Eindhoven University of Technology

MASTER

Automated return-to-flow blood pressure measurement using finapres as a detector

Ruijgrok, G.B.P.

Award date:1995

Link to publication

DisclaimerThis document contains a student thesis (bachelor's or master's), as authored by a student at Eindhoven University of Technology. Studenttheses are made available in the TU/e repository upon obtaining the required degree. The grade received is not published on the documentas presented in the repository. The required complexity or quality of research of student theses may vary by program, and the requiredminimum study period may vary in duration.

General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright ownersand it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain

EINDHOVEN UNIVERSITY OF TECHNOLOGYFACULTY OF ELECTRICAL ENGINEERINGDEPARTMENT OF MEDICAL ELECTRICAL ENGINEERING

AUTOMATED RETURN-TO-FLOWBLOOD PRESSURE MEASUREMENTUSING FINAPRES AS A DETECTOR

by G.B.P. Ruijgrok

This report is submitted in partial fulfilment of the requirements for the Master's degreeofElectrical Engineering (Ir.) at the University ofTechnology in Eindhoven (departmentofMedical Electrical Engineering).The research was done at TNO BioMedical Instrumentation (BM!) located at theAcademic Medical Centre in Amsterdam from 16 January 1995 until 30 November 1995,under supervision ofProf Ir. K.H. Wesseling and Dr. J. van Goudoever.BMI is a department ofthe TNO Institute ofApplied Physics (TPD) in Delft.

The department of Electrical Engineering of the Eindhoven University of Technologycannot be held responsible for the contents of training and graduation reports.

DANKWOORD

Van deze pagina wi! ik gebruik maken om degenen te bedanken die een bijdrage hebben geleverdaan de totstandkoming van dit verslag of op andere wijze tot steun zijn geweest tijdens mijnafstudeerperiode.

Karel, Jeroen en Jos bedankt voor juIlie begeleiding tijdens mijn afstudeerperiode.

Karel: voor de kritische beschouwingen op mijn verslag en voordrachten, het corrigeren vanmijn Engelse woorden/zinnen, de triggers om me van een zijspoor terug te loodsen naarhoofdspoor en je 'humorrijke' levenswijsheid.

Jeroen: voor de directe begeleiding, het oplossen van praktische problemen en de gezelligelunches.

Jos: vooral voor de aanwijzingen m.b.t. mijn voordrachten en je eigenschap van geduldigluisteraar.

Overig: voor het overbrengen van kennis, de bereikbaarheid als aanspreekpunt en het fungerenals proefpersoon.

A&A: voor de bijdrage wat betreft 'het' manchet (-gebeuren).

Diph, ondanks vaak moeilijke omstandigheden kon ik toch altijd mijn verhaal bij je kwijt.

Figure on title-page: Position of the finger in palpating the radial artery, as recommended byNorris (1916). (From Geddes L.A., The DIRECTand INDIRECTMEASUREMENT ofBLOOD PRESSSURE.)

CONTENTS

SUMMARy 1

INTRODUCTION 3

1 HOW DOES SYSTOLIC BLOOD PRESSURE OBTAINED WITH VISIUALLYDETECTED RETURN-TO-FLOW (RTF) IN FINGER PRESSURE COMPAREWITH SYSTOLIC BLOOD PRESSURE OBTAINED BY INDIRECT ANDDIRECT BLOOD PRESSURE MEASUREMENT? 5

1.1 Introduction..................................................... 51.2 Methods........................................................ 5

1.2.1 Patients 51.2.2 Non-invasive finger arterial blood pressure (FINAP) measurement 61.2.3 Riva-Rocci/Korotkoff (RRK) blood pressure measurement 61.2.4 Intra-arterial blood pressure (lAP) measurement 61.2.5 Data-analysis 61.2.6 Statistics 8

1.3 Results 8

1.3.1 RTF versus systolic RRK pressure 81.3.2 RTF versus systolic lAP 10

1.4 Discussion 12

1.4.1 Literature 121.4.2 Retum-to-flow 141.4.3 Patients 141.4.4 RTF versus RRK 141.4.5 RTF versus lAP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 16

1.5 Conclusion..................................................... 20

2 AUTOMATIC RETURN-TO-FLOW DETECTION 21

2.1 Description of the method and software 212.2 Results........................................................ 24

2.2.1 Comparison of the automatic RTF-value with the manual RTF-value .... 242.2.2 Comparison of the automatic RTF-value with intra-arterial systolic

pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

2.3 Discussion...................................................... 29

2.3.1 Detection algorithm 292.3.2 Automatic RTF versus manual RTF - " 30

2.4 Conclusion 30

3 INFLATION / DEFLATION TECHNIQUE OF THE ARM CUFF 31

3. 1 Introduction............................ . . . . . . . . . . . . . . . . . . . . . . .. 313.2 Requirements 31

3.2.1 Inflation height 313.2.2 Inflation time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313.2.3 Patients inconvenience 323.2.4 Deflation rate 33

