See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/5860160
`
`Can Photoplethysmography Variability Serve as
`an Alternative Approach to Obtain Heart Rate
`Variability Information?
`
`Article in Journal of Clinical Monitoring and Computing · March 2008
`
`DOI: 10.1007/s10877-007-9103-y · Source: PubMed
`
`CITATIONS
`92
`
`7 authors, including:
`
`READS
`258
`
`Sheng Lu
`State University of New York
`
`9 PUBLICATIONS 303 CITATIONS
`
`He Zhao
`Stony Brook University
`
`9 PUBLICATIONS 174 CITATIONS
`
`SEE PROFILE
`
`SEE PROFILE
`
`Kirk Harry Shelley
`Yale University
`
`79 PUBLICATIONS 1,174 CITATIONS
`
`hc ki
`Worcester Polytechnic Institute
`
`119 PUBLICATIONS 1,503 CITATIONS
`
`SEE PROFILE
`
`SEE PROFILE
`
`All content following this page was uploaded by Kirk Harry Shelley on 27 December 2016.
`
`The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
`and are linked to publications on ResearchGate, letting you access and read them immediately.
`
`APPLE 1013
`
`1
`
`

`

`Journal of Clinical Monitoring and Computing
`DOI: 10.1007/s10877-007-9103-y
`
`Ó Springer 2007
`
`CAN PHOTOPLETHYSMOGRAPHY VARIABILITY
`SERVE AS AN ALTERNATIVE APPROACH
`TO OBTAIN HEART RATE VARIABILITY
`INFORMATION?
`Sheng Lu, PhD1, He Zhao, MS1, Kihwan Ju, PhD1,
`Kunsoo Shin, PhD2, Myoungho Lee, PhD3,
`Kirk Shelley, PhD4 and Ki H. Chon, PhD1
`
`From the 1Department of Biomedical Engineering, State University
`of New York, SUNY@ Stony Brook HSC T18, Rm. 030, Stony
`Brook, NY, 11794-8181, USA; 2Samsung Advanced Institute of
`Technology, Yongin-Si, South Korea; 3Department of Electrical
`and Electronics Engineering, Yonsei University, Seoul, South
`Korea; 4Department of Anesthesia, Yale University, New Haven,
`CT, USA.
`
`Received 17 August 2007. Accepted for publication 24 October
`2007.
`
`Address correspondence to K. H. Chon, Department of Bio-
`medical Engineering, State University of New York, SUNY@
`Stony Brook HSC T18, Rm. 030, Stony Brook, NY, 11794-8181,
`USA.
`E-mail: ki.chon@sunysb.edu
`
`Lu S, Zhao H, Ju K, Shin KS, Lee MH, Shelley K, Chon KH. Can
`photoplethysmography variability serve as an alternative approach to
`obtain heart rate variability information?
`
`J Clin Monit Comput 2007
`
`ABSTRACT. Heart rate variability (HRV), extracted from an
`electrocardiogram, is known to be a noninvasive indicator
`reflecting the dynamic interplay between perturbations to
`cardiovascular function and the dynamic response of the
`cardiovascular
`regulatory system. Photoplethysmography
`(PPG) is a noninvasive method to monitor arterial oxygen
`saturation on a
`continuous
`basis. Given the
`rich
`cardiovascular
`information in the PPG signal, and the
`ubiquity and simplicity of pulse oximetry, we
`are
`investigating
`the
`feasibility
`of
`acquiring
`dynamics
`pertaining to the autonomic nervous
`system from PPG
`waveforms. To do this, we are quantifying PPG variability
`(PPGV). Detailed algorithmic approaches
`for extracting
`accurate PPGV signals are presented. We compare PPGV
`to HRV by computing time and frequency domain
`parameters often associated with HRV measurements, as
`well as approximate entropy calculations. Our
`results
`demonstrate that
`the parameters of PPGV are highly
`correlated with the parameters of HRV. Thus, our results
`indicate that PPGV could be used as an alternative
`measurement of HRV.
`KEY WORDS. autonomic nervous system, heart rate variability,
`pulse oximeter.
`
`INTRODUCTION
`
`As a non-invasive means to monitor arterial oxygen sat-
`uration (SaO2) on a continuous basis, pulse oximetry is a
`well-established technology based on photoplethysmog-
`raphy that has become one of the most commonly used
`patient monitoring devices during anesthesia and in
`intensive care units. Given the ubiquity and simplicity of
`pulse oximetry, it is desirable to maximize its potential by
`exploring additional measurements we can derive from
`the pulse oximeter. In the present study, our goal was to
`determine if variations in the PPG signal can be used in
`lieu of heart rate variability (HRV) to extract dynamics
`pertaining to the autonomic nervous system.
`HRV reflects the dynamic interplay between pertur-
`bations to cardiovascular function and the dynamic re-
`sponse of cardiovascular regulatory systems. It is clear
`from a work by Akselrod et al. [2] in the early 1980s, as
`well as numerous publications since, that maintaining
`ANS balance is
`important
`for cardiovascular health.
`Numerous studies have shown that altered variability in
`
`2
`
`

