Photoplethysmography-Based Continuous Systolic Blood Pressure Estimation Method for Low Processing Power Wearable Devices

applied sciences

Article Photoplethysmography-Based Continuous Systolic Blood Pressure Estimation Method for Low Processing Power Wearable Devices

Rolandas Gircys 1, Agnius Liutkevicius 1,* , Egidijus Kazanavicius 1, Vita Lesauskaite 2, Gyte Damuleviciene 2 and Audrone Janaviciute 1

1 Centre of Real Time Computer Systems, Kaunas University of Technology, Barsausko str. 59-A316, LT-51423 Kaunas, Lithuania; [email protected] (R.G.); [email protected] (E.K.); [email protected] (A.J.) 2 Clinical Department of Geriatrics, Lithuanian University of Health Sciences, Josvainiu str. 2, LT-47144 Kaunas, Lithuania; [email protected] (V.L.); [email protected] (G.D.) * Correspondence: [email protected]

Received: 13 April 2019; Accepted: 28 May 2019; Published: 30 May 2019

Abstract: Regardless of age, it is always important to detect deviations in long-term blood pressure from normal levels. Continuous monitoring of blood pressure throughout the day is even more important for elderly people with cardiovascular diseases or a high risk of stroke. The traditional cuff-based method for blood pressure measurements is not suitable for continuous real-time applications and is very uncomfortable. To address this problem, continuous blood pressure measurement methods based on photoplethysmogram (PPG) have been developed. However, these methods use specialized high-performance hardware and sensors, which are not available for common users. This paper proposes the continuous systolic blood pressure (SBP) estimation method based on PPG pulse wave steepness for low processing power wearable devices and evaluates its suitability using the commercially available CMS50FW Pulse Oximeter. The SBP estimation is done based on the PPG pulse wave steepness (rising edge angle) because it is highly correlated with systolic blood pressure. The SBP estimation based on this single feature allows us to significantly reduce the amount of data processed and avoid errors, due to PPG pulse wave amplitude changes resulting from physiological or external factors. The experimental evaluation shows that the proposed SBP estimation method allows the use of off-the-shelf wearable PPG measurement devices with a low sampling rate (up to 60 Hz) and low resolution (up to 8-bit) for precise SBP measurements (mean difference MD = 0.043 and standard deviation SD = 6.79). In contrast, the known methods for − continuous SBP estimation are based on equipment with a much higher sampling rate and better resolution characteristics.

Keywords: blood pressure estimation; photoplethysmogram; pulse wave; pulse oximeter; wearable device

1. Introduction It is very important to be able to determine the moment when physiological parameters deviate from normal levels in order to prevent chronic diseases. The dynamics of observed physiological parameters are more important over the long-run than accurate instant values, which are used in an emergency or cases of critical health disorders. It is important to detect the degree of a physiological parameter’s deviation from normal during a day, week, month or even a year, as well as the length of such a deviation. If deviation persists for a long time, this can suggest the risk of irreversible changes.

Appl. Sci. 2019, 9, 2236; doi:10.3390/app9112236 www.mdpi.com/journal/applsci Appl. Sci. 2019, 9, 2236 2 of 16

Appl. Sci. 2019, 9, x FOR PEER REVIEW 2 of 16 Timely detection of such dangerous deviations from normal and an immediate start of treatment can helpchanges. to avoid Timely negative detection and irreversible of such dangerous health changes. deviations from normal and an immediate start of treatmentTherefore, can the help long-term to avoid negative constant and observation irreversible of physiological health changes. parameters, such as blood pressure allowsTherefore, us to reduce the the long healthcare‐term constant costs, asobservation discussed inof Referencephysiological [1], minimizesparameters, the such risk as of blood visiting secondarypressure allows healthcare us to institutions,reduce the healthcare and ensures costs, a as better discussed quality in Reference of life for [1], a minimizes longer period. the risk Blood of pressurevisiting measurements secondary healthcare made atinstitutions, home indicate and ensures the state a better of the quality cardiovascular of life for systema longer better period. than thoseBlood obtained pressure during measurements regular visitsmade toat ahome doctor, indicate as discussed the state of in the Reference cardiovascular [2]. A verysystem significant better proportionthan those of obtained chronic during diseases regular are cardiovascular visits to a doctor, diseases. as discussed Atherosclerosis in Reference and [2]. endothelial A very significant function changesproportion are accompaniedof chronic diseases by a are rise cardiovascular in blood pressure. diseases. As Atherosclerosis suggested in and Reference endothelial [3], function increasing bloodchanges pressure are accompanied values, measured by a rise at in home, blood canpressure. be considered As suggested as a in prognostic Reference sign[3], increasing of atherosclerosis blood orpressure endothelial values, dysfunction. measured Long-term at home, can blood be pressureconsidered variability, as a prognostic obtained sign during of atherosclerosis daily home blood or pressureendothelial measurement dysfunction. (HBP), Long‐ canterm show blood internal pressure organ variability, damage obtained or risk during of cardiovascular daily home blood events, aspressure reported measurement in Reference (HBP), [4]. The can authors show internal of [5] showorgan thatdamage increased or risk bloodof cardiovascular pressure is events, a potential as dementiareported illness in Reference risk factor; [4]. thus,The authors registering of [5] the show dynamics that increased of blood pressureblood pressure for elderly is a potential people can dementia illness risk factor; thus, registering the dynamics of blood pressure for elderly people can serve as a preventive mental health disorder measure. Hence, the continuous monitoring of blood serve as a preventive mental health disorder measure. Hence, the continuous monitoring of blood pressure is critical for the prevention and diagnosis of hypertension and cardiovascular diseases. pressure is critical for the prevention and diagnosis of hypertension and cardiovascular diseases. Both SBP and diastolic blood pressure (DBP) are important to assess the state of the cardiovascular Both SBP and diastolic blood pressure (DBP) are important to assess the state of the system, but according to References [6–8], systolic hypertension is much more common than diastolic cardiovascular system, but according to References [6–8], systolic hypertension is much more hypertension,common than and diastolic systolic hypertension, blood pressure and systolic contributes blood pressure more tocontributes the huge more global to the disease huge global burden attributabledisease burden to hypertension attributable thanto hypertension diastolic pressure. than diastolic Systolic pressure. blood pressureSystolic blood is much pressure more is important much formore elderly important people for and elderly their healthpeople state and their evaluation, health state as discussed evaluation, in Referencesas discussed [ 9in–12 References]. Since elderly [9– people12]. Since are the elderly major people group are of the geriatrics major group patients, of geriatrics which are patients, subjects which of this are study,subjects the of estimationthis study, of systolicthe estimation blood pressure of systolic is the blood main pressure focus ofis the this main paper. focus of this paper. TheThe traditional traditional auscultatory auscultatory blood blood pressurepressure measurement method method is is not not continuous continuous and and is quite is quite uncomfortable,uncomfortable, because because it usesit uses a cu aff cuffwrapped wrapped around around the armthe arm which which is inflated is inflated repeatedly. repeatedly. To address To thisaddress problem, this continuous problem, continuous blood pressure blood measurement pressure measurement methods have methods been developedhave been recently.developed The mostrecently. popular The methodsmost popular for continuousmethods for andcontinuous cuffless and blood cuffless pressure blood monitoringpressure monitoring are based are on based pulse transiton pulse time (PTT)transit [ 13time–18 ].(PTT) However, [13–18]. PTT However, is calculated PTT using is calculated two signals using (usually two signals electrocardiogram (usually andelectrocardiogram pulse wave), which and ispulse quite wave), difficult which to implement is quite difficult in wearable to implement devices forin wearable daily use. devices for dailyBlood use. pressure estimation methods which are based on one-channel pulse wave measurements are lessBlood complex pressure and allow estimation us to createmethods wearable which are devices based which on one are‐channel more comfortable pulse wave measurements for the end users. A pulseare less wave complex is often and obtained allow us using to create the wearable photoplethysmogram devices which (PPG) are more optical comfortable measurement for the method. end Bloodusers. pressure A pulse measurement wave is often methods obtained basedusing onthe one-channel photoplethysmogram PPG are implemented (PPG) optical usingmeasurement stationary method. Blood pressure measurement methods based on one‐channel PPG are implemented using medical devices [10,19–22], original PPG measurement devices, such as those in References [23–25], stationary medical devices [10,19–22], original PPG measurement devices, such as those in References or even smartphones [26,27]. The smartphone approach is not truly continuous, since the user does [23–25], or even smartphones [26,27]. The smartphone approach is not truly continuous, since the not wear a PPG sensor. Meanwhile, stationary medical devices or devices created in laboratories are user does not wear a PPG sensor. Meanwhile, stationary medical devices or devices created in hardly wearable or are available only for a very small group of potential users. laboratories are hardly wearable or are available only for a very small group of potential users. Therefore,Therefore, this this research research investigates investigates thethe possibility of of using using commercial commercial-o‐off‐ffthe-the-shelf‐shelf wrist wrist pulse pulse oximetersoximeters for for continuous continuous cu cufflessffless bloodblood pressure measurement, measurement, since since they they are are certified, certified, quite quite cheap, cheap, and,and, most most importantly, importantly, accessible accessible to to everyone everyone (see(see FigureFigure 11).).

