Estimating blood pressure trends and the nocturnal dip from photoplethysmography

Mustafa Radha (Corresponding author), Koen de Groot, Nikita Rajani, Cybele C.P. Wong, Nadja Kobold, Valentina Vos, Pedro Fonseca, Nikolaos Mastellos, Petra A. Wark, Nathalie Velthoven, Reinder Haakma, Ronald M. Aarts

Research output: Contribution to journalArticleAcademicpeer-review

1 Citation (Scopus)

Abstract

OBJECTIVE: Evaluate a method for the estimation of the nocturnal systolic blood pressure (SBP) dip from 24 h blood pressure trends using a wrist-worn photoplethysmography (PPG) sensor and a deep neural network in free-living individuals, comparing the deep neural network to traditional machine learning and non-machine learning baselines.

APPROACH: A wrist-worn PPG sensor was worn by 106 healthy individuals for 226 d during which 5111 reference values for blood pressure (BP) were obtained with a 24 h ambulatory BP monitor and matched with the PPG sensor data. Features based on heart rate variability and pulse morphology were extracted from the PPG waveforms. Long- and short term memory (LSTM) networks, dense networks, random forests and linear regression models were trained and evaluated in their capability of tracking trends in BP, as well as the estimation of the SBP dip.

MAIN RESULTS: Best performance for estimating the SBP dip were obtained with a deep LSTM neural network with a root mean squared error (RMSE) of 3.12 [Formula: see text] 2.20 [Formula: see text] mmHg and a correlation of 0.69 [Formula: see text]. This dip was derived from trend estimates of BP which had an RMSE of 8.22 [Formula: see text] 1.49 mmHg for systolic and 6.55 [Formula: see text] 1.39 mmHg for diastolic BP (DBP). While other models had similar performance for the tracking of relative BP, they did not perform as well as the LSTM for the SBP dip.

SIGNIFICANCE: The work provides first evidence for the unobtrusive estimation of the nocturnal SBP dip, a highly prognostic clinical parameter. It is also the first to evaluate unobtrusive BP measurement in a large data set of unconstrained 24 h measurements in free-living individuals and provides evidence for the utility of LSTM models in this domain.

Original languageEnglish
Article number025006
Number of pages14
JournalPhysiological Measurement
Volume40
Issue number2
DOIs
Publication statusPublished - 26 Feb 2019

Fingerprint

Photoplethysmography
Blood pressure
Blood Pressure
Long-Term Memory
Short-Term Memory
Data storage equipment
Sensors
Wrist
Linear Models
Blood Pressure Monitors
Pressure measurement
Linear regression

Keywords

  • ambulatory blood pressure
  • free-living protocol
  • neural networks
  • photoplethysmography

Cite this

Radha, Mustafa ; de Groot, Koen ; Rajani, Nikita ; Wong, Cybele C.P. ; Kobold, Nadja ; Vos, Valentina ; Fonseca, Pedro ; Mastellos, Nikolaos ; Wark, Petra A. ; Velthoven, Nathalie ; Haakma, Reinder ; Aarts, Ronald M. / Estimating blood pressure trends and the nocturnal dip from photoplethysmography. In: Physiological Measurement. 2019 ; Vol. 40, No. 2.
@article{3671e031d1f54b51a4471ee81c7ab5e9,
title = "Estimating blood pressure trends and the nocturnal dip from photoplethysmography",
abstract = "OBJECTIVE: Evaluate a method for the estimation of the nocturnal systolic blood pressure (SBP) dip from 24 h blood pressure trends using a wrist-worn photoplethysmography (PPG) sensor and a deep neural network in free-living individuals, comparing the deep neural network to traditional machine learning and non-machine learning baselines.APPROACH: A wrist-worn PPG sensor was worn by 106 healthy individuals for 226 d during which 5111 reference values for blood pressure (BP) were obtained with a 24 h ambulatory BP monitor and matched with the PPG sensor data. Features based on heart rate variability and pulse morphology were extracted from the PPG waveforms. Long- and short term memory (LSTM) networks, dense networks, random forests and linear regression models were trained and evaluated in their capability of tracking trends in BP, as well as the estimation of the SBP dip.MAIN RESULTS: Best performance for estimating the SBP dip were obtained with a deep LSTM neural network with a root mean squared error (RMSE) of 3.12 [Formula: see text] 2.20 [Formula: see text] mmHg and a correlation of 0.69 [Formula: see text]. This dip was derived from trend estimates of BP which had an RMSE of 8.22 [Formula: see text] 1.49 mmHg for systolic and 6.55 [Formula: see text] 1.39 mmHg for diastolic BP (DBP). While other models had similar performance for the tracking of relative BP, they did not perform as well as the LSTM for the SBP dip.SIGNIFICANCE: The work provides first evidence for the unobtrusive estimation of the nocturnal SBP dip, a highly prognostic clinical parameter. It is also the first to evaluate unobtrusive BP measurement in a large data set of unconstrained 24 h measurements in free-living individuals and provides evidence for the utility of LSTM models in this domain.",
keywords = "ambulatory blood pressure, free-living protocol, neural networks, photoplethysmography",
author = "Mustafa Radha and {de Groot}, Koen and Nikita Rajani and Wong, {Cybele C.P.} and Nadja Kobold and Valentina Vos and Pedro Fonseca and Nikolaos Mastellos and Wark, {Petra A.} and Nathalie Velthoven and Reinder Haakma and Aarts, {Ronald M.}",
year = "2019",
month = "2",
day = "26",
doi = "10.1088/1361-6579/ab030e",
language = "English",
volume = "40",
journal = "Physiological Measurement",
issn = "0967-3334",
publisher = "Institute of Physics",
number = "2",