`

`of using the PPG instead of the ECG is that this allows
`multi-functionality of the pulse oximetry, which would
`simplify many facets of monitoring systems,
`reduce
`healthcare costs, provide a more compact system, and
`the home health care monitoring more attainable.
`
`METHOD
`
`Ten healthy subjects were involved in the study
`(26 ± 7.47 years). Twenty minutes (10 min in the upright
`position and 10 min in the supine position) of PPG
`waveform data (pulse oximeter module MP506, Nellcor
`Puritan Bennett) and ECG signal (HP 78354A patient
`monitor) were collected simultaneously and digitized with
`a sampling rate of 400 Hz (Powerlab 4SP, ADInstru-
`ments). HRV was obtained from ECG signals after R-
`wave peaks were detected,
`followed by cubic spline
`interpolation.
`The PPG waveforms were denoised and detrended
`with empirical mode decomposition method (EMD) [12].
`The EMD is based on a concept that any signal x(t) can be
`decomposed into a finite number of
`‘‘intrinsic mode
`functions’’ (IMFs):
`IMFi þ rn
`
`Xn
`
`i¼1
`
`xðtÞ ¼
`
`where rn is a residue. An IMF is a function that satisfies
`two conditions:
`
`(1) The number of extrema and the number of zero
`crossings must either be equal or differ at most by one.
`(2) At any point the mean value of the envelope defined
`by the local maxima and the envelope defined by the
`local minima is zero.
`
`IMFs represent oscillatory modes embedded within
`data where each IMF involves only one mode of oscil-
`lation. The EMD algorithm has been widely applied
`[7, 8], and thus, we only briefly summarize the following
`four steps:
`
`(1) Given the signal x(t), identify the successive extrema
`of x(t). Extract the upper and lower envelopes by
`interpolation and compute the average, and denote it
`as m1:
`
`m1 ¼ emaxðtÞ þe minðtÞ
`
`2
`
`ð1Þ
`
`(2) Subtract the envelope mean signal from the signal
`h1ðtÞ ¼ xðtÞ m 1
`ð2Þ
`
`Journal of Clinical Monitoring and Computing
`
`the cardiovascular system is associated with a range of
`cardiovascular diseases and increased mortality [1, 9].
`Cardiovascular variables such as heart rate, arterial blood
`pressure, and stroke volume, as well as lung volume,
`fluctuate on a beat-to-beat basis. The variability in car-
`diovascular signals reflects the homeostatic interplay be-
`tween perturbations in cardiovascular functions and the
`dynamic responses of the cardiovascular regulatory sys-
`tems [1, 6]. Thus, the normal heart rate is determined by
`many interacting systems, including the nervous system,
`baro- and chemoreceptors, local feedback loops in the
`heart, and several hormonal systems. These systems work
`on different
`time scales, and the complexity of
`this
`regulation is reflected in the apparently random fluctu-
`ations in heart rate [5]. There is considerable interest in
`these fluctuations because their simple statistical measures
`such as the standard deviation of the interbeat intervals
`(the R–R-intervals), have been shown to be some of the
`strongest
`independent predictors of mortality after
`myocardial infarction [10]. Moreover, other techniques
`such as spectral analysis and nonlinear analysis of the
`R–R-intervals of heart rate have been widely used in
`HRV studies, and on some occasions they have been
`shown to discriminate between subjects with different
`cardiac conditions as well as to predict mortality in some
`groups of patients [3, 9].
`To date, only a few studies have demonstrated the use
`of PPG variability as a noninvasive measure of
`the
`dynamics pertaining to the autonomic nervous system,
`but none of these studies have provided detailed and
`direct quantitative comparison between the PPG and
`ECG signals [4, 11]. These studies were mainly based on
`the use of the pulse oximeter to examine specific car-
`diovascular problems that are associated with the auto-
`nomic nervous
`system. For example, one study has
`noted that in diabetic autonomic neuropathy patients,
`the rate of fall in percentage of oxygen saturation was
`significantly lower, less intense, and with delayed sub-
`sequent recovery compared to normal subjects [11]. In a
`separate study, recordings of the pulse oximeter wave-
`forms during the Valsava maneuver were useful in rapid
`pre-operative identification of patients who have an
`autonomic neuropathy [4]. It is not surprising that only
`scant reports exist of using PPG variability as a nonin-
`vasive measure of the dynamics underlying the auto-
`nomic nervous system, since the rich information the
`pulse oximeter provides has only recently been appre-
`ciated. As the rise and fall of the PPG signal reflects the
`fluctuations of the heart beats, it is possible that PPG
`variability does reflect the dynamics of the autonomic
`nervous
`system. Our
`results
`show that
`this very
`hypothesis
`is feasible and that we can obtain similar
`information to that from HRV analysis. The significance
`
`3
`
`