3.3 Implementation 333.4 Discussion 36

LITERATURE REFERENCES 37

SUMMARY

Several methods exist to estimate blood pressure non-invasively at the upper arm. They all usean arm encircling, inflatable cuff, but detection criteria differ. Well known are the oscillometrictechnique ofMarey for mean pressure and the auscultatory technique ofRiva-Rocci/Korotkofffor systolic and diastolic blood pressure. Another is the return-to-flow (RTF) method for systolicblood pressure.In this report the return-to-flow technique is investigated using a new way of detection for thereturn ofblood flow: the Finapres. To show the usefulness of the RTF method the RTF pressuremanually obtained from the Finapres signal has been compared with two other methods formeasuring blood pressure: the Riva-RocciIKorotkoff (RRK) method and the direct method.An implementation of the RTF technique is described using computer pattern recognition todetect the moment of retum-to-flow. Results with this program are compared with the manuallydetected retum-to-flow pressure and intra-arterial systolic blood pressure on a set of pre-recordedpatient registrations.The manual RTF pressure compared to RRK has an error offset of 0.5 mmHg and a scatter of7mmHg, whereas the automatic RTF compared to manual has an offset of 0.3 mmHg and a scatterof2 mmHg. Compared to intra-arterial systolic blood pressure the manual and the automatic RTFtechnique underestimate pressures systematically by 6% to a mean offset of -12 mmHg and ascatter of5 mmHg. We conclude that the three cuff methods (RRK, manual RTF and automaticRTF) give comparable results in terms of error offset and scatter and all underestimate intra­arterial systolic pressure 6%. Our automatic detection of return-to-flow works well except incases when the upper arm cuff is either over- or underinflated.

1

INTRODUCTION

In 1976 TNOIBMI started the development ofa non-invasive blood pressure measuring devicecalled Finapres (for FINger Arterial PRESsure). This device records continuous finger arterialpressure in an indirect way.

During evaluation studies of this device versus intra-arterial and non-invasive upper arm bloodpressures it was observed that the moment ofKorotkoffphase I detection coincided remarkablyclosely to return ofblood flow detected with Finapres. This report describes a systematic studyof return-to-flow detection with Finapres measuring blood pressure distal ofan upper arm cuff

Chapter one describes a study of the return-to-flow method to obtain systolic blood pressure incomparison to conventional methods ofmeasuring blood pressure.Software has been developed to detect the return-to-flow pulse automatically. This software willbe discussed in chapter two.Chapter three deals with the technique of inflating and deflating the arm cuff

3

1 HOW DOES SYSTOLIC BLOOD PRESSURE OBTAINED WITH VISlUALLYDETECTED RETURN-TO-FLOW (RTF) IN FINGER PRESSURE COMPAREWITH SYSTOLIC BLOOD PRESSURE OBTAINED BY INDIRECT ANDDIRECT BLOOD PRESSURE MEASUREMENT?

1.1 Introduction

Several studies have been done to compare systolic blood pressure obtained with the return-to­flow (RTF) method with indirect and direct blood pressure measurement. The RTF method wasone ofthe earliest techniques to measure blood pressure indirectly. In 1897 Riva-Rocci introducedthe use ofthe air inflated occluding cuff [21]. The cuff pressure at disappearance or reappearanceof the radial pulse or the average of those two values was taken as the systolic blood pressure.The detection of the RTF pulse was done by palpation ofa radial artery (figure on title page) orlater by a second cuff distal to the occluding cuff. This second cuff was inflated to a lowerpressure and connected with a sensitive pressure gauge to detect the return-to-flow pulse [12].Later on also other techniques for the detection ofRTF were used. RTF could be recorded usingpiezoelectric, impedance or photoelectric pulse detectors [12]. In 1965 the Doppler ultra-soundtechnique was used for the detection ofRTF [27], which coined the name RTF and soon after theintroduction ofthe pulse oximeter into clinical anesthesia (1983) several investigators studied thepossibility to detect RTF with this device to obtain systolic blood pressure [5, 9, 10, 17, 8,24,25,26,33].

Since we desire to detect RTF by Finapres it is important to know how systolic pressure thusobtained correlates with conventional methods for the determination of the systolic bloodpressure.To gain experience with the methods involved, a subset of the patient data ofW.J.W. Bos et al.[8] was used. This allowed us to look at special cases in more detail and to show the usefulnessof the RTF-value. We compared to the Riva-RocciIKorotkoff method and to the invasivedetermination of blood pressure in the brachial artery of the contralateral arm. These are thepresent standards for indirect and direct blood pressure measurement [2, 31].

1.2 Methods

1.2.1 Patients

The research group consisted of hypertensive patients with therapy resistant hypertension,published before [7]. Patient characteristics are summarized in table 1.1.

5

lAP, average during 30 s controlperiod before cuff inflation

n age gendersystolic diastolic

mean range male female mean range mean range

14 66 52-79 9 5 194 128-240 90 65-122

Table 1.1: Patient characteristics.

1.2.2 Non-invasive finger arterial blood pressure (FINAP) measurement

Finger pressure was measured by a TNO Finapres model 5 device [14, 28]. The finger cuffwasapplied to the index or middle finger ofthe dominant hand. Subjects maintained the cuffed fingerat the level of the intra-arterial pressure transducer.

1.2.3 Riva-Rocci/KorotkofT (RRK) blood pressure measurement

In each patient two successive RRK measurements were performed on the dominant arm usinga 38 x 14 cm standard AMC ann cuffand a calibrated mercury sphygmomanometer. A constantcuff deflation rate of 2.5 mmHg/sec was achieved by using an electronic control valve. Cuffpressure was continuously recorded. The two RRK measurements were separated at least oneminute. At the detection ofKorotkoff phase I (KI) and Korotkoffphase V (K5) an electronicmarker was recorded manually to indicate systolic and diastolic readings from thesphygmomanometer (figure 1.1).

1.2.4 Intra-arterial blood pressure (lAP) measurement

Intra-arterial blood pressure was measured using a cannula inserted in the brachial artery of thenon-dominant arm. The natural frequency and damping coefficient of the fluid filled system wereapproximately 25 Hz and 0.30, at least adequate according to Gardner [11].

