Aortic and arterial stiffness determination by ... - Core

Procedia Chemistry Procedia Chemistry 1 (2009) 1243–1246 www.elsevier.com/locate/procedia

Proceedings of the Eurosensors XXIII conference

Aortic and arterial stiffness determination by photoplethysmographic technique M. Huotaria*, N. Yliaskab, V. Lanttoa (emer), K. Määttäb, J. Kostamovaarab University of Oulu, Department of Electrical and Information Engineering, a Microelectronics and Materials Physics Laboratories b Electronics Laboratory, Oulu, Pentti Kaiteran tie 1, FIN-90014 OULU UNIVERSITY; FINLAND

Abstract Photoplethysmography is an interesting biomedical technique and combined with pulse wave analysis it becomes an informationrich and non-invasive diagnosis method. A new constructed biomedical device can record pulse waveforms by a photonic sensor from finger and toe, and then we mathematically decompose these pulse waveforms into their four primaries. Based on these divided pulse waves is possible to estimate vessel elasticity which is called arterial stiffness. Both aortic and arterial stiffness (AS1 and AS2) in addition to the percussion to tidal index are calculated from pulse wave decomposition products which comprend four primary wave forms. These indices can be a sensitive biomarker as a cardiovascular risk factor.

Keywords; Digital volume pulse; photoplethysmography; pulse wave analysis; arterial stiffness index 1. Introduction

This paper concerns measurement technique of the digital volume pulse wave and pulse wave analysis procedure. The light absorption, reflection, refraction, and transmission of tissue blood in the visible and infrared range is partly caused by the oxidized and reduced hemoglobin of which blood concentrations can be measured /1/. The measurement principle is here applied for photoplethysmography (PPG). PPG enables a non-invasive measurement of the peripheral pulse wave and assessment of arterial stiffness. PPG is simple and only requires of SMA LEDs and photodiodes to be applied to the skin. In contrast to a conventional biomedical measurement which requires several sticky transducers, e.g. electrodes, to be carefully placed on the body, whereas PPG measurements needs no physical contacts and can help reduce external effects on subjects and attenuate the electrical noise. Pulse wave analysis is fulfilled with the optimization algorithm which solution was found by Levenberg and Marquardt and coded in the Origin 7.5. In the optimization procedure, simplex algorithm is also used for finding the best fit with minimum residual error. PPG waveforms begin to shed light on the information hidden in PPG signals on arterial stiffness /2/.

* Corresponding author. Tel.: +358-8-5537984; fax: +358-8-5532728. E-mail address: [email protected].

1876-6196/09 © 2009 Published by Elsevier B.V. Open access under CC BY-NC-ND license. doi:10.1016/j.proche.2009.07.310

1244

M. Huotari et al. / Procedia Chemistry 1 (2009) 1243–1246

1.1. Photoplethysmographic measurement principle When a part of the body is illuminated, it is possible to obtain a waveform called photoplethysmograph (PPG). The resulting signal is photoplethysmogram. PPG technique has not been fully validated and it may measure also many hemodynamic parameters, which are not yet known. To obtain a noiseless PPG, we can choose a LED as a light source and a photo diode as a detector. If the PPG is obtained from the reflected light, the source and the detector will be housed on the same geometry, e.g., in Beurer PM 100 heart rate meter which is based on a light reflection probe. Earlier works have concluded that the optimum distance between the source and detector in the sensor lies in the range of 2 to 20 mm depending on the light wavelength. This type of sensor can be used over the skin on any part of the body. On the other hand, if one has to obtain a PPG using the transmitted light, the source and sensor need to be arranged on the two different sides of the organ in the parallel plane, as on the most PPG devices, e.g., Hokanson MD6RP Photo Plethysmograph which is based on a light transmission probe. Our PPG measurements are done by the university lab-made PPG device based also on the transmission probe. In practice, all the subjects were recorded in the supine position for 300 s, with the transmission PPG probe placed on the left index finger and on the second toe by a four channel HP 35670A digital signal analyzer in its capture mode. The infrared LED light has a peak wavelength of 940 nm. To eliminate motion artifact, the subjects were encouraged to keep their fingers and toes relaxed. The data were sampled at 200 Hz. A total of 50 data files from both male and female subjects were selected for the analysis.

