Wearable Estimation of Central Aortic Blood Pressure: Feasibility Study.

  • Germán Andrés Fierro Universidad de la República
  • Ricardo Luis Armentano, R. A. Departamento de Ingeniería Biológica, CENUR Litoral Norte, Universidad de la República, Uruguay.
  • Fernando Silveira, F. S. Instituto de Ingeniería Eléctrica, Facultad de Ingeniería, Universidad de la República, Montevideo, Uruguay.

Abstract

Introduction: Central aortic blood pressure (CABP) estimation from electrocardiogram (ECG) and ballistocardiogram (BCG) acquisitions might be feasible, according to previous research. The result is based on acquisitions with off-the-shelf equipment and some custom electronics. This work aims to evaluate whether a proposed wearable device is capable of achieving similar feasibility results on CABP estimation, but also the study aims to situate the scope of this method to predict CABP readings. Methods: The study used data from one healthy subject involving three days of intermittent CABP observations, and waveforms from ECG and BCG signals. The methodology was evaluated in two levels, from a simple perspective to evaluate feasibility of the method for the CABP estimation ( Evaluation level-1: CABP human-model (HM) is constructed and tested using the same data-sets); to moderate-hard by evaluating the capability of the calibrated CABP-HM to predict unseen CABP data accurately (Evaluation level-2). CABP variables were assessed non-invasively by the use of the SphygmoCor XCEL system (AtCor Medical, Sydney, Australia) during hemodynamic maneuvers. Results: Level-1 evaluation presented strong correlations of r ≈ 0.9, and strong agreement (linear regression parameters) m ≈ 0.8 and y ≈ 20mmHg between CABP measurements and estimations. Root mean square error of RMSE ≈ 2.3mmHg. The level-2 evaluation showed significantly degraded performances when the same figures of merit were assessed. A three-day calibration interval was considered for the level-2 evaluation. Conclusions: Findings in this paper showed that results achieved with off-the-shelf equipment could be replicated by using a proposed wearable device. CABP estimation from the proposed wearable device could be feasible by using three feature times studied in this work (RI, RJ, and IJ intervals) as CABP surrogates. CABP could be accurately predicted by the proposed methodology when (in the order of) daily calibrations are performed.
Published
2020-12-30
Section
Scientific articles