FigureFigure 1. 1.The The CMS50FW CMS50FW wearablewearable oxygen saturation (SpO2) (SpO2) and and pulse pulse rate rate device device with with wireless wireless communicationcommunication for for real-time real‐time vital-sign vital‐sign monitoring.monitoring. Appl. Sci. 2019, 9, 2236 3 of 16 Appl. Sci. 2019, 9, x FOR PEER REVIEW 3 of 16

If aa pulse pulse oximeter oximeter is capableis capable of PPGof PPG signal signal transmission transmission over Bluetoothover Bluetooth or Wi-Fi, or thenWi‐Fi, it is then possible it is topossible send PPG to send values PPG to values a mobile to a phonemobile or phone another or another device fordevice further for further SBP estimation SBP estimation and SBP and value SBP recording.value recording. This allows This allows us to implement us to implement a 24/7 a system 24/7 system for SBP for monitoring. SBP monitoring. Other authors usually obtain PPG signal values using a 125–1000 Hz sampling rate and 8–12-bit8–12‐bit resolution analog digital converter (ADC) [[10,13,15,19–25].10,13,15,19–25]. However, energy consumption is the key feature of wearable devices, and a lower sampling rate and resolution allow us to significantly signiﬁcantly reduce ADCADC energyenergy consumption,consumption, as discussed in Reference [[28].28]. Moreover, the energy used for thethe data transmission is lower as well, because of the lower number of bytes per sample and the samples per second rate. This paperpaper proposes proposes a novela novel approach approach for thefor estimation the estimation of SBP, of using SBP, a using much lowera much PPG lower sampling PPG ratesampling and resolution rate and resolution than other than authors other andauthors analyzing and analyzing not all of not the all PPG of the pulse PPG wave, pulse butwave, its but rising its edgerising only, edge thereby only, thereby reducing reducing the number the number of calculations of calculations (and energy (and energy consumption) consumption) to obtain to obtain a result. a Experimentalresult. Experimental evaluation evaluation is done is using done a using CMS50FW a CMS50FW wrist pulse wrist oximeter pulse oximeter with a 60 with Hz samplinga 60 Hz frequencysampling frequency and 8-bit resolution, and 8‐bit resolution, which is widely which available, is widely quite available, cheap quite and is cheap capable and of is transmitting capable of PPGtransmitting values over PPG Bluetooth. values over Bluetooth.

2. Materials and Methods The methodology proposed inin thisthis paperpaper isis depicteddepicted inin FigureFigure2 2..

Figure 2. The block diagram of the proposed methodology. PPG: Photoplethysmogram.

The PPG signal preprocessingpreprocessing isis performed,performed, includingincluding normalization and filtering,filtering, pulse wave peak andand valleyvalley point point detection, detection, as as well well as as base base line line correction correction and and peak peak line line correction, correction, as as described described in thein the following following sections. sections. The The systolic systolic blood blood pressure pressure values values are are estimated estimated from from PPG PPG according according toto thethe mathematical model proposedproposed inin ourour previousprevious workwork [[29].29]. 2.1. Systolic Blood Pressure Estimation Model 2.1. Systolic Blood Pressure Estimation Model The mathematical–physical model for SBP estimation proposed in Reference [29] is based on The mathematical–physical model for SBP estimation proposed in Reference [29] is based on Hook’s law as defined in Equation (1): Hook’s law as defined in Equation (1):  σ = ∙Eε∙ℎh //11 θ2,, (1) · · − where d is the diameter of the artery, d0 is the initial diameter of the artery, h is the thickness of the where d is the diameter of the artery, d0 is the initial diameter of the artery, h is the thickness of the artery wall,  = d/d0 is strain,  is stress, E is the Young’s modulus, and  is Poisson’s ratio. artery wall, ε = d/d0 is strain, σ is stress, E is the Young’s modulus, and θ is Poisson’s ratio. Although pressure (p), stress and Young’s modulusmodulus are didifferentfferent physicalphysical quantities,quantities, theythey have −1 −2 the same dimensions dim(p)dim(p) == dim(dim(σ) )= =dim(E)dim(E − ) L LMT1MT, where2, where L is length,L is length, M is Mmass,is mass, and T and is time;T is − − − time;hence, hence, an equals an equals sign can sign be canwritten be written between between them: p them:  E  p. Hence,E σ .Young’s Hence, modulus Young’s modulusin Equation in ≡ ≡ Equation(1) can be (1) changed can be with changed pressure with without pressure losing without physical losing physicalmeaning. meaning. Changing Changing one physical one physicalquantity quantityto another to anotherwhen their when dimensions their dimensions are equal are equal is quite is quite common common practice, practice, which which can can be be seen seen by comparing the Bramwell–Hill and Moens–Korteweg equations describing pulse wave velocity. The Moens–Korteweg equation usesuses Young’sYoung’s modulus,modulus, whilewhile Bramwell–HillBramwell–Hill uses uses pressure. pressure. Let us change stress inin EquationEquation (1)(1) toto pressurepressure asas defineddefined inin EquationEquation (2):(2): 𝑝  ∙  ∙ℎ /1  . (2) p = E ε h/ 1 θ2 . (2) When blood pressure p(t) changes during· the· cardiac− cycle, the diameter of artery d(t) changes as well; therefore, Equation (2) can be re‐written as a time function as given by Equation (3): Appl. Sci. 2019, 9, 2236 4 of 16

When blood pressure p(t) changes during the cardiac cycle, the diameter of artery d(t) changes as well; therefore, Equation (2) can be re-written as a time function as given by Equation (3):

d(t) h p(t) = E , (3) 2 · d0 · (1 θ ) − where pAppl.(t) is Sci. the 2019 blood, 9, x FOR pressure PEER REVIEW change function during the cardiac cycle (blood pressure4 pulse of 16 wave). Function p(t) is called the pressure pulse wave (PPW) in the remainder of this paper, while relative deformation variation d(t)/d0 is the deformation pulse wave (DPW), where d0 is the artery diameter 𝑝𝑡 𝐸∙ ∙ , (3) at the beginning of the cardiac cycle. Let us denote DPW as D(t) = d(t)/d , and the remainder of 0 Equationwhere (3) asp(t) a constantis the bloodK =pressureE h/ 1changeθ2 function, leading during to the the following cardiac cycle Equation (blood (4):pressure pulse wave). Function p(t) is called the pressure− pulse wave (PPW) in the remainder of this paper, while relative deformation variation d(t)/d0 is thep( deformationt) = K D(t) pulse. wave (DPW), where d0 is the artery (4) diameter at the beginning of the cardiac cycle. Let us· denote DPW as D(t) = d(t)/d0, and the remainder of Equation (3) as a constant 𝐾𝐸ℎ/1𝜃, leading to the following Equation (4): If the stress–strain characteristic is linear, i.e., E = const., then, if K is known, p(t) values can be calculated for the whole cardiac cycle interval.𝑝𝑡 Usually, 𝐾 ∙ 𝐷𝑡we. are interested in ﬁnding the maximum(4) PPW value,If systolic the stress–strain blood pressure characteristic value is plinear,i = max i.e.,(p (Et )),= const. which, then, is achieved if K is known, during p(t) timevalues period can be ts at the i-th cardiaccalculated cycle. for the whole cardiac cycle interval. Usually, we are interested in finding the maximum PPW value, systolic blood pressure value pi = max(p(t)), which is achieved during time period ts at the If K = const. in Equation (4), then the DPW steepness tan(γ)i = DPW(t)/t during i-th cardiac cycle is i‐th cardiac cycle. tan(γ)i = const., while the systolic blood pressure value pi = max(p(t)), when 0 t ts, can be obtained If K = const. in Equation (4), then the DPW steepness tan(γ)i = DPW(t)/t during≤ i‐th≤ cardiac cycle as followingis tan(γ Equation)i = const., while (5): the systolic blood pressure value pi = max(p(t)), when 0 ≤ t ≤ ts, can be obtained as following Equation (5): p = K tan(γ) ts, (5) i · · 𝑝 𝐾∙tan𝛾∙𝑡, (5) where pi and tan(γ)i change equally during each cardiac cycle. Thewhere DPW pi formand tan( dependsγ)i change on equally the stress–strain during each characteristic,cardiac cycle. which is not linear in the case of blood vessels. BecauseThe DPW of form this, depends pressure on andthe stress–strain deformation characteristic, pulse waveforms which is not are linear di ﬀinerent, the case while of bloodK , const. The authorsvessels. of Because [30–32 of] show this, pressure that in and the deformation part where pulse PPW waveforms and DPW are aredifferent, changing while K most ≠ const. rapidly, The their authors of [30–32] show that in the part where PPW and DPW are changing most rapidly, their trajectories concur (see Figure3). trajectories concur (see Figure 3).

(a) (b)

FigureFigure 3. (a) 3. PPW (a) PPW and and DPW DPW trajectories; trajectories; ( b(b)) normalizednormalized PPW PPWn =n PPW/max(PPW)= PPW/max(PPW) and DPW andn DPW= n = DPW/max(DPW)DPW/max(DPW) trajectories trajectories coincide. coincide.