1.2. Photoplethysmographic waveform analysis method PPG waveforms were recorded from volunteer subjects in the morning before no coffee intake. It is hypothesized that caffeine induced sympathetic activation and therefore attenuates also significant low frequency spectral components in the variability spectra. Each lasted for about five minutes from which sequential 20 to 30 pulse waves were selected. The signal analysis procedure is implemented in Origin 7.5 (Microcalc). To decompose PPG waves into the individual physiological components, advanced waveform decomposition techniques can be adopted. The use of waveform decomposition techniques can also enhance the baseline restoration, in particular the artifact introduced by the fluctuation in the finger or toe with the PPG sensor. One potential technique for waveform decomposition is Levenberg-Marquardt optimization algorithm (LMO) for non-linear fitting. Preliminary trial of the LMO technique is performed on a PPG data recorded from the elderly and young subjects in supine position. In the young subjects, e.g., radial wave is easy for analysis purposes, whereas in the elderly, it is very difficult to be analyzed completely. When studying wave reflections, it is possible to decompose the measured PPG waveform into the percussion wave and the four reflected components, e.g., the tidal wave, the dichrotic wave, and the peripheral reflection wave for index calculation. For fully characterizing the time and real nature of wave reflection and also allowing accurate determination of the time of arrival of the reflected waves we realized a trial and error procedure at first. This analysis can be done in the time domain using principles general optimization procedures. This technique has its pros and cons, but the final fitting result based on the residual error is a very good method. When analysis the tidal wave we can see that in the younger persons the width of the tidal wave is within the limits under the value of 0.2 and that is the limit, whereas in the case of the elderly persons the width of the tidal wave is fixed to 0.2 but it has a snowball effect.

2. Results In Figures 1 and 2 there are pictured the PPG waveforms analyzed based on LMO algorithm procedure. The black line is the measured PPG wave, the green is the percussion, the blue is the tidal, the magneta is the dicrotic, and the navy is the peripheral reflections wave. In these cases we can determine the reflection index, A, and the stiffness indices (relative areas). In elder individuals, the tidal wave of the PPG is different and decomposition by the current software is difficult.


0.075

t=183.07 ms 0,2746

0 ±0 0.20194 0.85509 0.02228 0.29159 0.22675 0.00214 0.45434 0.14968 0.00467 0.59819 0.10652 0.00136

0,11481

PPG1(V)

PPG1(V)

0.025

y0 xc1 w1 A1 xc2 w2 A2 xc3 w3 A3 xc4 w4 A4

±0.00306 ±0.00641 ±0.00021 ±0.00567 ±0.01536 ±0.00012 ±0.00279 ±0.00731 ±0.00011 ±0.00648 ±0.00807 ±0.00017

Chi^2/DoF = 0.00013 R^2 = 0.99538

0.4

0,44008 0.3 0.2

t=114.8 ms; t=0 ms; A=2 AS =0.535; AS =20.654 1 2

0.1

0.0025

AS1=0.210; AS2=2.182

0.1

0.2

0,74623

0,22961

t=183.07 ms; t=91.54 ms; A=3

0.000 0.0

Data: Data1_D JaRu24PPG1ac Model: LogNormal Equation: y = y0 + A/(sqrt(2*PI)*w*x)*exp(-(ln(x/xc))^2/(2*w^2)) Weighting: y No weighting

0.5

Chi^2/DoF = 1.3243E-6 R^2 = 0.99702

0.050

DJaRu24c

0.6

Data: Data1_F Model: LogNormal Equation: y = y0 + A/(sqrt(2*PI)*w*x)*exp(-(ln(x/xc))^2/(2*w^2)) Weighting: y No weighting

0,09153

1245


0 ±-0.26526 ±0.00995 0.9152 ±0.01439 0.21193 ±0.00518 0.2419 ±0.00426 0.14257 ±0.01853 0.00549 ±0.00083 0.48122 ±0.00914 0.30625 ±0.01637 0.11339 ±0.00319 0.78027 ±0.00586 0.18831 ±0.00734 0.06522 ±0.00515

0.0

0.3

0.4

0.5

0.7 t(s) 0.8

0.6

0.0 0.02

Residual error

0.2

0.4

0.6

0.8

1.0

t(s)

1.2

0.4

0.6

0.8

1.0

t(s)

1.2

Residual error

0.01 0.00

0.0000

-0.01

-0.0025 0.0

-0.02

0.1

0.2

0.3

0.4

0.5

0.7 t(s) 0.8

0.6

(a)

0.0

0.2

(b)

Fig. 1. (a) PPG wave with its residual error for a 13 y male person (AS1=0.210; AS2=2.182; A=3); (b) PPG wave with its residual error for a 24 y male person (AS1=0.535; AS2=20.654; A=2). (AS1=dichrotic area/percussion area, AS2=dichrotic area/tidal area).

0.6 0.5 0,13308

0.1 0.0 0.0 0.025

2.0

t=133.07 ms; t=-10 s; A=2 AS1=0,334; AS2=9,896

0 ±0 0.23124 ±0.00578 0.72056 ±0.01154 0.16497 ±0.00271 0.27502 ±0.00436 0.12427 ±0.0163 0.00557 ±0.00067 0.48177 ±0.0094 0.20078 ±0.01566 0.05512 ±0.00227 0.62574 ±0.00721 0.09639 ±0.02348 0.0071 ±0.00299

PPG1(V)

PPG1(V)

0.3


Chi^2/DoF = 0.00079 R^2 = 0.99889

1.5 0,41525

1.0 t=116.79 ms; t=0 ms; A=2 0,23358

0.5 AS1=0,322; AS2=3,803

0,26615

0.2

Data: Data1_D Model: LogNormal Equation: y = y0 + A/(sqrt(2*PI)*w*x)*exp(-(ln(x/xc))^2/(2*w^2)) Weighting: y No weighting