`

`Lu et al.: Can Photoplethysmography Variability Serve as an Alternative Approach to Obtain Heart Rate Variability Information?
`
`Third, perform a second moving average of the signal c:
`þM
`
`cðjÞ
`
`Xi
`
`dði M þ 1Þ ¼
`
`1
`2M þ 1
`j¼iM
`where i = M, M + 1, M + 2,...,N ) 3M with M = fs/8.
`Similar to the first moving average procedure, this facili-
`tates choosing more accurate local minima.
`Fourth, local minima and their positions are obtained
`by using an adaptive threshold (Tha) and a moving
`window length, W. Local minima obtained on the
`boundary of a moving window are neglected. Local
`minima should be no larger
`than the set adaptive
`threshold value. The length of a window should not
`exceed half of
`the cardiac cycle to avoid including
`down-slope segments from different cardiac cycles. In
`this study, the moving window length, W, was selected
`as fs/4 = 100. The adaptive threshold was obtained by
`setting Tha(i) = ((i ) 1th)
`local minimum)/3, where
`Tha(1) is defined as the mean of the first 5 local minima,
`where i = 1, 2, 3, ..., N. The moving average parameter
`M was selected as fs/8. Justification for this selection will
`be shown in the Results section.
`To compare the similarity indices between PPGV and
`HRV, the following widely used time- and frequency-do-
`main parameters are calculated and compared: the standard
`deviation of normal-to-normal R–R intervals (SDNN), the
`root-mean square of the difference of successive R–R
`intervals (RMSSD), the ratio of the low-to-high frequency
`spectra (LF/HF, LF: 0.04–0.15 Hz, HF: 0.15–0.4 Hz) and
`the approximate entropy (ApEn). ApEn is denoted as
`ApEn(m, r, N) whereN is the length of the series to be
`analyzed and m determines the length of the sequences to be
`compared, which can be estimated by calculating the false
`nearest neighbor. The parameter r is the tolerance threshold
`for accepting similar patterns between two segments, and has
`been recommended to be within 0.1–0.2 times the standard
`deviation of the data [13]. ApEn(m, r, N) is the average of the
`logarithms of the conditional probabilities that sequences of
`m beats which are ‘‘close’’ (within ± r) will remain ‘‘close’’ at
`the (m + 1)th beat:
`P
`Nm
`ðrÞ
`ln Cmþ1
`i ðrÞ
`i
`Cm
`i¼1
`ApEn ¼
`ð7Þ
`N m
`r ðiÞ ¼ VmðiÞ=ðN m þ 1Þ where VmðiÞ ¼
`where Cm
`no: of d½XðiÞ; Xðjފ  r:
`The student-t test was used to determine whether there
`is a statistical significance between PPGV and HRV.
`Correlation coefficients wewe used to measure the linear
`relation between PPGV and HRV. p-values less than 0.05
`were considered to be statistically significant.
`
`ð6Þ
`
`Treat h1(t) as a new set of data, and repeat steps 1–2
`until h1(t)
`is converged. Then set C1(t) = h1(t). The
`process stops when the difference between two consec-
`utive shifts is smaller than a selected threshold SD, de-
`fined by
`"
`#
`