1.2.5 Data-analysis

Upper arm cuff pressure, systolic and diastolic markers, lAP and finger pressure were recordedon Sl'n'p c'n' '. ,- _. -...- - -- .•! -.•~- - TL. --- -----.-- _:~_l_ •••_-... AIn ..........."'""',.1 .. t .. .,"tnnl;no r!'ltparr anu UII IIIi:lgllt:ll~ l.al-'t:. 111l;~ }Jll;:>:>Ul ~ :>15-"WO> yy,-. '" r>J~ ,",VII .." ••w ......... ~""'I:""'O • _

of iDO Hz with a resoiution uf 0.25 mmHg. These digitized data files were aml!yzed.The manual RTF-value is the arm cuffpressure at the moment that the first pressure pulse (retum­to-flow pulse) was visible in the finger blood pressure recording during cuffdeflation (momentofRTF = upstroke ofRTF pulse, see figure 1.1).The RTF-values were compared with the systolic blood pressures obtained from the RRKreadings and from the intra-brachial measurements. To compare the RTF-value with intra-arterialsystolic blood pressure, the one beat systolic pressure preceding the moment ofRTF was used.

6

200.00

mmHg

1150.00

ann cuffpressure ~

100.00

100.0080.00

marker for K5

60.00, ,

brachial lAP0.00+---------,,.---~---'-'--___,_-----~----____,

20.00

50.00

-----7 s

200.00

mmHg

1150.00

100.00

systolic lAP

54.0053.0052.0051.00

0.00 +- ,.- -----L.---_L---'-_------. ---,

50.00-----7 s

Figure 1.1: Measurement cycle with cuff, finger and brachial pressure. Cuff pressures atKorotkoffphase I (1<1) and V (1<5) and at return-to-flow are indicated.

7

1.2.6 Statistics

Systolic blood pressure values were compared by correlation analysis and degree of agreementas suggested by Bland and Altman [1, 4].The correlation of systolic blood pressure values was compared using least squares linearregression and Pearson's product moment correlation (symbol r). To verify if the RTF-value canbe used as an alternative detection of brachial, systolic blood pressure differences between themethods have been plotted against the mean of two methods (RTF-RRK vs. (RTF+RRK)/2 andRTF-lAP vs. (RTF+IAP)/2). From the collected systolic blood pressure values the 95%confidence interval in differences between two methods was calculated by the mean ± 2*standarddeviation. If the differences were normally distributed, 95% ofthe differences would fall withinthe range ofmean ± 2 *standard deviation. The agreement between the two methods is consideredgood if these limits of agreement are within ± 10 mmHg. The AAMI (Association for theAdvancement ofMedical Instrumentation) standard for acceptability of the mean difference andstandard deviation are 5 and 8 mmHg, respectively [31].

1.3 Results

1.3.1 RTF versus systolic RRK pressure

The scatter diagram of the manual RTF-value versus systolic RRK blood pressure (figure 1.2)shows that correlation is good. Figure 1.3 is a Bland-Altman diagram showing the differencebetween versus average ofmanual RTF-value and systolic RRK pressure. The differences appearnormally distributed. The 95% confidence interval (mean ± 2*standard deviation) is also shown.

8

300 -r-----------------------."

250

.....011

:%: 200S,g

r...~ 150

..

'.

.'

50n = 28r = 0.985 (p < 0.001)y = 0.986*x+3.10

o 50 100 150 200

RRK (mmHg)250 300

Figure 1.2: Scatter dagram of manual RTF vs. systolic RRK pressure.

3U.....-----------------------,

.....=20-

!~~ 10-

~

mean + 2 -00 = 13.8

• 1

.2

mean = 0.536

...r... 0 +-=:"'::""::"'::""::"::""::"::"":=-=-=-=-=-=-=-=-=-=-=-=-..=....:0."::"::'''::''::'..::....:0.-"=""::'--=--=--="=--=--=-j

~ ..~

~ .

=28=0.536=6.61=-9=21

nmeansdmmmax

......ell -10-~ ~~n:}!~::~J _s::ell

S-20-

- 30 -t-,.,-.---r---r-r-T--.-,.,-....-r---'-""'--'--T'"'T--,--,rr-.---r---'-T'"'T--.-r-r-io 50 100 150 200 250 300

(manual RTF + RRK)/2 (mmHg)

Figure 1.3: Bland-Altman diagram; difference between vs. average of manual RTF andsystolic RRK pressure. Mean difference and 95% confidence interval(mean ± 2 *sd) are indicated by dashed lines.

9

1.3.2 RTF versus systolic lAP

The scatter diagram in figure 1.4 shows the correlation between the manual RTF-value andsystolic lAP in the contralateral ann.The increase in the difference seems to be systematic withincreasing pressure. It is approximately 6% of systolic lAP.Figure 1.5 shows the difference against mean ofmanual RTF and systolic lAP. The differencesare normally distributed. The bias is -12.0 mmHg. The 95% confidence interval is also shown. Itexceeds ± 10 mmHg.

10

JOO --.-------------------------"

250

.......110

:J::ag 200

//

//

/

//

.. ,

50/

/

//

//

n =28r = 0.992 (p < 0.001)y =0.944*x - 0.926

o 50 100 150 200

lAP (mmHg)250 300

Figure 1.4: Scatter diagram of manual RTF vs. systolic lAP.

30--.--------------------~

Oil 20­

:J::a.!0.. 10­

<-..

nmeansdnunmax

=28= -12= 5.3= -26= -2

~ 0 +--_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_---1_E-< mean + 2"00 = -1.40 1: 2~

......~-10-

='C~

S-20-

, ,, .-----------------------------

mean = -12.0 ' ...'3-----------------------------

mean - 2 *00 = -22.7·4

- 30 +-,......,...,.---,r-........l-r--r-r-rl----.-.,........,--.-........1--'-"""""''''''1----'-'''''''-'--'-'''''''1---,-.,......,..--1o 50 100 150 200 250 300

(manual RTF + IAP}/2 (mmHg)

Figure 1.5: Bland-Altman diagram; difference between va. average of manual RTF and systoliclAP.

