Development and Validation of a Novel Cuff-Less Blood Pressure Monitoring Device

Abstract

Ordinary cuff-based blood pressure-monitoring devices remain a technical limitation that disturbs activities of daily life. Here we report a novel system for the cuff-less blood pressure estimation (CLB) that requires only 1 sensor for photoplethysmography. The present study is the first report to validate and assess the clinical application of the CLB in accordance with the latest wearable device standard (issued by the Institute of Electrical and Electronics Engineers, standard 1708-2014). Our CLB is expected to offer a flexible and wearable device that permits blood pressure monitoring in more continuous and stress-free settings.

Keywords: AAMI, Association for the Advancement of Medical Instrumentation; ABPM, ambulatory blood pressure monitoring; BP, blood pressure; CB, cuff-based blood pressure measurement; CI, confidence interval; CLB, cuff-less blood pressure estimation; DBP, diastolic blood pressure; ECG, electrocardiogram; HF, high-frequency; HR, heart rate; ICC, intraclass correlation coefficient; IEEE, Institute of Electrical and Electronics Engineers; LF, low-frequency; MAD, mean absolute difference; PTG, photoplethysmogram; SBP, systolic blood pressure; ambulatory blood pressure monitoring; blood pressure; diagnosis.

Figures

Graphical abstract

Figure 1

Circuit Diagram for the PW Sensor and BP Estimation System (A) The light-emitting diode (LED) emits light with a wavelength of 940 nm. The light penetrates the finger and arrives at the photo detector (PD). The PD detects blood flow changes, which correspond to the natural pulsation of the blood flow. Baseline fluctuation is removed by a high-pass filter (HPF) with a cutoff frequency of 0.3 Hz, and noise is removed by a low-pass filter (LPF) with a cutoff frequency of 30 Hz. The output signal is digitized at a sampling frequency of 200 Hz and a resolution of 12 bits. (B to E) Schematic diagram (B and D) and illustration (C and E) used to develop the BP estimation algorithm. ∗Leg clamp (E). ∗∗Arithmetic calculation. A total of 887 participants were enrolled (a histogram of participant age and gender is displayed in Supplemental Figure 1A). The obtained pulse waves were analyzed, and feature parameters were extracted and collected. A database of feature parameters was analyzed to generate a BP estimation algorithm. (D) Validation protocol based on Institute of Electrical and Electronics Engineers (IEEE) standard 1708-2014. BP data were obtained by using cuff-less BP estimation (CLB) with simultaneous recording by a cuff-type sphygmomanometer (CB) as a reference conducted at the time point indicated by the closed circle. Calibration was performed at the beginning of each measurement using a cuff-type sphygmomanometer to take 3 measurements at 60-s intervals. After calibration, simultaneous BP monitoring was performed by using CLB and CB under (C) static conditions followed by dynamic measurements using (E) leg stretching and a clamp. ADC = analog to digital converter; Amp = amplifier; PC = personal computer.

Figure 2

Validation and Reproducibility of CLB According to the IEEE 1708-2014 Standard Under Static and Dynamic Conditions Cumulative systolic blood pressure (SBP) dataset was obtained under (A and B) static (n = 386), (C and D) BP rise (n = 182), and (E) reproducible (retaking BP measurements on the same examinee 1 month later, n = 262) conditions. Linear correlation analysis was used to assess the association between the 2 measures. The correlation coefficient is given by r. The results of analyzing diastolic BP data are shown in Supplemental Figure 2. To assess the agreement of BP data measured by using CLB with those recorded by using a cuff-based sphygmomanometer, all BP data were assessed by using a Bland-Altman plot. Scatter plots of difference in SBP between CB and CLB under (B) static conditions, (D) dynamic conditions, and (F) for reproducibility assessment are shown. The average difference in SBP between the CLB and the CB is indicated by solid lines. Dotted lines indicate the 95% limits of agreement (mean ± 2 SDs of the difference in SBP) between the CB and the CLB. The agreement limits of SBP were: (in mean ± 2 SDs): (B) static, −0.3 ± 13.2; (D) BP rise, −2.1 ± 15.0; and (F) reproducibility, −0.8 ± 18.4. Abbreviations as in Figure 1.

Figure 3

Validation of the Cuff-less BP Monitoring Device Under Decreasing BP Conditions During Coronary Arterial Angiography To address the precision requirements outlined by IEEE 1708-2014 for decreasing BP, we conducted simultaneous BP recording using CLB and a standard CB (A). Image of the recording system used during coronary angiography is shown in the left panel in A. (B) Representative recording of changes in intra-arterial BP during coronary angiography (B). Intracoronary arterial administration of nitroglycerin (1.5 mg/bolus shot) reduced arterial BP by >15 mm Hg in accordance with the validation criteria described in IEEE 1708-2014. (C) Typical results from the simultaneous BP recordings of CLB (red line) and CB (blue line). (D and E) Correlation and agreement of SBP data measured by using CB and CLB. Agreement of CLB with CB was analyzed by using a Bland-Altman plot (E) during coronary arterial angiography. Average difference in SBP between the CLB and the CB is indicated by the solid line. The 95% limits of agreement (mean ± 2 SDs of the difference) between CLB and CB are indicated by dotted lines. The agreement limit for SBP was 1.00 ± 20.3. Assessments of DBP are shown in Supplemental Figure 2. Abbreviations as in Figure 1.

Figure 4

Effect of Ambulatory BP Measurement on Sleep Quality The effect of BP measurement during sleep was evaluated in terms of discomfort using (B) a questionnaire about sleep quality and (C to F) physiological parameters (heart rate [HR], high-frequency [HF], and low-frequency [LF]/HF). (A) Examinees were randomly assigned and subjected to a bedtime BP monitoring study using standard ambulatory BP monitoring (ABPM) (Study 1). More than 2 months later, the same examinees were subjected to bedtime BP monitoring using the CLB (Study 2). To avoid any bias resulting from device order, the second study was performed after a long interval. There were no significant differences in the mean SBP or DBP recorded by using either of these devices (Supplemental Figure 5A). (B) The effects on sleep quality of standard ABPM and CLB were compared with a questionnaire. The effect of ABPM and CLB on the sleep quality of 35 participants was assessed using ratings provided on a scale of 0 to 2. A higher score indicates better sleep quality during BP monitoring. A score of 2 (white area) indicates fair or usual sleep quality; a score of 1 (gray area) indicates sleep that was mildly disturbed by BP measurement; and a score of 0 (black area) indicates sleep that was significantly disturbed. P < 0.001 according to the McNemar test. (C to F) Changes in physiological parameters. To assess the effect of a BP cuff on sleep quality, HR variability was analyzed. Typical recordings of HR, HF, and LF/HF during (C) Study 1 and (D) Study 2 are shown. (E) Time course of changes in mean HR during sleep by CB (Study 1, blue line) and by CLB (Study 2, red line) are displayed. The mean HR was significantly lower when using CLB (line) during the first hour after going to bed (F), presumably indicating the time to sleep onset. Abbreviations as in Figure 1.