`
`

`h1ðk1ÞðtÞ h 1kðtÞ
`1ðk1ÞðtÞ
`h2
`
`2
`
`XT
`
`t¼0
`
`SD ¼
`
`ð3Þ
`
`(3) Calculate the residue R1(n) = x(n) ) C1(n); Treat
`R1(n) as a new set of data, and repeat steps 1–2 until
`the residue becomes a constant or a monotonic
`function.
`(4) xðnÞ ¼ C1ðnÞ þC 2ðnÞ þ þ CNðnÞ þR NðnÞ:
`Ci(n) is an intrinsic oscillatory mode and RN(n) is the
`residue. Thus, the signal can be reconstructed by
`summing up all intrinsic oscillatory modes.When all
`the IMFs are extracted, the cross-correlation between
`each IMF and x(t) is calculated. The most correlated
`N number of IMFs whose dominant frequencies are
`larger than 0.5 Hz will be used to reconstruct the
`signal. In this way, most low frequency trends and
`high frequency noise can be removed. In our study,
`we used N = 4.
`
`Once the low frequency trends and high frequency
`noise were removed via the EMD algorithm, the PPG
`variability series p was obtained by taking the first deriv-
`ative of the down-slope phase of the PPG waveforms
`followed by selecting the resulting largest negative values.
`To ensure the robustness of the algorithm, two moving
`average processes were applied in the PPG variability
`detection procedure.
`First, PPG waveforms are denoted as a(i), i = 1, 2, 3,
`±, ... ,N and the moving average of the signal a is per-
`formed:
`
`aðjÞ
`
`ð4Þ
`
`Xi
`
`þM
`
`bði M þ 1Þ ¼
`
`1
`2M þ 1
`j¼iM
`where i = M, M +1, M + 2, ... , N ) M with M = fs/
`8. The fs refers to the sampling frequency. This ensures
`that the local minima of the PPG signals are accentu-
`ated when the derivative is taken in the proceeding
`step.
`Second, define c as
`derivative) series of b:
`
`the successive difference (first
`
`cðiÞ ¼ Db ¼ bði þ 1Þ bðiÞ
`i ¼ 1; 2; 3; . . . ; N 2M
`
`ð5Þ
`
`4
`
`