11

1.4 Discussion

1.4.1 Literature

In 1954 Van Bergen et al. [3] investigated the relationship between systolic pressure recordeddirectly and that measured by the palpatory method. They found that palpatory systolic pressurewas underestimating direct systolic pressure considerably (about 30 mmHg below intra-arterialsystolic pressure at 120 mmHg) and variation was high. To reduce the large spread ofvalues moresensitive detectors for the distal RTF pulse were used. Figure 1.6 illustrates a graphic recordingof the cutfpressure and the peripheral pulse detected with several sensitive pulse detectors [12].

RADIAL PU.SE (PIEZa CRYSTAL)

FINGER PlA.SE (IMPEDANCE)

FINGER PULSE (PHOTOELECTRIC)

ISO

100

SO

o

SYSTOLIC PRESSlM137 """iii

TIME MARKS m 1 SECOND

I II " I " " , I " " " " " " Ii" " " "i " " iii " " " " " " , i " I

Figure 1.6: Graphic record of the return-to-flow method using piezoelectric, impedance andphotoelectric pulse detectors.

After the introduction ofthe Doppler ultra-sound technique to measure arterial blood flow (1960)\Vare used trjs techrjque for the detection of RTF to obtain systolic blood pressure [27].'T'~~~~l.~_ .,,:~l. v"'........"..."'.. "'f .. l 1,,,, C'tl1~;"'~ th", ",,,,,,,ihiliti..,, "fthi" n1pthnti r1 ~ 1fi1 Thf>v found.J.VC'-'l.ll\".l "Vl"ll.!.~"'JJU11"'1"". "" \.tol. J"'" ~"W\,6&""U. 1'''''..,.., ..,. va. •• &.A., _ _- L--' - - ----OJ

excellent correlation with the direct method but a significant difference.

The first trials to measure systolic blood pressure with a pulse oximeter have been reported in1987 by Wallace et al. and Korbon et al. [17,26]. In these two studies RTF detection using apulse oximeter was compared with systolic blood pressure measurement using the Dopplerultrasound technique. Their results showed that both methods correlated well.

12

In 1990 Talke et al. [25] were the first who compared the RTF method using a pulse oximeterwith intra-arterial blood pressure measurement and with the RRK method. In 20 healthyvolunteers (aged 24 to 48 yrs.) and 26 anesthetized patients (260 measurements) comparisonswere made between systolic RRK pressures and systolic pressures obtained with a pulse oximeter.They found for the equation of linear regression: RRK = 0.875 *RTF + 15.1 for the healthyvolunteers and RRK =0.961 *RRK + 5.37 for the anesthetized patients. Correlation was: r = 0.88and r =0.958, respectively. In 12 patients ofthe anesthetized group they measured systolic bloodpressure by pulse oximetry and by an intra-arterial cannula. The equation oflinear regression was:lAP = 0.825 *RTF + 27.6 and correlation was: r = 0.880.Their conclusion was that the particular pulse oximeter was useful in measuring systolic bloodpressure intraoperatively and that the values obtained with this method correlated well with thoseofother conventional methods. The intra-arterial readings were not at the same moment and site(lAP in radial artery) as the RTF readings.In 1991 they compared this RTF method with intra-arterial systolic blood pressure measurementduring helicopter flight [24]. The radial artery was cannulated in 10 patients (73 measurements;age 23 to 63 yrs.). The equation of linear regression was: RTF = 0.53*IAP + 51.18 andcorrelation was: r = 0.78. They concluded that the pulse oximeter could be used to measuresystolic blood pressure during helicopter flight and that indirectly measured systolic bloodpressure may be significantly lower than the intra-arterial systolic blood pressure.

Block et al. [5] (July 1991) were the first who evaluated automatic non-invasive systolic bloodpressure determination with the Ohmeda 2120 "return-to-flow" method. This device also uses apulse oximeter to detect RTF. Besides correlation analysis they also applied the statisticaltechnique ofBland & Altman [4] as well as the method of Lee et al. [19]. From 16 patients 1484paired determinations ofsystolic blood pressure with the automatic RTF method were comparedwith systolic lAP. The equation of linear regression was: RTF = 0.85*IAP + 9.49 and meandifference was: -9.1 ± 9.9 mmHg.They stated that the correlation between this automatic RTF method and intra-arterialmeasurement was good but not perfect. The Ohmeda 2120 monitor tended to underestimatesystolic pressures as determined by the arterial cannula, particularly at higher blood pressures.

The most recent study about the RTF method using a pulse oximeter was in 1994 ofLangbaumet aI. [18]. They compared intra-arterial systolic blood pressure with RTF using a pulse oximeterand with the oscillometric method in 50 sick neonates aged 24 to 40 weeks. In their results thedisappearance of pulse and the average of disappearance and reappearance of the pulse wasanalyzed. They found good correlation between intra-arterial and RTF readings. Correlationcoefficients were: r = 0.973 for pulse disappearance versus systolic lAP and r = 0.98 for theaverage of pulse disappearance and reappearance versus systolic lAP. The mean difference ofRTF minus lAP was -2.0 ± 1.6 mmHg for pulse disappearance and -1.9 ± 1.6 mmHg for theaverage ofwaveform loss and reappearance.The RTF method using a pulse oximeter gave better correlation and agreement with intra-arterialpressure than the oscillometric method. They concluded that systolic blood pressuremeasurements with pulse oximetry were accurate, easily obtainable and reliable in the neonatalpopulation.