Hence,Hence,tan(γ )tan(max_iγ)max_icorrelates correlates with with the the systolicsystolic blood blood pressure pressure value value of the of i‐ theth cardiaci-th cardiac cycle, and cycle, and this SBPthis value SBP value can be can calculated be calculated using using Equation Equation (5) and and taking taking the the steepest steepest part of part the of DPW the rising DPW rising edge: tan(γ)max_i = max(tan(γj)), 1 ≤ j ≤ M−1, where M is the number of sampled points of DPW rising edge: tan(γ)max_i = max(tan(γj)), 1 j M 1, where M is the number of sampled points of DPW rising edge, γj is the angle between the≤ base≤ line− and the line drawn through the j‐th and (j+1)‐th points of edge, γjrisingis the edge, angle and between i is the pulse the wave base number. line and the line drawn through the j-th and (j + 1)-th points of rising edge, and i is the pulse wave number. 2.2. Signal Preprocessing 2.2. Signal Preprocessing SBP is estimated in two steps. The initial step includes the calculation of systolic blood pressure SBPvalues is estimated as an average in two value steps. obtained The from initial 15 pulse step waves. includes Later the medians calculation of these of values systolic are averaged blood pressure once more (see the “Systolic blood pressure estimation algorithm” subsection and Equation (6) for values as an average value obtained from 15 pulse waves. Later medians of these values are averaged details) to get the final result. Averaging is common practice to reduce variability. In order to get a once moremean (see value, the which “Systolic reflects blood the real pressure value with estimation a probability algorithm” higher than subsection p = 0.5, more and than Equation 10 values (6) for details)should to get be the taken final for result. averaging. Averaging The time isperiod common of the PPG practice record, to which reduce contains variability. at least In15 pulse order to get waves, is called the blood pressure estimation window (BPEW) (see Figure 4). The time span for Appl. Sci. 2019, 9, 2236 5 of 16 a mean value, which reflects the real value with a probability higher than p = 0.5, more than 10 values should be taken for averaging. The time period of the PPG record, which contains at least 15 pulse waves, is called the blood pressure estimation window (BPEW) (see Figure4). The time span for BPEW is based on the minimum possible heart rate of elderly patients with bradycardia, which can fall to 40 bpm. In such a case, one pulse wave fits in 60/40 = 1.5 s (the time interval between two heartbeats). Therefore, BPEW should not be less than 15 1.5 23 s (we used a 25 s BPEW for better results). The × ≈ BPEW duration remains constant during the whole PPG recording time. The signal being recorded is influenced by various noise sources: Breath, PPG sensor motion noise, high voltage lines’ inducted noise, low amplitude PPG signal, and premature ventricular contraction, as reported in Reference [33]. This can lead to a situation where BPEW does not contain 15 pulse waves suitable for processing. In such a case, BPEW is dropped, and the next 25 s BPEW is taken for further calculations (non-overlapping moving window principle). This is completely acceptable for soft real-time self-measured blood pressure (SMBP) monitoring, since the final SBP value is estimated from several BPEW (in our case, one SBP value is obtained in a 5-min period, as explained later). Additionally, to reduce such cases, PPG sensor motion noise is eliminated using common signal processing methods. The width of the signal preprocessing window (SPW) is chosen according to the breath period, because it is the lowest-frequency noise in the signal. The inhalation phase in the PPG signal gives a rising/falling trend. Normally, in a calm state, a person performs around 10–20 breath cycles per minute, while inspiration takes approximately 5 s, as discussed in References [34–37]. Therefore, SPW = 5 s is chosen for signal preprocessing in BPEW (see Figure4).

Figure 4. PPG with signal preprocessing window (SPW), blood pressure estimation window (BPEW) and peak detection window (PDW).

The following steps are performed for PPG signal preprocessing: (1) Normalization: The PPG signal in SPW is replaced by a signal with a zero mean and variability equal to one. (2) Filtering: Although the power of the high-frequency harmonics is small compared to the harmonics in low and mid-frequency bands (see Figure5b), their influence on the form of the signal in the time domain is still obvious (see Figure5a). Various algorithms are proposed for noise filtering: Empirical mode decomposition [38], discrete wavelet transform [39–41], infinite impulse response (IIR) [10] and fast Fourier transform [26]. The FFT technique is chosen in this work, the output of which is shown in Figure6. The largest part of the energy of the PPG signal is accumulated in the frequency band up to 16 Hz. In this frequency band, the average frequencies describe the pulse wave structure, while the high frequencies define the details of the pulse waveform, as discussed in Reference [42]. The FFT filter is implemented by zeroing the frequency band above 16 Hz according to Reference [43]. The filtered signal (see Figure7) is obtained by reverse FFT; then, the signal is normalized, making its average equal zero. (3) Eliminating the rising/falling trend of the signal: As mentioned before, the inspiration phase takes approximately 5 s, and therefore SPW contains PPG, which is influenced by breathing (see Figure4), resulting in a rising/falling signal trend. To eliminate this linear trend in SPW, the least squares method is used to find the straight line (see Figure8a) of this trend. Then, each PPG signal value is decreased Appl. Sci. 2019, 9, x FOR PEER REVIEW 5 of 16

BPEW is based on the minimum possible heart rate of elderly patients with bradycardia, which can fall to 40 bpm. In such a case, one pulse wave fits in 60/40 = 1.5 s (the time interval between two heartbeats). Therefore, BPEW should not be less than 15 × 1.5 ≈ 23 s (we used a 25 s BPEW for better results). The BPEW duration remains constant during the whole PPG recording time. The signal being recorded is influenced by various noise sources: Breath, PPG sensor motion noise, high voltage lines’ inducted noise, low amplitude PPG signal, and premature ventricular contraction, as reported in Reference [33]. This can lead to a situation where BPEW does not contain 15 pulse waves suitable for processing. In such a case, BPEW is dropped, and the next 25 s BPEW is taken for further calculations (non‐overlapping moving window principle). This is completely acceptable for soft real‐time self‐measured blood pressure (SMBP) monitoring, since the final SBP value is estimated from several BPEW (in our case, one SBP value is obtained in a 5‐min period, as explained later). Additionally, to reduce such cases, PPG sensor motion noise is eliminated using common signal processing methods. The width of the signal preprocessing window (SPW) is chosen according to the breath period, because it is the lowest‐frequency noise in the signal. The inhalation phase in the PPG signal gives a rising/falling trend. Normally, in a calm state, a person performs around 10–20 breath cycles per minute, while inspiration takes approximately 5 s, as discussed in References [34–37]. Therefore, SPW = 5 s is chosen for signal preprocessing in BPEW (see Figure 4).

Figure 4. PPG with signal preprocessing window (SPW), blood pressure estimation window (BPEW) and peak detection window (PDW).

Appl. Sci. 2019,The9, 2236 following steps are performed for PPG signal preprocessing: 6 of 16 (1) Normalization: The PPG signal in SPW is replaced by a signal with a zero mean and variability equal to one. by the corresponding(2) Filtering: valueAlthough of thisthe power straight of line,the high resulting‐frequency in aharmonics detrended is small PPG compared signal, as to depicted the in Figure8harmonicsb. in low and mid‐frequency bands (see Figure 5b), their influence on the form of the signal in the time domain is still obvious (see Figure 5a).

Appl. Sci. 2019, 9, x FOR PEER REVIEW 6 of 16

Appl. Sci. 2019, 9, x FOR PEER REVIEWFigure 5. (a) Normalized PPG signal in SPW and (b) its spectrum. 6 of 16

Various algorithmsFigure 5. are (a) proposed Normalized for noise PPG filtering:signal in Empirical SPW and mode (b) its decomposition spectrum. [38], discrete wavelet transform [39–41], infinite impulse response (IIR) [10] and fast Fourier transform [26]. The VariousFFT technique algorithms is chosen are proposed in this work, for the noise output filtering: of which Empirical is shown in mode Figure decomposition 6. [38], discrete wavelet transform [39–41], infinite impulse response(a) (IIR) [10] and fast Fourier transform [26]. The FFT technique is chosen in this work, the output of which is shown in Figure 6.

Appl. Sci. 2019, 9, x FOR PEER REVIEW 6 of 16

Figure 6. TheFigure block 5. (diagrama) Normalized of signal PPG fast signal Fourier in SPWtransform and ( b(FFT)) its spectrum. filtering algorithm.

Various algorithms are proposed for noise filtering: Empirical mode decomposition [38], discrete The largest part of the energy of the PPG signal(b) is accumulated in the frequency band up to 16 Hz.wavelet In this transform frequency [39–41], band, theinfinite average impulse frequencies response describe (IIR) [10] the andpulse fast wave Fourier structure, transform while [26]. the high The frequenciesFFT Figuretechnique define6. TheisFigure chosen theblock details 5. in diagram(a this) Normalized of work, the of pulse signalthe outputwaveform, PPG fast signal ofFourier which as in discussed SPWtransform is shown and in in ( b(FFT)Reference )Figure its spectrum.filtering 6. [42]. algorithm.The FFT filter is implemented by zeroing the frequency band above 16 Hz according to Reference [43]. The filtered Thesignal largest (see part Figure of 7)the is obtainedenergy ofby thereverse PPG FFT; signal then, is the accumulated signal is normalized, in the frequency making its averageband up to 16 Hz. In thisequal frequency zero. band, the average frequencies describe the pulse wave structure, while the high frequencies define the details of the pulse waveform, as discussed in Reference [42]. The FFT filter is implemented by zeroing the frequency band above 16 Hz according to Reference [43]. The filtered signal (see Figure 7) is obtained by reverse FFT; then, the signal is normalized, making its average equal zero. FigureFigure 6. The 6. blockThe block diagram diagram of of signal signal fastfast Fourier transform transform (FFT) (FFT) filtering ﬁltering algorithm. algorithm.

