorded by carrying out an in vivo study of 20 subjects was trained and validated on three ML and one Deep Learning (DL) models. The proposed technique uses a combination of 71 unique features of local arterial dimensions and blood pulse propagation characteristics for continuous cuff-less BP estimation. Out of the proposed techniques Multivariate adaptive regression spline (MARS) exhibited an acceptable accuracy, with a root-mean- square-error of 1.4 mmHg and 4.77 mmHg for Diastolic blood pressure (DBP) and Systolic blood pressure(SBP) respectively. The proposed technique could be used to develop a single PPG based cuff-less BP monitoring system with an accuracy that can be clinically and practically employed.