In 1995 W.J. Bos et al. [8] published about the reconstruction of brachial artery pressure fromnon-invasive finger pressure measurements. This could be established with a frequency dependent

13

transfer function [13] which corrects for peripheral pulse wave distortion, but not for the pressuregradient caused by flow in the resistive vascular tree. W.J. Bos et al. described methods to correctfor this pressure gradient. Correction formulas were derived from the difference between fingerpressure and RRK diastolic, RTF systolic and oscillometric mean pressure measurements. Bestcorrection for the pressure gradient was obtained with the transfer function in combination withan equation based on RTF measurements from finger pressure readings ofFinapres distal of thearm cuff.

1.4.2 Return-to-flow

When Finapres is used to detennine RTF, the first small pressure pulse appearing during deflationof the arm cuff is detected. To cause this pulse, a fraction of the normal blood flow to theextremity is necessary. Although this fraction is very small there stilI need to be some pressuredifference between arterial pressure proximal to the arm cuff and the pressure in the arm cuff toestablish this flow. Thus RTF pressure would always measure too low.Also, as the RTF pulse travels underneath the arm cuff to the finger, it will be affected by thepressure gradient along the distal vascular tree and by pulse wave distortions. This should notinfluence the RTF measurement.There is also a time delay between the pulse passing underneath the arm cuff and arriving at thefinger. This delay can also be neglected.

1.4.3 Patients

In general, indirect systolic blood pressure measurement can underestimate direct readingssignificantly [2]. Our pressure measurements were obtained in an elderly population (mean age66 yrs.) ofpatients with hypertension and vascular disease. In such a group the pressure differencebetween left and right arm may be high. In this particular group of 14 patients there was no greatdifference between left and right arm RRK measurements, and this cannot be the cause of suchdifferences in this patient group.

1.4.4 RTF versus RRK

As can be seen in figure 1.2 there is good correlation (r =0.99) between the RTF-value andsystolic RRK blood pressure. Since both methods used the same arm and cuffand both detect thereturn ofblood flow with a deflating arm cuff, this seems explained. The difference between thetwo methods is that the RTF method detects the return to blood flow by an arterial pressurechange in the finger distal to the arm cuffand RRK detects the return to blood flow by the soundsthat are produced. Thus, different phenomena are detected caused by the same return to bloodflow.Though correlation is high, there are still some large differences, as can be seen in figure 1.3. The95% confidence interval is wide (-12.7 to 13.8 mmHg) so the degree of agreement is not verygood. RRK measurement can give rise to problems when applied to elderly people whose bloodvessels are stiffer and Korotkoff sounds less clear [30]. Moreover in the elderly blood pressureis less stable and arrhythmias occur more often. Therefore, it is harder to get good accuracy of

14

the RRK measurement.

In figure 1.3 there are three cases that exceed a difference of 10 mmHg. The differences are 11,15 and 21 mmHg, labeled 1,2 and 3, respectively. These cases are shown in figure 1.7a to 1.7c.

300.00

mmHg

I

RTF

200.00

100.00

a

65.0060.0055.00

0.00+--------.----------'--.--------~--­

50.00--7 S

300.00

mmHg

RTF

200.00

100.00

b

15

60.0055.0050.00

0.00+------------,----'---'----------r-----------,

45.00------7 s

200.00

mmHg

1150.00

RTF

100.00

50.00

c

50.00 55.00--7 s

45.0040.00

0.00+--------,r------L------,--------.----------,

35.00

Figure 1.7: Outliers from figure 1.3,a (case 1): low inflationb (case 2): short disappearance ofK-soundsc (case 3): weak K-sounds.

In figure 1.7a the inflation pressure is too low so K-sounds may not be correctly detected. Afterfirst RTF, systolic blood pressure decreases and K-sounds disappear for a while.The case of figure 1.7b is similar to figure 1.7a. The inflation pressure is also too low. The firstmarker agrees better with RTF. The K-sounds disappear for a short time due to the suddendecrease in blood pressure after RTF so a second systolic marker is placed when the K-soundsreappeared.In the case of figure 1.7c the RRK marker for systolic blood pressure is too late with regard toRTF, due to weak K-sounds.

1.4.5 RTF versus lAP

Although correlation between the RTF-value and the corresponding systolic lAP in thecontralateral arm is high (r =0.99), the RTF-values underestimate lAP considerably. The increasein the difference as systolic blood pressure rises, is systematic, at approximately 6% ofsystoliclAP.Studies have shown that indirectly measured systolic arterial pressures consistently underestimatedirect readings with discrepancies quoted from 3 to 24.6 mmHg [2]. It is clear that by their nature

16

indirect measurements are at a disadvantage for accuracy. This may be caused in part since directmeasurement reflects in fact pressure. Indirect measurement, on the other hand, is flow dependentand reflects an acceleration of flow causing a lateral pressure on the arterial wall [2]. In fact, thepressure in the ann cuff at the moment ofRTF is always lower than the systolic lAP causing theRTF pulse, because otherwise there could not be a RTF pulse at all.

One could also compare the RTF-value with the average of systolic lAP of several beats orseconds before inflation of the arm cuff. Or with the average of systolic lAP during the inflationand deflation of the arm cuff. In this particular group of patients the systolic lAP at Korotkoffphase I was 2.9 ± 3.9 mmHg higher than the average systolic lAP during the RRK measurementand 5.1 ± 5.6 mmHg higher than the average systolic lAP during a control period of30 secondsjust before cuff inflation. So the systolic lAP at the moment ofK1 (and at the moment ofRTF)is higher than the average systolic lAP during the surrounding period. Therefore, the generallyobserved underestimation of the systolic lAP by RRK readings will decrease when the averagesystolic IAP over a certain period is used, rather than a single value of the lAP at Kl (or at RTF)[6].