The largest part of the energy of the PPG signal is accumulated in the frequency band up to 16

Hz. In this frequency band, the average frequencies describe the pulse wave structure, while the high frequencies define the details of theFigure pulse 7.waveform, Filtered signal as discussed in SPW. in Reference [42]. The FFT filter is implemented by zeroing the frequency band above 16 Hz according to Reference [43]. The filtered (3) Eliminating the rising/falling trend of the signal: As mentioned before, the inspiration phase signal (see Figure 7) is obtained by reverse FFT; then, the signal is normalized, making its average takes approximately 5 s, and therefore SPW contains PPG, which is influenced by breathing (see equal zero. Figure 4), resulting in a rising/falling signal trend. To eliminate this linear trend in SPW, the least squares method is used to find the straight line (see Figure 8a) of this trend. Then, each PPG signal value is decreased by the corresponding value of this straight line, resulting in a detrended PPG signal, as depicted in Figure 8b. FigureFigure 7. 7. FilteredFiltered signal signal in in SPW. SPW.

(3) Eliminating the rising/falling trend of the signal: As mentioned before, the inspiration phase takes approximately 5 s, and therefore SPW contains PPG, which is influenced by breathing (see

Figure 4), resulting in a rising/falling signal trend. To eliminate this linear trend in SPW, the least Figure 7. Filtered signal in SPW. squares method is used to find the straight line (see Figure 8a) of this trend. Then, each PPG signal value is decreased(3) Eliminating by the the corresponding rising/falling trend value of the ofsignal this: As straight mentioned line, before, resulting the inspiration in a detrended phase PPG signal, astakes depicted approximately in Figure 5 s, 8b. and therefore SPW contains PPG, which is influenced by breathing (see Figure 4), resulting in a rising/falling signal trend.(a) To eliminate this linear trend in SPW, the least squares method is used to find the straight line (see Figure 8a) of this trend. Then, each PPG signal value is decreased by the corresponding value of this straight line, resulting in a detrended PPG signal, as depicted in Figure 8b.

(b) (a) Figure 8. (a) PPG signal in SPW with the linear trend; (b) detrended PPG signal. (a)

(b) (b) Appl. Sci. 2019, 9, 2236 7 of 16

(4) Pulse wave peak point detection is performed using the adaptive peak detection window (PDW) method [44], which starts by setting PDW as the minimum heart contraction period. Then, PDW is divided into three parts, and according to the proposed algorithm, the heart rate is calculated. PDW is increased, divided into three parts and the heart rate is calculated again. This process is repeated until the PDW width reaches its maximum possible period. The histogram is constructed from the obtained heart rates, and the period corresponding to the most commonly found heart rate is used as optimal PDW. Such an algorithm sometimes gives incorrect results when more than two pulse waves fall into PDW; for example, if a heart rate is 200 bpm and PDW width reaches 1.5 s (40 bpm), then 5 pulse waves will fall into the PDW. Besides, if the heart rate is low, this algorithm is not efficient in terms of calculation time. We modified this method by changing the PDW definition rules. These rules are applied for each SPW separately, as defined in Algorithm 1.

Algorithm 1. PDW size calculation 1 procedure pdw (p1, p2) A: Input Signal ∇ mHR: min Heart Rate ∇ 2 Tmin = (1/mHR)*1/2; L = round(Tmin/Fs); 3 i = 1; N = length(A)/L; 4 while (i+1)*L <= N do 5 [k1, c] = linaprox(A(i*L-L+1: i*L)) 6 [k2, c] = linaprox(A(i*L:(i+1)*L)) 7 if k1 > 0 and k2 < 0 8 then p1 =i*L-L+1; p2 = (i+1)*L; 9 goto 16 line 10 end 11 if (k1 < 0 and k2 < 0) or (k1 < 0 and k2 > 0) 12 then i = i + 1; goto 15 line 13 if (k1 > 0 and k2 > 0) 14 then L = L + 0.1*L; 15 end /*While*/ 16 end

The first part of PDW can contain a pulse wave falling edge (or part of it), while the rising edge of the wave appears at the end of PDW. In such a case, the pulse wave peak will not fall into PDW. To detect the beginning of PDW (p1) and its end (p2), PDW = 2L signal values are taken from SPW, where L is number of values fitting in half of the minimum period Tmin. The least squares method is used to derive straight lines through A(p1:L) and A(L:p2). If the slopes of these lines are negative (k1 < 0, k2 < 0), then the values belong to the falling edge of the pulse wave (see Figure9a), and PDW is shifted to the right by the L values. If the slope coe fficients are positive (k1 > 0, k2 > 0), then the values belong mostly to the rising edge, but they do not contain peak values (see Figure9c). In this case, L is increased by 1 /10 and new slope coefficients are found. When k1 < 0 and k2 > 0 values belong to the rising and falling edges of the pulse wave, but do not contain a peak (see Figure9b), the PDW is shifted to the right by the L values. Only when k1 > 0 and k2 < 0 does the PDW contain a peak value (see Figure9d). (5) Pulse wave valley point detection is performed using the algorithm proposed in Reference [45]. (6) Base line correction: This step is performed when the starting points of pulse waves in BPEW are found. When a PPG signal is filtered, the unwanted harmonics are eliminated, while base line corrections give a signal whose values vary about a constant value. However, the line connecting the starting points of each pulse waveform a piecewise linear curve. This means that the amplitude values of the pulse wave’s starting and ending points are different. For this reason, incorrect start Appl. Sci. 2019, 9, x FOR PEER REVIEW 8 of 16

The first part of PDW can contain a pulse wave falling edge (or part of it), while the rising edge of the wave appears at the end of PDW. In such a case, the pulse wave peak will not fall into PDW. To detect the beginning of PDW (p1) and its end (p2), PDW = 2L signal values are taken from SPW, where L is number of values fitting in half of the minimum period Tmin. The least squares method is used to derive straight lines through A(p1:L) and A(L:p2). If the slopes of these lines are negative (k1 < 0, k2 < 0), then the values belong to the falling edge of the pulse wave (see Figure 9a), and PDW is shifted to the right by the L values. If the slope coefficients are Appl.positive Sci. 2019 (k, 19 ,> 2236 0, k2 > 0), then the values belong mostly to the rising edge, but they do not contain peak8 of 16 values (see Figure 9c). In this case, L is increased by 1/10 and new slope coefficients are found. When k1 < 0 and k2 > 0 values belong to the rising and falling edges of the pulse wave, but do not points,contain peak a peak points, (see and Figure rising 9b), edge the slopesPDW is of shifted the pulse to the wave right can by be the found; L values. thus, Only base when line correctionk1 > 0 and is implementedk2 < 0 does the as PDW explained contain in a Reference peak value [46 (see] (see Figure Figure 9d). 10 a).

(a) (b)

Appl. Sci. 2019, 9, x FOR PEER REVIEW 9 of 16 (c) (d) PPG registration and distorts the rising edge trajectory time–amplitude proportions. Therefore, the peak FigurelineFigure (connecting 9. 9.( a()a) PPG PPG signal’s signal’spulse fallingfallingwave edgepeak in in points)PDW; PDW; (b (b,correction c, )c PPG) PPG signal’s signal’s helps falling fallingto eliminateand and rising rising edgerandom edge without without amplitude a changesa peakpeak (see pointpoint Figure in PDW; 10b), ( (d dwithout)) PPG PPG signal signal influencing with with a apeak peak the point point form in in PDW.of PDW. the PDW: pulse PDW: Peak wave. Peak detection detection window. window.

5) Pulse wave valley point detection is performed using the algorithm proposed in Reference [45]. 6) Base line correction: This step is performed when the starting points of pulse waves in BPEW are found. When a PPG signal is filtered, the unwanted harmonics are eliminated, while base line corrections give a signal whose values vary about a constant value. However, the line connecting the starting points of each pulse waveform a piecewise linear curve. This means that the amplitude values of the pulse wave’s starting and ending points are different. For this reason, incorrect start points, peak points, and rising edge slopes of the pulse wave can be found; thus, base line correction is implemented as explained in Reference [46] (see Figure 10a). (a) 7) Peak line correction: This correction is performed when pulse wave peaks are found in BPEW. The form of the DPW rising edge trajectory (not the amplitude) carries information about systolic blood pressure as explained in Reference [29]. Therefore, it is important to ensure uniform time‐ amplitude proportions of all DPW rising edges. When the DPW sampling rate is 60 Hz, the variability of duration of the rising edge is low, as reported in Reference [47], which results in a constant value of abscissa (time) and does not affect the time–amplitude proportions. The motion of the arms or body of the patient (with sensors put on them) results in an uneven amplitude fluctuation during

(b)

Figure 10. ((aa)) PPG PPG signal signal after after base base line line correction; correction; ( (b)) PPG signal after peak line correction.