0,11679

Chi^2/DoF = 0.00018 R^2 = 0.99219

0.4

0.2

2.5

Data: Data1_D IA27PPG1a Model: LogNormal Equation: y = y0 + A/(sqrt(2*PI)*w*x)*exp(-(ln(x/xc))^2/(2*w^2)) Weighting: y No weighting

0.4

0.6

0.0 0.0

t(s) 0.8

0.04

Residual error

0.02

0.000

0.2


0 ±0 0.19196 ±0.00403 0.69006 ±0.00904 0.67846 ±0.01064 0.24275 ±0.00247 0.17432 ±0.01093 0.05738 ±0.00422 0.43829 ±0.00243 0.2108 ±0.00837 0.21824 ±0.00724 0.62372 ±0.00558 0.09651 ±0.01176 0.0264 ±0.0038

0,62288

0.4

0.6

t(s)

0.8

0.4

0.6

t(s)

0.8

Residual error

0.00 -0.02

-0.025 0.0

(a)

0.2

0.4

0.6

-0.04 0.0

t(s) 0.8

0.2

(b)

Fig. 2. (a) PPG wave with its residual error for a 27 y female person (AS1=0.334; AS2=9.896; A=2); (b) PPG wave with its residual error for a 39 y male person (AS1=0.322; AS2=3.803; A=2). Numerical values calculated as above.

3. Discussion PPG waveform analysis provides much more information relating to arterial structure and function compared with ultrasound analysis on which the intima media thickness measurement is nowadays based. Because the PPG wave is determined by an individual complex hemodynamic properties, so it is clear that the interpretation of arterial stiffness indices in terms of the biomechanical properties of arteries is complex and not yet done. The PPG based time index A and the stiffness indexes, AS1 and AS2, represent an attempt to relate pulse wave indexes to arterial function and structure, respectively. One challenge for the future is to determine under what circumstances these indexes provide reliable measure of arterial functioning and structure, and whether their reliability can be improved in older individuals or in individuals with a waveform that exhibits new characteristics. Another challenge is to define the relationship between indices derived from the PPG and indices derived from the pressure and flow pulses. Because current PPG wave analysis and PPG indices are determined from small group of people, we need more


1246

analysis materials. In Figure 3 it is pictured the A index as a function of age. This index strongly depends on age and it has a small variability in the elderly subjects.

3.25

The index A as a function of age

A 3.00 2.75 2.50 2.25 2.00 1.75

B C D Linear Fit of Data1_B Linear Fit of Data1_C Linear Fit of Data1_D

1.50 1.25

Linear Regression for Data1_B: Y=A+B*X Param Value Error A 2,82436 0,23716 B -0,02139 0,005 -----------------------------------------------------------R SD N P ------------------------------------------------------------0,868 0,278 8 0,00524 -----------------------------------------------------------Linear Regression for Data1_C: Y=A+B*X -----------------------------------------------------------A 2,63682 0,07983 B -0,01905 0,0017 -----------------------------------------------------------R SD N P ------------------------------------------------------------0,984 0,0892 6 3,59483E-4 -----------------------------------------------------------Linear Regression for Data1_D: Y=A+B*X -----------------------------------------------------------A 2,51752 0,19065 B -0,01606 0,00425 -----------------------------------------------------------R SD N P ------------------------------------------------------------0,936 0,139 4 0,06342 ---------------------------------------------------------

1.00 0

10

20

30

40

50

60

70

80

age (years)

Fig. 3. The tidal peak time divided by the percussion peak time (the index A) as a function of human age for 8 cases.

Perhaps the most exciting application of PPG waveform analysis is the possibility of providing a rapid biophysical measure of diseases or ageing process. Because of PPG measurement simplicity, it can be employed in longitudinal epidemiological studies and be used to assess effects of cures on arterial stiffness. Determination of the PPG is a particularly simple method for performing automatically pulse wave analysis. Just as the pressure pulse, the PPG is influenced by large artery stiffness. Pulse waveform analysis of the PPG provides a rapid means of assessing arterial stiffness in health care centers. Measurements include the assessment of endothelial function, arterial stiffness on different organs, e.g. upper and lower parts, and characterization of arterial ageing. Future work will involve complete clinical studies, optimization of the measurement device, and evaluation of the LMO for online analysis.

Acknowledgements The minister of Finnish Education is gratefully acknowledged.

References 1. Van Kraitl, J. et al., An optical device to measure blood components by a photoplethysmographic method. J. Opt. A: Pure Appl. Opt. (2005), 318-324., doi: 10.1088/1464-4258/7/6/010 2. Strunk Kim, Y.-S., et al., Effects of aging on the cerebrovascular orthostatic response. Neurobiol. Aging (2009), doi:10.1016/j.neurobiolaging.2009.02.019