`

`reconstruction was based on using the four most corre-
`lated IMFs whose dominant frequency components are
`higher than 0.5 Hz. We note that both high and low
`frequency trends and noise are removed in the recon-
`structed signal. The EMD procedure was successful in
`removing noise and low frequency trends in all of the
`signals analyzed, and this procedure facilitates down-slope
`peak detection of the PPG waveforms.
`Figure 2 shows a representative result obtained using
`the procedure outlined in the Methods section for the
`detection of the largest negative peak of the PPG wave-
`form. Panel (a) is the detrended PPG waveform, panel (b)
`represents the moving average of the signal in panel (a),
`panel (c) represents the first derivative of the signal in panel
`(b), and panel (d) represents the moving average of the
`signal in panel (c). Details regarding the moving average
`and the derivative parameters are provided in the Methods
`section. The circles in panel (d) represent the position of
`the largest negative values. Note that as compared to panel
`(a), the negative values are better defined in panel (d).
`Finally, the PPGV signal is obtained by measuring the
`distance between two consecutive PPG signals in panel (d).
`
`Journal of Clinical Monitoring and Computing
`
`Original Series
`
`Reconstructed series
`
`1.4
`
`1.2
`
`1
`
`0.8
`
`0.6
`
`0.4
`
`0.2
`
`0
`
`)stlov( mrofevaW GPP
`
`-0.2
`0
`
`2.5
`
`5
`
`7.5
`
`12.5 15
`10
`Time (S)
`
`17.5 20
`
`22.5
`
`25
`
`the original PPG and the reconstructed PPG
`Fig. 1. Comparison of
`waveforms based on the use of empirical mode decomposition.
`
`RESULTS
`
`A comparison between the original and the reconstructed
`PPG signal via the use of EMD is shown in Figure 1. The
`
`180
`
`b
`
`160
`
`140
`
`120
`
`100
`
`50
`
`d
`
`0
`
`5
`
`10
`Time (s)
`
`15
`
`20
`
`-50
`0
`
`5
`
`10
`Time (s)
`
`15
`
`20
`
`1.6
`
`a
`
`1.4
`
`1.2
`
`1
`
`0.8
`
`0.5
`
`c
`
`0
`
`-0.5
`
`-1
`0
`
`)stlov( edutilpmA
`
`Fig. 2. Smoothing via moving average filter and the first derivative used to accentuate better detection of the local minima of the processed PPG signal. Panels
`a, b, c and d represent the PPG signal, moving average filter applied to the signal shown in panel a, derivative of the signal shown in panel b, and a second
`moving average filter on the signal shown in panel c, respectively.
`
`5
`
`