This(7) Peak correction line correction is done: This for correctioneach pulse is wave performed (when when base pulseline correction wave peaks is performed, are found in and BPEW. all pulseThe form wave of the starting DPW risingpoint edgevalues trajectory become (not zero) the amplitude)by dividing carries its values information by maximum about systolic value, blood as explainedpressure as in explained Reference in [48]. Reference [29]. Therefore, it is important to ensure uniform time-amplitude proportions of all DPW rising edges. When the DPW sampling rate is 60 Hz, the variability of duration 2.3.of the Systolic rising Blood edge Pressure is low, as Estimation reported Algorithm in Reference [47], which results in a constant value of abscissa Systolic blood pressure is estimated using Equation (5). Coefficient K is found for each subject experimentally from one 5 min duration training period and later is used for all testing periods. Coefficient K is determined using Equation (5) in such a way that the difference between estimated (from PPG record) and measured (using a conventional cuff‐based meter) blood pressure values would be minimal. SBPei (0 ≤ i ≤ length (BPEW), i—value number) is estimated for each pulse wave rising edge in BPEW as described in Figure 11.

Figure 11. Estimation of SBPei = K∙tan(gmax_i)∙ts, where ts is the duration of the rising edge, Ts is the

sampling period, gmax_i is the maximum steepness of the rising edge, K = SBPm / tan(gmax_i)∙ts, and SBPm is the systolic blood pressure which is measured using a conventional cuff‐based meter during the initial training session.

Since estimated values are very scattered—i.e., standard deviation SDSBPei = ±10.6—SBPei values in BPEW are replaced by their median values SBPmdj (see Figure 12). Appl. Sci. 2019, 9, x FOR PEER REVIEW 9 of 16

PPG registration and distorts the rising edge trajectory time–amplitude proportions. Therefore, the peak line (connecting pulse wave peak points) correction helps to eliminate random amplitude changes (see Figure 10b), without influencing the form of the pulse wave.

(a)

Appl. Sci. 2019, 9, 2236 9 of 16

(time) and does not affect the time–amplitude proportions. The motion of the arms or body of the patient (with sensors put on them) results in an uneven amplitude fluctuation during PPG registration and distorts the rising edge trajectory time–amplitude(b) proportions. Therefore, the peak line (connecting pulse waveFigure peak 10. points) (a) PPG correction signal after helpsbase line to correction; eliminate ( randomb) PPG signal amplitude after peak changes line correction. (see Figure 10b), without influencing the form of the pulse wave. This correctioncorrection is is done done for for each each pulse pulse wave wave (when (when base base line line correction correction is performed, is performed, and all and pulse all wavepulse startingwave starting point values point becomevalues zero)become by dividingzero) by itsdividing values byits maximumvalues by value, maximum as explained value, inas Referenceexplained [in48 Reference]. [48].

2.3. Systolic Blood Pressure Estimation Algorithm Systolic blood pressure is estimatedestimated using EquationEquation (5). CoeCoefficientfficient K is found forfor each subjectsubject experimentally from from one one 5 min5 min duration durationtraining training period periodand later and is usedlater foris allusedtesting for periodsall testing. Coe periodsfficient. KCoefficientis determined K is determined using Equation using (5) Equation in such a(5) way in such that thea way diff thaterence the between difference estimated between (from estimated PPG record)(from PPG and measuredrecord) and (using measured a conventional (using a cu conventionalff-based meter) cuff blood‐based pressure meter) values blood would pressure be minimal. values SBPei (0 i length (BPEW), i—value number) is estimated for each pulse wave rising edge in BPEW would be≤ minimal.≤ SBPei (0 ≤ i ≤ length (BPEW), i—value number) is estimated for each pulse wave asrising described edge in in BPEW Figure as 11 described. in Figure 11.

Figure 11.11. EstimationEstimation ofof SBPSBPei == K∙tan(gmax_i)∙)tst,s ,where where tst sisis the the duration duration of of the the rising edge, Ts is thethe ei · max_i · sampling period,period,g gmax_i isis the the maximum steepness ofof thethe risingrising edge, edge,K K= = SBPSBPmm/ tan/ tan(g(gmax_i))t∙tss,, andand SBPSBPmm max_i max_i · isis thethe systolicsystolic blood pressure which isis measuredmeasured using aa conventionalconventional cucuffﬀ-based‐based meter during thethe initialinitial trainingtraining session.session.

Since estimatedestimated valuesvalues are are very very scattered—i.e., scattered—i.e., standard standard deviation deviationSD SDSBPeiSBPei= = ±10.6—10.6—SBPSBPeiei values ± inin BPEW are replacedreplaced byby theirtheir medianmedian valuesvalues SBPSBPmdjmdj (see(see Figure Figure 12).12). Appl. Sci. 2019, 9, x FOR PEER REVIEW 10 of 16

Figure 12. Estimated SBPeiei blood pressure values (curved(curved line)line) andand SBPSBPmdjmdj median values (stepwise line), calculated from SBPeiei inin BPEW.

When blood pressure is measured a few times per day at home or in the hospital, it is assumed that it remains constant between measurements. X 𝑆𝐵𝑃 1 ∑M 𝑆𝐵𝑃 . (6) SBPe = SBPmdj. (6) M j=1 Therefore, in this work, for a 5‐min PPG period, we calculate one SBPe value using Equation (6), SBP whichTherefore, is equal to in the this mean work, of for the a medians 5-min PPG (see period, Figure we 13): calculate one e value using Equation (6), which is equal to the mean of the medians (see Figure 13):

Figure 13. Estimated SBPe blood pressure value (dot line) and SBPmdj in BPEW (stepwise line).

2.4. Experiment Setup A commercial blood pressure meter with a cuff (OMRON M1) was used as the gold standard in this work to measure SBP on the left brachial artery. The PPG signal was collected using the CMS50FW Pulse Oximeter without automatic gain control (AGC) function. The PPG signal sampling rate was 60 Hz and the quantization resolution was 8 bits. The optical sensor with both red (wavelength 660 nm) and infrared (wavelength 880 nm) light sources were used. The collected signal was sent via Bluetooth to a PC, where SBP estimations were performed. This study was performed at the Clinical Department of Geriatrics of the Lithuanian University of Health Sciences. Nineteen subjects took part in the research (12 women and 7 men), including 6 persons aged 70–77, 11 persons aged 82–85, and 2 persons aged 88 years old with body mass index BMI = 22.4 ± 1.3. All subjects, but two, used blood pressure‐regulating drugs. Five subjects had ischemic heart disease. SBP measurements and PPG records were done for each subject 5 days in a row (one per day), at any convenient time for the subject. The subjects sat with their left hand placed on a support and a PPG sensor placed on their left‐hand finger, as shown in Figure 14a. The PPG signal recording was always done first using the pulse oximeter. This procedure took 5 min. The SBP measurements were done immediately after that, using the blood pressure meter, as shown in Figure 14b. Standard cuff type for an arm circumference ranging from 220 to 320 mm was used. Since measurements were performed in a calm state, the records are without movement artefacts. Appl. Sci. 2019, 9, x FOR PEER REVIEW 10 of 16

Figure 12. Estimated SBPei blood pressure values (curved line) and SBPmdj median values (stepwise

line), calculated from SBPei in BPEW.

When blood pressure is measured a few times per day at home or in the hospital, it is assumed that it remains constant between measurements. 𝑆𝐵𝑃 ∑ 𝑆𝐵𝑃 . (6)

Appl. Sci.Therefore,2019, 9, 2236 in this work, for a 5‐min PPG period, we calculate one SBPe value using Equation10 of (6), 16 which is equal to the mean of the medians (see Figure 13):

FigureFigure 13.13. Estimated SBPee blood pressure value (dot line) and SBPmdj inin BPEW BPEW (stepwise (stepwise line). line).