In figure 1.5 there are four cases with large deviations from the mean difference. Themeasurements ofthe cases with label 1 and 2 are from the same patient and so are the cases withlabel 3 and 4. The recordings of the first two cases are shown in figure 1.8a and 1.8b.

300.00

mmHg

1200.00

RTF

100.00

a

I

II

II

II

II

I

r-r-......-.~-_....... ........~----++-..,............,..-----~I

II

0.00 --'------,-------r-'--L--l-----.,........,:,.......---------,

44.00 46.00 48.00I -7 SI

v48.21 s

50.00

17

300.00 RTF

100.00

bI

1

I

0.00 I

I

40.00 45.00I

50.00 55.00I

I--7 sI

v48.59 s

300.00 RTF

mrnHg

I200.00

100.00

I I "o.oot=:

I'\.

-it __r--J "'-c

40.00 42.00 44.00 46.00 48.00~s

18

d

300.00

nunHg

I200.00

100.00

RTF

0.00 --'------r-----.---------.-----L-J...+-------,

54.00 56.00 58.00 60.00----7 s

62.00

Figure 1.8: Outliers from figure 1.5,a (case 1) and b (case 2): decision of manual RTF is too earlyc (case 3) and d (case 4): patient with considerable systolic blood pressure variability.

The manual RTF detection is too early in both cases. When RTF is taken at 48.21 s in the caseof figure 1.8a and at 48.59 s in figure 1.8b, the differences between the RTF-value and thepreceding systolic lAP are -11 and -10 mmHg respectively.The cases with label 3 and 4 are shown in figure 1.8c and 1.8d. In both cases RTF detection iscorrect but systolic blood pressure in this patient has considerable variability.

In most cases the systolic peak ofthe lAP recording is far above cuff pressure before a RTF pulsecan be seen. A certain threshold value appears to be necessary between arterial pressure proximalto the arm cuffand the pressure in the arm cuff (potential energy) to cause a pulse ofblood flow(kinetic energy) distal of the arm cuff. In these cases it seems that this pressure difference is notsufficient to send the pulse completely under the arm cuff. This may be an indication that the cuffis too wide. This could explain in part the underestimation of the RTF-value compared with thesystolic lAP. It is not clear, however, why this underestimation increases as blood pressureIncreases.

19

1.5 Conclusion

The scatter in data between manual RTF-value and systolic RRK blood pressure is mostly due tothe difficulties of the RRK method, especially-applied to an elderly population.Comparison of the manual RTF-value with systolic IAP gives better precision, but there is a largebias. If the difference between the manual RTF-value and systolic lAP is corrected for this biasof-12 mmHg, there is an acceptable degree ofagreement between both methods (95% confidenceinterval from -10.7 to lOA mmHg). The manual RTF-value underestimated systolic lAPsystematically (6% for the analyzed range ofsystolic IAP from 127 to 253 mmHg), thus the RTF­value can be used if corrected for this 6% underestimation.

20

2 AUTOMATIC RETURN-TO-FLOW DETECTION

2.1 Description of the method and software

Software has been written to implement an automatic detection of the RTF-value out of the fingerpressure signal of Finapres. The language used for this program is Modula-2 [32]. It is thesuccessor of Pascal. The program consists of two modules: 'detRTF' containing the mainprogram and 'RTF' containing procedures and functions that are used in the main program. InModula-2 the exported variables and procedure definitions are stored in a separate file (.DEF).The program works offline. Data are read from a so-called ADC-file (analog digital converted)[29]. The ADC-files in our case contain data of four analog channels, sampled at 100 Hz, with2.5 mV resolution:- channel 1 =arm cuff pressure (dominant arm)- channel 2 = finger cuff pressure (left hand)- channel 3 = finger cuff pressure (right hand)- channel 4 =intra-arterial brachial pressure (non-dominant arm).Depending on whether the left or right arm is dominant either channel 2 or 3 gives the fingerpressure. The samples in the ADC-files are two's complement 16 bits integers. An integer valueof400 units corresponds with 1 V analog signal and with 100 rnmHg pressure.For comparison with the manual RTF-value and with lAP systolic pressure, the ADC-files of thepatients described in chapter 1 have been used. With the aid of the program 'WAVEVIEW' [29]the RRK measurements were extracted from the ADC-files so each new file consisted of onecomplete RRK measurement. One of the two channels for finger pressure (channel 2 or 3)contains no data. The program reads the data samples from the file, displays the signals on screenand determines the moment ofRTF. The corresponding pressure in the arm cuff is denoted as theautomatic RTF-value (figure 2.1).Our algorithm for the detection ofRTF is quite simple. A moving average of255 finger pressuresamples is calculated. When the arm cuffhas start deflation the minimum ofthis moving averagecurve is determined simply by comparing the new calculated average with the previous one. Theaverages are compared in rnmHg. If the new calculated average is higher than the previousaverage, the RTF-value is taken as the arm pressure that is just read from the ADC-file. This isshown in figure 2.1.

21

200.00

nunHg

I150.00

arm cuffpressure

~ automatic RTF-value

marker for Kl

100.00

Finapres

51.0050.00 52.00 53.00) ---7 s

------, -----------~

previous ~indow of255finger pressure sampies

I

49. aI

I.,

0.00 +- -r++ .-- ----. '+-----,---..JL..l-__~

48.00

current window of255finger pressure samples

IF (average of current 255 finger pressure samples) >(previous average of255 finger pressure samples)

THEN (RTF is found)

Figure 2.1: Plot ofthe cuffand finger pressue showing the algorithm for the automatic detectionofRTF.