`

`Lu et al.: Can Photoplethysmography Variability Serve as an Alternative Approach to Obtain Heart Rate Variability Information?
`
`another indication that there is close similarity between
`the two signals.
`As indicated in the Methods section, the selection of
`moving average parameter M is important to the accuracy
`of identifying the PPG variability time series. Figure 4
`shows simulation results of the incorrect identification
`ratio of the PPGV as a function of the size of the moving
`average parameter. As shown, the incorrect identification
`ratio reaches the minimum of 0.54% when M = fs/8 but
`there is not much difference between M = fs/6–fs/9.
`Tables 1 and 2 show comparison of the HRV and
`PPGV derived time- and frequency-domain parameters
`consisting of SDNN, RMSSD, LF/HF and ApEn during
`supine and upright positions, respectively. As shown,
`there is no significant difference between parameters of
`PPGV and HRV for both supine and upright positions
`(p > 0.05). Correlation coefficients between parameters of
`PPGV and HRV are shown in Tables 1 and 2, and they
`indicate
`strong
`correlation
`(p < 0.00001)
`between
`parameters of PPGV and HRV, in all cases. We note that
`the correlation coefficients of parameters in the supine
`position are generally higher than those in the upright
`position, indicating that PPGV signals measured from the
`upright position are more affected by motion artifacts. List
`of abbreviations are provided in Table 3.
`
`DISCUSSION AND CONCLUSIONS
`
`The promise of early detection and diagnosis of lethal
`cardiac arrhythmias based on noninvasive assessment of
`the autonomic nervous system via heart rate variability has
`led to significant interest among many researchers. In the
`biosignal processing field, the research focus is to develop
`more accurate algorithms
`that can be used for early
`diagnosis of malignant cardiac arrhythmias. In the field of
`cardiac monitoring systems, the current emphasis is on
`miniaturization of a system that
`is wireless and cost
`effective so that the measuring devices can be deployed to
`a large population. One such approach to accomplish the
`above-described three capabilities is to come up with a
`sensor
`that can extract many different physiological
`parameters, as the current environment is such that a
`sensor is limited to only one particular task. In this study,
`given the ubiquity and simplicity of the pulse oximeter,
`we investigate whether the PPGV can be used in lieu of
`heart
`rate variability to quantify and assess dynamic
`characteristics of the autonomic nervous system. If this is
`possible, then the pulse oximeter can have at least three
`different functionalities: blood oxygenation level, heart
`rate variability information, and respiratory rate. Having
`one sensor to perform all three tasks also facilitates com-
`pactness of the monitoring system facilitates making it
`
`Fig. 3. Comparison of the HRV and PPG signals in the top panel shows
`that the signals are nearly identical. The bottom panel shows a segment of
`the upper panel, which further illustrates the closeness of the two signals.
`
`The top panel of Figure 3 represents the overlap of the
`PPGV and HRV signals for a representative subject, and
`the bottom panel is a close-up one segment of PPGV
`(dotted line) and HRV (solid line) series. The solid and
`dotted lines match closely for all times, as Figure 3 shows.
`To quantitatively measure the similarity between the two
`signals, we computed the ratio between the power of the
`difference between PPGV and HRV to the power of
`HRV. The ratio is 1.35 ± 1.43% for the supine position,
`3.03 ± 2.78% for the upright position, and the mean ratio
`between upright and supine is 2.19 ± 2.32%. Thus, this is
`
`20
`
`15
`
`10
`
`5
`
`0
`
`-5
`
`)%( oitaR noitacifitnedisiM
`
`4
`
`5
`
`6
`7
`8
`9
`10
`11
`M (Moving average Parameter)
`
`12
`
`Fig. 4. Incorrect identification ratio of the PPG as a function of the size of
`the moving average parameter. Note that the incorrect identification ratio is
`relatively low when M is between fs/6 and fs/9.
`
`6
`
`