2.4. Experiment Experiment Setup A commercial blood pressure meter with a cuff cuff (OMRON M1) was used as the gold standard in this workwork to to measure measure SBP SBP on theon leftthe brachialleft brachial artery. artery. The PPG The signal PPG was signal collected was usingcollected the CMS50FWusing the CMS50FWPulse Oximeter Pulse without Oximeter automatic without automatic gain control gain (AGC) control function. (AGC) Thefunction. PPG signalThe PPG sampling signal sampling rate was rate60 Hz was and 60 the Hz quantization and the resolutionquantization was resolution 8 bits. The was optical 8 sensorbits. The with optical both red sensor (wavelength with both 660 nm)red (wavelengthand infrared (wavelength660 nm) and 880infrared nm) light (wavelength sources were 880 nm) used. light The sources collected were signal used. was The sent collected via Bluetooth signal wasto a PC,sent where via Bluetooth SBP estimations to a PC, where were performed. SBP estimations were performed. This study was performed at the Clinical Department of Geriatrics of the Lithuanian University of Health Sciences. Nineteen Nineteen subjects subjects took took part part in in the the research research (12 (12 women women and and 7 7men), men), including including 6 persons6 persons aged aged 70–77, 70–77, 11 11 persons persons aged aged 82–85, 82–85, and and 2 2 persons persons aged aged 88 88 years years old old with with body body mass index BMI = 22.4 1.3. All subjects, but two, used blood pressure-regulating drugs. Five subjects had BMI = 22.4 ± 1.3. All subjects, but two, used blood pressure‐regulating drugs. Five subjects had ischemic heart disease. SBP measurements andand PPGPPG recordsrecords were were done done for for each each subject subject 5 5 days days in in a rowa row (one (one per per day), day), at atany any convenient convenient time time for thefor the subject. subject. The The subjects subjects sat with sat with their their left hand left hand placed placed on a support on a support and a PPGand asensor PPG sensor placed placed on their on left-hand their left finger,‐hand asfinger, shown as shown in Figure in 14Figurea. The 14a. PPG The signal PPG recording signal recording was always was alwaysdone first done using first the using pulse the oximeter. pulse oximeter. This procedure This procedure took 5 took min. 5 The min. SBP The measurements SBP measurements were donewere doneimmediately immediately after that, after using that, theusing blood the pressureblood pressure meter, asmeter, shown as inshown Figure in 14 Figureb. Standard 14b. Standard cuff type cuff for typean arm for circumference an arm circumference ranging from ranging 220 to from 320 mm220 wasto 320 used. mm Since was measurementsused. Since measurements were performed were in Appl.performeda calm Sci. state, 2019 ,in 9 the, ax FORcalm records PEER state, REVIEW are the without records movement are without artefacts. movement artefacts. 11 of 16

(a) (b)

Figure 14. a b Figure 14. ((a)) PPG PPG signal signal recording recording posture; ( b) SBP measurements posture. The experimental protocol of the study was approved by the Kaunas Regional Biomedical Research The experimental protocol of the study was approved by the Kaunas Regional Biomedical Ethics Committee (permission No. BE-2-53), allowing the use of a pulse oximeter CMS50FW at the Research Ethics Committee (permission No. BE‐2‐53), allowing the use of a pulse oximeter CMS50FW Clinical Department of Geriatrics. According to the Association for the Advancement of Medical at the Clinical Department of Geriatrics. According to the Association for the Advancement of Instrumentation (AAMI) protocol [49], the blood pressure values were measured by experienced Medical Instrumentation (AAMI) protocol [49], the blood pressure values were measured by medical staﬀ. experienced medical staff. All subjects were informed about the performed research and their written consent was obtained. All subjects were informed about the performed research and their written consent was obtained.

3. Results Nineteen subjects took part in the research. For each of them, 5 SBP values were measured using the oscillometric technique, and 5 SBP values were estimated using the proposed method. Therefore, the measured values vector x and estimated values vector y contains N = 95 values. The difference between measured and estimated SBP was calculated for each examination, including the mean difference (MD), the standard deviation (SD), and the relative error (er), in order to evaluate the performance of the proposed method (Table 1):

Table 1. Performance of the proposed method.

r (p < 0.001) MD SD er,  0.86 −0.043 6.79 0.025 ± 5.44 A Pearson’s correlation value shows that the estimated and measured SBP values correlated well (r = 0.86, p < 0.001). This statistically significant tight correlation shows that the proposed method is suitable for continuous systolic blood pressure estimation using PPG recording devices with a low sampling rate (up to 60 Hz) and low resolution (up to 8 bits). The measured MD and SD values are within the requirements of the AAMI standard (MD = ±5 mmHg and SD = ±8 mmHg) [49], since the mean difference MD = −0.043 and standard deviation SD = 6.79, though the study does not comply with the AAMI protocol, which requires more subjects with different blood pressures and auscultatory technique as a reference method to evaluate the accuracy of the proposed method. The Bland–Altman plot, as discussed in Reference [50], is commonly used for the comparison of clinical results obtained by using different methods. In Figure 15, we present the agreements of the measured and estimated blood pressure values. The horizontal part of the plot corresponds to the mean values of the measured and estimated SBP, while the vertical axis shows the difference between these values. The mean difference of all the subjects is calculated and represented as a horizontal line, and this horizontal line means better results if it is closer to zero. Appl. Sci. 2019, 9, 2236 11 of 16

3. Results Nineteen subjects took part in the research. For each of them, 5 SBP values were measured using the oscillometric technique, and 5 SBP values were estimated using the proposed method. Therefore, the measured values vector x and estimated values vector y contains N = 95 values. The diﬀerence between measured and estimated SBP was calculated for each examination, including the mean diﬀerence (MD), the standard deviation (SD), and the relative error (er), in order to evaluate the performance of the proposed method (Table1):

Table 1. Performance of the proposed method.

r (p < 0.001) MD SD er,% 0.86 0.043 6.79 0.025 5.44 − ±

A Pearson’s correlation value shows that the estimated and measured SBP values correlated well (r = 0.86, p < 0.001). This statistically significant tight correlation shows that the proposed method is suitable for continuous systolic blood pressure estimation using PPG recording devices with a low sampling rate (up to 60 Hz) and low resolution (up to 8 bits). The measured MD and SD values are within the requirements of the AAMI standard (MD = 5 mmHg and SD = 8 mmHg) [49], since the mean difference MD = 0.043 and standard deviation ± ± − SD = 6.79, though the study does not comply with the AAMI protocol, which requires more subjects with different blood pressures and auscultatory technique as a reference method to evaluate the accuracy of the proposed method. The Bland–Altman plot, as discussed in Reference [50], is commonly used for the comparison of clinical results obtained by using different methods. In Figure 15, we present the agreements of the measured and estimated blood pressure values. The horizontal part of the plot corresponds to the mean values of the measured and estimated SBP, while the vertical axis shows the difference between these values. The mean difference of all the subjects is calculated and represented as a horizontal line, andAppl. this Sci. horizontal 2019, 9, x FOR line PEER means REVIEW better results if it is closer to zero. 12 of 16

FigureFigure 15. 15.Bland–Altman Bland–Altman plot plot presenting presenting the values of the the difference diﬀerence (SBP (SBPe–SBPe–SBPm)m as) as a function a function of the of the

meanmean of ofSBP SBPm mand andSBP SBPee..

TheThe comparison comparison of estimated estimated and and measured measured values values in Figure in Figure 15 shows 15 showsthat the that error the bias, error −0.043 bias, 0.043 mmHg between the gold standard and measured SBP values, is close to zero, while the values − mmHg between the gold standard and measured SBP values, is close to zero, while the values of the ofSD the of SD bias of 6.79 bias mmHg 6.79 mmHg for SBP for are SBP very are similar very similar to the results to the of results other of authors. other authors. The systematic The systematic trends trendsin Figure in Figure 15 show, 15 show,that there that are there some are errors some caused errors by caused the method by the itself, method most itself, likely mostdue to likely rounding due to roundingprocedures. procedures. TableTable2 shows 2 shows the the comparison comparison of the of proposedthe proposed SBP estimationSBP estimation method method with with other other authors’ authors’ results: results:

Table 2. Evaluation of the performance of the systolic blood pressure estimation methods (MD— mean difference, SD—standard deviation, SPS—samples per second).

Systolic Blood Pressure Method Performance Methods MD SD SPS Proposed method −0.043 6.79 60 Hz Ding [15] 0.37 5.21 1 KHz Lin [13] 3.22 8.02 1 KHz Yan [25] −0.37 4.3 200 Hz Kurylyak [22] 3.24 3.47 125 Hz Choudhury [20] 0.78 13.1 100 Hz Khalid [19] −1.1 5.7 100 Hz Xing [21] −1.67 2.46 100 Hz Mousavi [10] Method 1 0.25 6.7 125 Hz Method 2 0.66 7.5 125 Hz Method 3 0.19 4.17 125 Hz Method 4 0.2 4.73 125 Hz

As can be seen in Table 2, the proposed SBP estimation method gives similar or better results even when the sampling rate is almost two times lower.

4. Discussion

4.1. Advantage The original PPG based method for the continuous estimation of systolic blood pressure using ADC with a 60 Hz sampling rate and 8‐bit resolution is proposed in this paper. The main advantage of the method is its high accuracy, which is achieved using lower sampling rates and less capable hardware than other methods. Appl. Sci. 2019, 9, 2236 12 of 16

Table 2. Evaluation of the performance of the systolic blood pressure estimation methods (MD—mean diﬀerence, SD—standard deviation, SPS—samples per second).

Systolic Blood Pressure Method Performance Methods MD SD SPS Proposed method 0.043 6.79 60 Hz − Ding [15] 0.37 5.21 1 KHz Lin [13] 3.22 8.02 1 KHz Yan [25] 0.37 4.3 200 Hz − Kurylyak [22] 3.24 3.47 125 Hz Choudhury [20] 0.78 13.1 100 Hz Khalid [19] 1.1 5.7 100 Hz − Xing [21] 1.67 2.46 100 Hz − Mousavi [10] Method 1 0.25 6.7 125 Hz Method 2 0.66 7.5 125 Hz Method 3 0.19 4.17 125 Hz Method 4 0.2 4.73 125 Hz

As can be seen in Table2, the proposed SBP estimation method gives similar or better results even when the sampling rate is almost two times lower.