}

Radha, M, de Groot, K, Rajani, N, Wong, CCP, Kobold, N, Vos, V, Fonseca, P, Mastellos, N, Wark, PA, Velthoven, N, Haakma, R & Aarts, RM 2019, 'Estimating blood pressure trends and the nocturnal dip from photoplethysmography', Physiological Measurement, vol. 40, no. 2, 025006. https://doi.org/10.1088/1361-6579/ab030e

Estimating blood pressure trends and the nocturnal dip from photoplethysmography. / Radha, Mustafa (Corresponding author); de Groot, Koen; Rajani, Nikita; Wong, Cybele C.P.; Kobold, Nadja; Vos, Valentina; Fonseca, Pedro; Mastellos, Nikolaos; Wark, Petra A.; Velthoven, Nathalie; Haakma, Reinder; Aarts, Ronald M.

In: Physiological Measurement, Vol. 40, No. 2, 025006, 26.02.2019.

Research output: Contribution to journalArticleAcademicpeer-review

TY - JOUR

T1 - Estimating blood pressure trends and the nocturnal dip from photoplethysmography

AU - Radha, Mustafa

AU - de Groot, Koen

AU - Rajani, Nikita

AU - Wong, Cybele C.P.

AU - Kobold, Nadja

AU - Vos, Valentina

AU - Fonseca, Pedro

AU - Mastellos, Nikolaos

AU - Wark, Petra A.

AU - Velthoven, Nathalie

AU - Haakma, Reinder

AU - Aarts, Ronald M.

PY - 2019/2/26

Y1 - 2019/2/26

N2 - OBJECTIVE: Evaluate a method for the estimation of the nocturnal systolic blood pressure (SBP) dip from 24 h blood pressure trends using a wrist-worn photoplethysmography (PPG) sensor and a deep neural network in free-living individuals, comparing the deep neural network to traditional machine learning and non-machine learning baselines.APPROACH: A wrist-worn PPG sensor was worn by 106 healthy individuals for 226 d during which 5111 reference values for blood pressure (BP) were obtained with a 24 h ambulatory BP monitor and matched with the PPG sensor data. Features based on heart rate variability and pulse morphology were extracted from the PPG waveforms. Long- and short term memory (LSTM) networks, dense networks, random forests and linear regression models were trained and evaluated in their capability of tracking trends in BP, as well as the estimation of the SBP dip.MAIN RESULTS: Best performance for estimating the SBP dip were obtained with a deep LSTM neural network with a root mean squared error (RMSE) of 3.12 [Formula: see text] 2.20 [Formula: see text] mmHg and a correlation of 0.69 [Formula: see text]. This dip was derived from trend estimates of BP which had an RMSE of 8.22 [Formula: see text] 1.49 mmHg for systolic and 6.55 [Formula: see text] 1.39 mmHg for diastolic BP (DBP). While other models had similar performance for the tracking of relative BP, they did not perform as well as the LSTM for the SBP dip.SIGNIFICANCE: The work provides first evidence for the unobtrusive estimation of the nocturnal SBP dip, a highly prognostic clinical parameter. It is also the first to evaluate unobtrusive BP measurement in a large data set of unconstrained 24 h measurements in free-living individuals and provides evidence for the utility of LSTM models in this domain.

AB - OBJECTIVE: Evaluate a method for the estimation of the nocturnal systolic blood pressure (SBP) dip from 24 h blood pressure trends using a wrist-worn photoplethysmography (PPG) sensor and a deep neural network in free-living individuals, comparing the deep neural network to traditional machine learning and non-machine learning baselines.APPROACH: A wrist-worn PPG sensor was worn by 106 healthy individuals for 226 d during which 5111 reference values for blood pressure (BP) were obtained with a 24 h ambulatory BP monitor and matched with the PPG sensor data. Features based on heart rate variability and pulse morphology were extracted from the PPG waveforms. Long- and short term memory (LSTM) networks, dense networks, random forests and linear regression models were trained and evaluated in their capability of tracking trends in BP, as well as the estimation of the SBP dip.MAIN RESULTS: Best performance for estimating the SBP dip were obtained with a deep LSTM neural network with a root mean squared error (RMSE) of 3.12 [Formula: see text] 2.20 [Formula: see text] mmHg and a correlation of 0.69 [Formula: see text]. This dip was derived from trend estimates of BP which had an RMSE of 8.22 [Formula: see text] 1.49 mmHg for systolic and 6.55 [Formula: see text] 1.39 mmHg for diastolic BP (DBP). While other models had similar performance for the tracking of relative BP, they did not perform as well as the LSTM for the SBP dip.SIGNIFICANCE: The work provides first evidence for the unobtrusive estimation of the nocturnal SBP dip, a highly prognostic clinical parameter. It is also the first to evaluate unobtrusive BP measurement in a large data set of unconstrained 24 h measurements in free-living individuals and provides evidence for the utility of LSTM models in this domain.

KW - ambulatory blood pressure

KW - free-living protocol

KW - neural networks

KW - photoplethysmography

UR - http://www.scopus.com/inward/record.url?scp=85064094609&partnerID=8YFLogxK

U2 - 10.1088/1361-6579/ab030e

DO - 10.1088/1361-6579/ab030e

M3 - Article

C2 - 30699397

AN - SCOPUS:85064094609

VL - 40

JO - Physiological Measurement

JF - Physiological Measurement

SN - 0967-3334

IS - 2

M1 - 025006

ER -