`

`Journal of Clinical Monitoring and Computing
`
`Table 1. Quantitative comparison of the PPGV and HRV during supine position using linear time and frequency domain parameters as well
`as the nonlinear quantity ApEn. There were no statistical differences in all parameters shown, suggesting that the PPGV and HRV are
`similar
`
`Supine
`
`SDNN (s)
`
`RMSSD (s)
`
`LF/HF (unitless)
`
`ApEn (2,0.15)
`(unitless)
`
`PPGV
`
`HRV
`
`PPGV
`
`HRV
`
`PPGV
`
`HRV
`
`PPGV
`
`HRV
`
`Sub 1
`Sub 2
`Sub 3
`Sub 4
`Sub 5
`Sub 6
`Sub 7
`Sub 8
`Sub 9
`Sub 10
`Student-t test, p
`r2
`
`0.0641
`0.0574
`0.0524
`0.0368
`0.0412
`0.0269
`0.0542
`0.0633
`0.141
`0.0207
`0.958
`0.9997*
`
`0.0632
`0.0572
`0.0526
`0.0368
`0.0404
`0.0251
`0.0519
`0.062
`0.1405
`0.0204
`
`0.0426
`0.0371
`0.0383
`0.0311
`0.0306
`0.0175
`0.0722
`0.0356
`0.149
`0.0118
`0.964
`0.999*
`
`0.0419
`0.0374
`0.0394
`0.0311
`0.0293
`0.0144
`0.0679
`0.0347
`0.1498
`0.0117
`
`1.9081
`3.3237
`0.6642
`0.6772
`1.4595
`0.208
`0.9485
`0.2684
`0.6791
`1.7337
`
`1.9061
`3.5228
`0.7316
`0.595
`1.4443
`0.2291
`0.9093
`0.3324
`0.6884
`1.6075
`0.983
`0.9964*
`
`1.2005
`1.0262
`1.3854
`1.3226
`0.9482
`1.2257
`1.009
`1.2585
`1.2554
`1.3844
`0.732
`0.9649*
`
`1.1941
`1.0661
`1.4121
`1.3153
`1.0095
`1.323
`1.0781
`1.233
`1.2676
`1.3443
`
`wireless as it reduces the bandwidth as well as battery
`consumption, and reduces
`the cost
`significantly as
`it
`eliminates the need for having three separate sensors and
`their associated supporting hardware requirements.
`The feasibility of the PPGV was verified by comparing
`both time and frequency-domain parameters of the HRV,
`which all demonstrated high correlation between the two
`signals. Likewise, ApEn values were not
`statistically
`
`different for both signals. Our results were aided by our
`preprocessing procedures of the PPG data which involved
`elimination of low and high frequency trends as well as
`simple derivative and moving average procedures
`to
`accentuate the detection of local minima.
`We observed that the PPG signal obtained during the
`upright position had more motion artifacts than during
`the supine position, as evidenced by a lower correlation
`
`Table 2. Quantitative comparison of the PPGV and HRV during upright position via linear time and frequency domain parameters as well
`as the nonlinear quantity ApEn. There were no statistical differences in all parameters shown, suggesting that the PPGV and HRV are
`similar
`
`Upright
`
`SDNN (s)
`
`RMSSD (s)
`
`LF/HF (unitless)
`
`ApEn (2,0.15) (un-
`itless)
`
`PPGV
`
`HRV
`
`PPGV
`
`HRV
`
`PPGV
`
`HRV
`
`PPGV
`
`HRV
`
`Sub 1
`Sub 2
`Sub 3
`Sub 4
`Sub 5
`Sub 6
`Sub 7
`Sub 8
`Sub 9
`Sub 10
`Student-t test, p
`r2
`
`0.0327
`0.0588
`0.0491
`0.0279
`0.0347
`0.0365
`0.0311
`0.0561
`0.0669
`0.0232
`0.827
`0.9977*
`
`0.0311
`0.0582
`0.048
`0.0271
`0.0327
`0.0365
`0.0279
`0.0541
`0.0642
`0.0225
`
`0.0132
`0.0229
`0.0178
`0.0119
`0.0193
`0.0286
`0.0254
`0.0289
`0.0394
`0.0109
`0.594
`0.9858*
`
`0.0104
`0.0222
`0.0157
`0.0097
`0.0173
`0.0289
`0.0195
`0.0269
`0.0374
`0.0079
`
`2.0875
`7.0911
`8.3334
`1.4044
`3.0242
`1.3458
`1.0841
`1.2307
`0.5892
`6.2982
`0.431
`0.9254*
`
`2.9932
`7.5436
`10.3396
`2.7836
`3.3079
`1.9998
`1.3121
`1.8484
`0.6656
`12.4375
`
`1.0722
`1.1471
`0.9648
`1.2213
`0.9086
`1.1851
`1.4042
`1.0484
`1.1899
`1.1184
`0.569
`0.967*
`
`1.1941
`1.0661
`1.4121
`1.3153
`1.0095
`1.1851
`1.3689
`0.9925
`1.1809
`0.9953
`
`7
`
`