4. Discussion

4.1. Advantage The original PPG based method for the continuous estimation of systolic blood pressure using ADC with a 60 Hz sampling rate and 8-bit resolution is proposed in this paper. The main advantage of the method is its high accuracy, which is achieved using lower sampling rates and less capable hardware than other methods. Since the resolution of SBP estimation is not linear (because the dependency between DPW and SBP is not linear, as explained in “Systolic Blood Pressure Estimation Model” section), the mean difference MD and standard deviation SD is used to evaluate the accuracy of the proposed method. The mean difference MD = 0.043 and standard deviation SD = 6.79 matches, or even improve, the − results of the known methods (see Table2). Though the proposed method does not meet the accuracy requirements of the AAMI standard (due to protocol violations), the MD and SD values are within the requirements of this standard (MD = 5 mmHg and SD = 8 mmHg). This shows the great potential ± ± of the method and additional experiments are planned in the future work to evaluate full compliance with the AAMI standard. There are many authors performing state-of-the-art research activities in the field of continuous and cuffless blood pressure monitoring. They admit that continuous blood pressure monitoring is a very important activity and propose various non-invasive PPG based methods to solve this problem. However, the main attention is given to the verification of the accuracy of the proposed methods and comparison with related works, while very little or no attention is paid to the possibilities to implement these methods in wearable systems [10,13,15,19,20,22,25]. There are known wearable (portable) blood pressure measurement systems based on smartphones [26,27], allowing to make measurements at any selected moment. But it is not continuous monitoring, since the idea of the continuous monitoring of physiological parameters (including blood pressure) is that the user does not need to pay any attention to registering those parameters. Therefore, the main distinction of the proposed method in comparison with other state-of-the-art approaches is that this method can be successfully applied for the implementation of a real life wearable continuous SBP monitoring systems. The method is successfully tested using a prototype system based on the widely available and quite cheap CMS50FW Pulse Oximeter, which is capable of transferring PPG values via Bluetooth to a mobile phone or any other remote device. In order to reduce the number Appl. Sci. 2019, 9, 2236 13 of 16 and complexity of calculations (and energy consumption), the proposed method analyzes the rising edge of the PPG pulse wave only. The rising edge is less than 1/3 of the duration of the PPG pulse wave; therefore, this significantly reduces the calculations compared to the methods which process the entire PPG pulse wave. This is relevant for the development of wearable monitoring systems which have very limited processing and energy resources.

4.2. Limitation The proposed method allows us to estimate systolic blood pressure only, and this can be done when subjects are in a calm state. Ideally, wearable devices should allow us to record systolic and diastolic blood pressure in diﬀerent situations (sitting, walking, driving, etc.) without distracting from daily activities. In order to make such systems, the proposed method should be upgraded by adding an additional processing stage, which would allow us to eliminate movement-induced noise, as well as noise caused by a raised or lowered arm.

5. Conclusions The possibility of estimating systolic blood pressure using a PPG collecting device with a low sampling rate and resolution was investigated in this paper. We believe that such a wearable estimation system will enable the on-demand home monitoring of systolic blood pressure as an important vital sign, resulting in an overall reduction in cardiovascular morbidity and mortality. Our results show that PPG recording devices with a low sampling rate (60 Hz) and low resolution (8-bit) ADC can be successfully applied for systolic blood pressure estimation using the proposed systolic blood pressure estimation method based on the pulse wave rising edge steepness.

Author Contributions: Conceptualization, R.G.; Formal analysis, R.G.; Investigation, R.G., V.L. and G.D.; Methodology, R.G. and A.L.; Project administration, E.K. and A.J.; Resources, G.D. and A.J.; Supervision, E.K. and V.L.; Validation, A.L.; Writing—original draft, A.L. and A.J. Funding: This research was funded by the Research Council of Lithuania, grant number SEN-06/2016. Conﬂicts of Interest: The authors declare no conﬂict of interest.

References

1. Jacob, V.; Chattopadhyay, S.K.; Proia, K.K.; Hopkins, D.P.; Reynolds, J.; Thota, A.B.; Jones, C.D.; Lackland, D.T.; Rask, K.J.; Pronk, N.P.; et al. Economics of self-measured blood pressure monitoring: A community guide systematic review. Am. J. Prev. Med. 2017, 53, e105–e113. [CrossRef][PubMed] 2. Sankar, J.; Mahesh, N.; Sharma, P.; Sankar, L.; Balasubramanian, A. Role of ambulatory blood pressure monitoring in chronic hypertensive patients on antihypertensive therapy-a cross-sectional study. J. Clin. Diagn. Res. 2018, 12, OC01–OC04. [CrossRef] 3. Liu, Z.; Zhao, Y.; Lu, F.; Zhang, H.; Diao, Y. Day-by-day variability in self-measured blood pressure at home: Effects on carotid artery atherosclerosis, brachial flow-mediated dilation, and endothelin-1 in normotensive and mild–moderate hypertensive individuals. Blood Press. Monit. 2013, 18, 316–325. [CrossRef] 4. Toriumi, S.; Hoshide, S.; Nagai, M.; Kario, K. Day-to-day blood pressure variability as a phenotype in a high-risk patient: Letters to the Editor. Geriatr. Gerontol. Int. 2014, 14, 1005–1006. [CrossRef] 5. Pan, J.; Zhang, Y. Improved blood pressure estimation using photoplethysmography based on ensemble method. In Proceedings of the 2017 14th International Symposium on Pervasive Systems, Algorithms and Networks & 2017 11th International Conference on Frontier of Computer Science and Technology & 2017 Third International Symposium of Creative Computing (ISPAN-FCST-ISCC), Exeter, UK, 21–23 June 2017; pp. 105–111. 6. Lewington, S.L.; Clarke, R.; Qizilbash, N.; Peto, R.; Collins, R. Age-specific relevance of usual blood pressure to vascular mortality: A meta-analysis of individual data for one million adults in 61 prospective studies. Lancet 2002, 360, 1903–1913. [PubMed] 7. Asia Pacific Cohort Studies Collaboration. Blood pressure indices and cardiovascular disease in the asia pacific region: A pooled analysis. Hypertension 2003, 42, 69–75. [CrossRef] Appl. Sci. 2019, 9, 2236 14 of 16

8. Lawes, C.M.; Hoorn, S.V.; Rodgers, A. Global burden of blood-pressure-related disease, 2001. Lancet 2008, 371, 1513–1518. [CrossRef] 9. Bose, S.S.N.; Kandaswamy, A. Sparse representation of photoplethysmogram using K-SVD for cuffless estimation of arterial blood pressure. In Proceedings of the 2017 4th International Conference on Advanced Computing and Communication Systems (ICACCS), Coimbatore, India, 6–7 January 2017; pp. 1–5. 10. Mousavi, S.S.; Firouzmand, M.; Charmi, M.; Hemmati, M.; Moghadam, M.; Ghorbani, Y. Blood pressure estimation from appropriate and inappropriate PPG signals using A whole-based method. Biomed. Signal Process. Control 2019, 47, 196–206. [CrossRef] 11. Taylor, B.C.; Wilt, T.J.; Welch, H.G. Impact of diastolic and systolic blood pressure on mortality: Implications for the definition of “normal”. J. Gen. Intern. Med. 2011, 26, 685–690. [CrossRef][PubMed] 12. Rothwell, P.M.; Howard, S.C.; Dolan, E.; O’Brien, E.; Dobson, J.E.; Dahlöf, B.; Sever, P.S.; Poulter, N.R. Prognostic significance of visit-to-visit variability, maximum systolic blood pressure, and episodic hypertension. Lancet 2010, 375, 895–905. [CrossRef] 13. Lin, W.-H.; Wang, H.; Samuel, O.W.; Li, G. Using a new PPG indicator to increase the accuracy of PTT-based continuous cuffless blood pressure estimation. In Proceedings of the 2017 39th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), Seogwipo, Korea, 11–15 July 2017; pp. 738–741. 14. Khan, N.; Mikael Eklund, J. A Highly integrated computing platform for continuous, non-invasive bp estimation. In Proceedings of the 2018 IEEE Canadian Conference on Electrical & Computer Engineering (CCECE), Quebec City, QC, Canada, 13–16 May 2018; pp. 1–5. 15. Ding, X.-R.; Zhang, Y.-T.; Liu, J.; Dai, W.-X.; Tsang, H.K. Continuous cuffless blood pressure estimation using pulse transit time and photoplethysmogram intensity ratio. IEEE Trans. Biomed. Eng. 2016, 63, 964–972. [CrossRef] 16. Zhang, J.M.; Wei, P.F.; Li, Y. A LabVIEW based measure system for pulse wave transit time. In Proceedings of the 2008 International Conference on Technology and Applications in Biomedicine, Shenzhen, China, 18 July 2008; pp. 477–480. 17. Liu, S.-H.; Cheng, D.-C.; Su, C.-H. A cuffless blood pressure measurement based on the impedance plethysmography technique. Sensors 2017, 17, 1176. [CrossRef] 18. Liu, H.; Ivanov, K.; Wang, Y.; Wang, L. Toward a smartphone application for estimation of pulse transit time. Sensors 2015, 15, 27303–27321. [CrossRef] 19. Khalid, S.G.; Zhang, J.; Chen, F.; Zheng, D. Blood pressure estimation using photoplethysmography only: comparison between different machine learning approaches. J. Healthc. Eng. 2018, 2018, 1–13. [CrossRef] 20. Choudhury, A.D.; Banerjee, R.; Sinha, A.; Kundu, S. Estimating blood pressure using Windkessel model on photoplethysmogram. In Proceedings of the 2014 36th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, Chicago, IL, USA, 26–30 August 2014; pp. 4567–4570. 21. Xing, X.; Sun, M. Optical blood pressure estimation with photoplethysmography and FFT-based neural networks. Biomed. Opt. Express 2016, 7, 3007–3020. [CrossRef] 22. Kurylyak, Y.; Barbe, K.; Lamonaca, F.; Grimaldi, D.; Van Moer, W. Photoplethysmogram-based Blood pressure evaluation using Kalman filtering and Neural Networks. In Proceedings of the 2013 IEEE International Symposium on Medical Measurements and Applications (MeMeA), Gatineau, QC, Canada, 4–5 May 2013; pp. 170–174. 23. Kao, Y.-H.; Chao, P.C.-P.; Tu, T.-Y.; Chiang, K.-Y.; Wey, C.-L. A new cuffless optical sensor for blood pressure measuring with self-adaptive signal processing. In Proceedings of the 2016 IEEE SENSORS, Orlando, FL, USA, 30 October–3 November 2016; pp. 1–3. 24. Cohen, Z.; Haxha, S. Optical-based sensor prototype for continuous monitoring of the blood pressure. IEEE Sens. J. 2017, 17, 4258–4268. [CrossRef] 25. Yan, Y.S.; Zhang, Y.T. Noninvasive estimation of blood pressure using photoplethysmographic signals in the period domain. In Proceedings of the 2005 IEEE Engineering in Medicine and Biology 27th Annual Conference, Shanghai, China, 17–18 January 2006; pp. 3583–3584. 26. Gaurav, A.; Maheedhar, M.; Tiwari, V.N.; Narayanan, R. Cuff-less PPG based continuous blood pressure monitoring—A smartphone based approach. In Proceedings of the 2016 38th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), Orlando, FL, USA, 16–20 August 2016; pp. 607–610. Appl. Sci. 2019, 9, 2236 15 of 16