The moving average curve is also displayed on screen when the program runs. With the abovementioned method the first local minimum of this moving average curve is determined. Themoment ofRTF is chosen at the end ofthe moving average window of255 samples, whereas the'IOC"l rnuum'• ,. •• - .. - _.,"- - :-- .--.- -- _••_.- 1: __ ---- .t._ _ :ri..11" "l'+t.:" ,,';n..-lnu, Thp tirnp ~t thp.... urn Ul Ul~ 1I1UVUI~ C1V~1 C1~C \,oUI V~ U,",,, U\,o<o' aU\,o uu v .. ~ ~~ ••• _ •• _ •••••__••__ ."

end of the window is the Cuffent moment in which the last four samples \vere read (o~e of el!chchannel). The RTF-value is the current arm cuff pressure sample that is just read out from theADC-file. This procedure is shown in figure 2.2.

22

60.00

mmHg

i

RTF

55.00

50.00 --lI---4-1-'-\1lr.J4IhA I

middle of current window

finger pressure curve

moving averagecurve

window of255 samples

50.00

,/

previous average =49 rnmHg

51.00 52.00

current average = 50 rnmHg

Figure 2.2: The moment of RTF is chosen at the end of the window of the moving averagethat is higher than the previous average.

The moment ofRTF and the RTF-value are displayed on screen. This moment ofRTF is displayedas the time relative to the beginning of the ADC-file ofone RRK measurement. A typical screendisplay is shown in figure 2.3.

23

1&9 5. 16'30E-+-1

Figure 2.3: Screen display of 'detRTF' run. Numbers in upper-left comer are automatic RTFpressure (mmHg) and moment of RTF (s).

The RTF time is written to a so-called event-me (.EVT). An EVT-file is an ASCII-file containingtime values and labels. This me can be read by the program 'WAVEVIEW' to mark certain eventsin the data of an ADC-file. This written time value is given the label 'RTF'.The moving average values and the corresponding relative time are also written to a file. This isa so-called user-file (.USR). This USR-file can be read by 'WAVEVIEW' so the moving averagecurve can be watched in detail using 'WAVEVIEW'.

2.2 Results

2.2.1 Comparison of the automatic RTF-value with the manual RTF-value

The RTF-vaiues delemlim:u by the software desCiibcd in the previous paragraph h<!ve beencompared with the manual RTF-values described in chapter one. The data of the same patientgroup described in chapter one was used.A scatter plot is shown in figure 2.4 and in figure 2.5 the difference against the mean of the twoRTF-values is displayed.

24

300,------------------------"

250

.....lIII:=a! 200

150C)....-'

Clj

8 100o-'::lClj

50 n =28r = 0.985 (p < 0.001)Y = 0.976 *x + 4.43

o 50 100 150 200 250

manual RTF (mmHg)

300

Figure 2.4: Scatter diagram ofautomatic RTF vs. manual RTF pressure.

30--,--------------------,

'Iii:=a!~ 20­

E-<P::: mean + 2.sd = 13.2

• 2

• 1-Clj 10-::lr::::Clj

8 0 mean = -0.07

~Eo-< -lO-

P::: mean - 2 .sd = -13.3

.. . . ...

C).... n =28-'Clj -20- mean =-0.078 sd =6.62o nUn =-21-' max = 19::l -30 IClj 0 50 100

(automatic RTF150 200 250 300

+ manual RTF)/2 (mmHg)

Figure 2.5: Bland-Altman diagram ofautomatic RTF and manual RTF pressure.

25

As can be seen in figure 2.4 the correlation between the automatic and the manual RTF-value ishigh. Figure 2.5 shows that the mean difference is small but the precision is not really good causedby four outliers. These outliers are labeled with numbers in figure 2.5. Case number one is shownin figure 2.6. To make the moving average curve better visible it is plotted with an offset of 10mmHg (right axis).

100.00

0.00 -'-------.------r---------.--------'-------,,-'-~-~_J

100.00

300.00 a mu a 300.00

nunHgt nR R

IT TF F

200.00200.00

30.00 35.00 40.00 45.00----7 s

50.00

Figure 2.6: Outlier from figure 2.5 (case 1). Finger pressure stays low for a long time due tothe high cuff inflation.

Due to the high cuffinflation (about 65 mmHg above systolic pressure) finger pressure stays lowfor a long time. Hence there has to be only a slight increase in finger pressure to cause thecalculated average over 255 samples to be higher than the previous average. This slight increasein finger pressure could also be a disturbance or moving artifact instead ofthe RTF pulse. Besides,systolic lAP shows large variability. As can be seen the systolic lAP drops after the moment ofRTF detection (autRTF) so the RTF pulse (manRTF) appears later.Case number tv/c is sl.'Y'.ilar to case one, that is finger pressure stays low for a long time althoughthe cuff inflation is not too high (about 50 mmHg above systolic pressure).In the cases three and four cuff inflation was too low. Case number three is shown in figure 2.7.

26

100.00

moving averagecurve

0.00 --'------,,-----,....------'-----.-"'-----T-------r--'

100.00

300.00 a300.00u

tmmHg R

T

I F

200.00200.00

45.00 50.00 55.00 60.00 65.00-----7 s

Figure 2.7: Outlier from figure 2.5 (case 3). Moving average is still decreasing after manualRTF due to the low inflation. The moving average curve is plotted with an offsetof 10 mmHg (right axis).

