Specific Delineation Algorithm Based on Two-Stage EWMA for Noninvasive Estimation of Blood Pressure

Ccallo M.L.N.; Contreras R.V.; Talavera J.E.R.; Espinoza E.S. · Revista Mexicana de Ingenieria Biomedica (2025) · DOI: 10.17488/rmib.46.1.1490

Resumen

Dynamic noise, due to its variability and intensity, prevents conventional peak detection methods in ECG and PPG signals based on fixed thresholds from performing effectively in wearable devices. The inflexibility of these fixed thresholds results in low sensitivity and positive predictive value. Therefore, this study proposes a specific delineation algorithm with an adaptive threshold based on the Two-Stage Exponentially Weighted Moving Average (EWMA) model, focusing on flexibility, precision, robustness against strong dynamic noise, and low computational load. The proposed algorithm demonstrated robust performance under high SNR conditions (24 dB and 18 dB), achieving 100 % sensitivity and positive predictive value. Under moderate noise conditions (12 dB), the algorithm maintained a high sensitivity of 99.39 % and a positive predictive value of 98.18 %, with a delineation error rate (DER) of 2.43 %. Even under low SNR conditions (6 dB), the algorithm significantly outperformed fixed-threshold-based approaches, which exhibited error rates exceeding 50 %. Furthermore, the algorithm was validated using a mathematical model to estimate blood pressure based on pulse transit time, with signals from the MIMIC database. The results showed a mean error of -1.422 mmHg for systolic blood pressure (SBP) and 0.577 mmHg for diastolic blood pressure (DBP), with standard deviations of 4.668 mmHg and 2.888 mmHg, respectively, meeting the standards of the Association for the Advancement of Medical Instrumentation (AAMI).

Instituciones: Universidad Nacional de San Agustín de Arequipa