`

`Lu et al.: Can Photoplethysmography Variability Serve as an Alternative Approach to Obtain Heart Rate Variability Information?
`
`Table 3. List of abbreviations
`
`REFERENCES
`
`Approximate Entropy
`ApEn
`Electrocardiogram
`ECG
`Empirical Mode Decomposition
`EMD
`High Frequency
`HF
`Heart Rate Variability
`HRV
`Intrinsic Mode Function
`IMF
`Low Frequency
`LF
`Photoplethysmography
`PPG
`Photoplethysmography Variability
`PPGV
`RMSSD Root-mean Square of the Successive Difference
`SDNN
`Standard Deviation of Normal-to-Normal
`
`In addition,
`between the HRV and PPGV signals.
`unless an adaptive filtering scheme such as Wiener fil-
`tering is used, the PPG signal is susceptible to significant
`motion artifacts which even the EMD has difficulty in
`eliminating. Thus, future studies need to address elimi-
`nation of significant motion artifacts in the PPG signal
`in situations where they are unavoidable. However,
`given the fact that pulse oximetry is ubiquitous, simple
`to use and has the potential to provide multi-physio-
`logical parameters, its advantages outweigh the current
`shortcomings. The current work strongly suggests a
`good alternative to understanding dynamics pertaining
`to the autonomic nervous system without the use of an
`ECG device. Most
`importantly,
`the multi-functional
`pulse oximeter-based
`device has
`the
`tremendous
`potential to be readily accepted by patients due to its
`simplicity and comfort, not to mention the significant
`potential to decrease health care costs as compared to
`the current practice of using multiple sensors and
`devices to obtain the data.
`
`1. Heart rate variability: standards of measurement, physiological
`interpretation and clinical use. Task force of the European
`society of cardiology and the north american society of pacing
`and electrophysiology. Circulation 1996; 93: 1043–1065.
`2. Akselrod S, Gordon D, Madwed JB, Snidman NC, Shannon
`DC, Cohen RJ. Hemodynamic regulation: investigation by
`spectral analysis. Am J Physiol 1985; 249: H867–H875.
`3. Bigger
`JT Jr., Fleiss
`JL, Rolnitzky LM, Steinman RC,
`Schneider WJ. Time course of recovery of heart period vari-
`ability after myocardial infarction. J Am Coll Cardiol 1991; 18:
`1643–1649.
`4. Broome IJ, Mason RA. Identification of autonomic dysfunc-
`tion with a pulse oximeter. Anaesthesia 1988; 43: 833–836.
`5. Davies CT, Neilson JM. Sinus arrhythmia in man at rest. J Appl
`Physiol 1967; 22: 947–955.
`6. Eckberg DL. Sympathovagal balance: a critical appraisal.
`Circulation 1997; 96: 3224–3232.
`7. Huang W, Shen Z, Huang NE, Fung YC. Engineering analysis
`of biological variables: an example of blood pressure over
`1 day. Proc Natl Acad Sci U S A 1998; 95: 4816–4821.
`8. Huang W, Shen Z, Huang NE, Fung YC. Use of intrinsic
`modes in biology: examples of indicial response of pulmonary
`blood pressure to +/) step hypoxia. Proc Natl Acad Sci USA
`1998; 95: 12766–12771.
`9. Huikuri HV, Koistinen MJ, Yli-Mayry S, Airaksinen KE,
`Seppanen T, Ikaheimo MJ, Myerburg RJ. Impaired low-fre-
`quency oscillations of heart rate in patients with prior acute
`myocardial infarction and life-threatening arrhythmias. Am J
`Cardiol 1995; 76: 56–60.
`10. Kleiger RE, Miller JP, Bigger JT Jr., Moss AJ. Decreased heart
`rate variability and its association with increased mortality after
`acute myocardial infarction. Am J Cardiol 1987; 59: 256–262.
`11. Modi KD, Sharma AK, Mishra SK, Mithal A. Pulse oximetry
`for the assessment of autonomic neuropathy in diabetic pa-
`tients. J Diabetes Complications 1997; 11: 35–39.
`12. Huang NE, SRL, Shen Z. The mechanism for frequency
`downshift in nonlinear wave evolution. Adv Appl Mech 1996;
`32: 59–111.
`13. Pincus SM. Approximate entropy as a measure of system
`complexity. Proc Natl Acad Sci USA 1991; 88: 2297–2301.
`
`
`
`View publication statsView publication stats
`
`8
`
`