27. Matsumura, K.; Rolfe, P.; Toda, S.; Yamakoshi, T. Cuffless blood pressure estimation using only a smartphone. Sci. Rep. 2018, 8, 7298. [CrossRef] 28. Tobola, A.; Streit, F.J.; Espig, C.; Korpok, O.; Sauter, C.; Lang, N.; Schmitz, B.; Hofmann, C.; Struck, M.; Weigand, C.; et al. Sampling rate impact on energy consumption of biomedical signal processing systems. In Proceedings of the 2015 IEEE 12th International Conference on Wearable and Implantable Body Sensor Networks (BSN), Cambridge, MA, USA, 9–12 June 2015; pp. 1–6. 29. Gircys, R.; Liutkevicius, A.; Vrubliauskas, A.; Kazanavicius, E. Blood pressure estimation accoording to photoplethysmographic signal steepness. Inf. Technol. Control 2015, 44, 443–450. [CrossRef] 30. Harada, A.; Okada, T.; Niki, K.; Chang, D.; Sugawara, M. On-line noninvasive one-point measurements of pulse wave velocity. Heart Vessel. 2002, 17, 61–68. [CrossRef][PubMed] 31. Diourté, B.; Siché, J.-P.; Comparat, V.; Baguet, J.-P.; Mallion, J.-M. Study of arterial blood pressure by a Windkessel-type model: Influence of arterial functional properties. Comput. Methods Programs Biomed. 1999, 60, 11–22. [CrossRef] 32. Sugawara, M.; Niki, K.; Furuhata, H.; Ohnishi, S.; Suzuki, S. Relationship between the pressure and diameter of the carotid artery in humans. Heart Vessel. 2000, 15, 49–51. [CrossRef] 33. Elgendi, M. Detection of c, d, and e waves in the acceleration photoplethysmogram. Comput. Methods Programs Biomed. 2014, 117, 125–136. [CrossRef][PubMed] 34. Song, H.-S.; Lehrer, P.M. The effects of specific respiratory rates on heart rate and heart rate variability. Appl. Psychophysiol. Biofeedback 2003, 28, 13–23. [CrossRef][PubMed] 35. AL-Khalidi, F.Q.; Saatchi, R.; Burke, D.; Elphick, H.; Tan, S. Respiration rate monitoring methods: A review. Pediatric Pulmonol. 2011, 46, 523–529. [CrossRef][PubMed] 36. Charlton, P.H.; Villarroel, M.; Salguiero, F. Waveform analysis to estimate respiratory rate. In Secondary Analysis of Electronic Health Records; Springer International Publishing: Cham, Switzerland, 2016; pp. 377–390. ISBN 978-3-319-43740-8. 37. Jennings, J.R.; McKnight, J.D.; Molen, M. Phase-sensitive interaction of cardiac and respiratory timing in humans. Psychophysiology 1996, 33, 514–521. [CrossRef] 38. Sadrawi, M.; Shieh, J.-S.; Fan, S.Z.; Lin, C.H.; Haraikawa, K.; Chien, J.C.; Abbod, M.F. Intermittent blood pressure prediction via multiscale entropy and ensemble artificial neural networks. In Proceedings of the 2016 IEEE EMBS Conference on Biomedical Engineering and Sciences (IECBES), Kuala Lumpur, Malaysia, 4–8 December 2016; pp. 356–359. 39. Bhoi, A.K.; Sarkar, S.; Mishra, P.; Savita, G. Pre-processing of PPG signal with performance based methods. Int. J. Comput. Appl. 2012, 4, 251–256. 40. Lee, C.M.; Zhang, Y.T. Reduction of motion artifacts from photoplethysmographic recordings using a wavelet denoising approach. In Proceedings of the IEEE EMBS Asian-Pacific Conference on Biomedical Engineering, Kyoto, Japan, 20–22 October 2003; pp. 194–195. 41. Boloursaz Mashhadi, M.; Asadi, E.; Eskandari, M.; Kiani, S.; Marvasti, F. Heart rate tracking using wrist-type photoplethysmographic (PPG) signals during physical exercise with simultaneous accelerometry. IEEE Signal Process. Lett. 2016, 23, 227–231. [CrossRef] 42. Wang, D.; Zhang, D.; Lu, G. A robust signal preprocessing framework for wrist pulse analysis. Biomed. Signal Process. Control 2016, 23, 62–75. [CrossRef] 43. Alian, A.A.; Shelley, K.H. Photoplethysmography: Analysis of the pulse oximeter waveform. In Monitoring Technologies in Acute Care Environments; Ehrenfeld, J.M., Cannesson, M., Eds.; Springer: New York, NY, USA, 2014; pp. 165–178. ISBN 978-1-4614-8556-8. 44. Kavsao˘glu,A.R.; Polat, K.; Bozkurt, M.R. An innovative peak detection algorithm for photoplethysmography signals: An adaptive segmentation method. Turk. J. Electr. Eng. Comput. Sci. 2016, 24, 1782–1796. [CrossRef] 45. Kazanavicius, E.; Gircys, R.; Vrubliauskas, A.; Lugin, S. Mathematical methods for determining the foot point of the arterial pulse wave and evaluation of proposed methods. Inf. Technol. Control 2005, 34, 29–36. 46. Banerjee, R.; Ghose, A.; Dutta Choudhury, A.; Sinha, A.; Pal, A. Noise cleaning and Gaussian modeling of smart phone photoplethysmogram to improve blood pressure estimation. In Proceedings of the 2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), South Brisbane, Queensland, Australia, 19–24 April 2015; pp. 967–971. 47. Akl, T.J.; Wilson, M.A.; Ericson, M.N.; Coté, G.L. Quantifying tissue mechanical properties using photoplethysmography. Biomed. Opt. Express 2014, 5, 2362–2375. [CrossRef][PubMed] Appl. Sci. 2019, 9, 2236 16 of 16

48. Datta, S.; Banerjee, R.; Choudhury, A.D.; Sinha, A.; Pal, A. Blood pressure estimation from photoplethysmogram using latent parameters. In Proceedings of the 2016 IEEE International Conference on Communications (ICC), Kuala Lumpur, Malaysia, 22–27 May 2016; pp. 1–7. 49. White, W.B.; Berson, A.S.; Robbins, C.; Jamieson, M.J.; Prisant, L.M.; Roccella, E.; Sheps, S.G. National standard for measurement of resting and ambulatory blood pressures with automated sphygmomanometers. Hypertension 1993, 21, 504–509. [CrossRef][PubMed] 50. Bland, J.M.; Altman, D.G. Statistical methods for assessing agreement between two methods of clinical measurement. Int. J. Nurs. Stud. 2010, 47, 931–936. [CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).