The RTF pulse (manRTF) appears just after the start of cuff deflation, but this pulse is notdetected by the program as RTF because the moving average is still decreasing. The minimum ofthe moving average curve lies further on (58.6 s).From this case it is clear that the arm cuffhas to be inflated high enough to let the finger pressuredrop to its baseline. But from the first two cases one could conclude that the inflation must notbe too high. The problem of these cases could be avoided by checking if the moving averagecurve keeps rising after the detection ofRTF. If not then RTF was detected too soon.When the four outliers are omitted the mean difference between automatic and manual RTF-valueis -0.29 ± 1.97 mrnHg.

2.2.2 Comparison of the automatic RTF-value with intra-arterial systolic pressure

Because lAP measurement is denoted as the 'gold standard' to some researchers, the automaticRTF-values have also been compared with the brachial systolic lAP, measured in the non­dominant arm. The results are plotted in figure 2.8 and 2.9.

27

300 -,---------------------------"

250......IlO

::z::S!

200

~

Eo-<0::C,) 150....

+oJ~

S //

/

0 100 /

+oJ /

;:l /

~/

50

o 50 100 150 200

lAP (mmHg)250 300

Figure 2.8: Scatter diagram of automatic RTF vs. systolic lAP.

30 -,-------------------------,

OJ) 20­:ta!

0.. 10­<..- mean +2*sd = 5.30

• 1

0+------------------------;

.•...

. ..8 -10- ~e~_=.:l~.~ ._ ~ .: • _+oJroSo .~ -2°1 . I

'" .,01 ~~.,~~~~:~~:~,.,.,. ,.,.,.,.,.,.,.,.,.,.,.,.,.,.,.,., jo 50 100 150 200 250 300

(automatic RTF + lAP}/2 (mmHg)

Figure 2.9: Bland-Altman diagram ofautomatic RTF and systolic lAP.

28

From the scatterplot it can be seen that the correlation between the automatic RTF-value andsystolic lAP is high. The regression line shows that there is a systematic underestimation ofsystolic lAP by RTF of [(lAP - RTF)IIAP] *100% = (9.7/182)*100% = 5.3%.There is one outlier labeled with number one. This is the same case as number two in figure 2.5.When this case is omitted the equation of the linear regression line becomes y = 0.946*x - 1.19and the systematic underestimation is (11.0/182)*100% = 6.0%. The mean difference thenbecomes -11.7 ± 5.0 mmHg.These results could be expected because the automatic RTF compared well with the manual RTFand from paragraph 1.3.2 it was found that the manual RTF also underestimated systolic lAPabout 6% and that the mean difference was -12.0 ± 5.3 mmHg.

2.3 Discussion

2.3.1 Detection algorithm

The algorithm for the detection ofRTF originated from the observation that the RTF pulse liesnear to the global minimum of an average curve through the finger pressure signal. The choicefor the window width started from the idea of averaging at least one hart beat. With a normalheart rate of60 beats/minute this would result in averaging 100 samples (sample frequency is 100Hz). A window width of 127 samples appeared to be too sensitive so RTF was detected too sooncompared with the manual RTF. The influence of little disturbances from background noise wastoo big. With a window width of255 samples this problem was nearly redressed. Only when theinflation was high the finger pressure signal dropped to its baseline and stayed low for a relativelong time, so little disturbances could still be detected as RTF. When the cuff inflation for the RTFmeasurement is automated (see chapter 3) the inflation height will be derived from the knownsystolic finger pressure measured by Finapres so there will be a good indication for the requiredinflation height. Besides, an improvement to the RTF detection algorithm can easily be obtainedby checking if the moving average keeps rising after the detection of RTF for a period of forexample two heartbeat intervals. If not then the detection of RTF will continue until the rightmoment ofRTF is found. Then the old RTF-value can be replaced by the new one. If the movingaverage curve will not rise at all then there will probably be a failure in the finger pressuremeasurement and a complete new RTF measurement should be done by inflating the arm cuffagain.A window width of511 samples has also been tested but then RTF was detected too late probablybecause the algorithm is too insensitive for the RTF pulse and the fact that the moment ofRTFis taken at the end of the moving average window. This moment should be chosen earlier whenthe window width is enlarged.

It is not tested what the influence is ofthe heart rate on the detection algorithm. One can imagingthat when the heart rate is low, the smoothing will be less if the width of the moving averagewindow stays the same. This might affect the RTF detection. The window width could be madeheart rate dependent ifthe influence of the heart rate would prove significant.

29

2.3.2 Automatic RTF versus manual RTF

When the automatic RTF-value is compared with the manual RTF-value it must be noted that thedetennination ofthe manual RTF-value is done by inspecting the finger pressure signal with thehuman eye with the aid ofthe program 'wAVEVIEW'. This is a subjective method and when thecuff pressure curve just passes a top of the brachial artery pressure curve, so cuff pressure isexactly equal with systolic arterial blood pressure, this cannot be seen with the human eye in thefinger pressure signal. With an automatic detection algorithm one could adapt for this problem.

2.4 Conclusion

Automatic detection ofRTF using Finapres is possible. The described algorithm is simple and theresults are good (automatic RTF versus manual RTF). The automatic RTF is similar to the manualRTF when compared with lAP.The described RTF detection algorithm can be improved by checking if the moving average of255 finger pressure samples keeps rising after the detection ofRTF.

30

3 INFLATIONIDEFLATION TECHNIQUE OF THE ARM CUFF

3.1 Introduction

To read brachial systolic blood pressure with the return-to-flow method an inflatable cuff is placedaround the upper arm. This arm cuffis rapidly inflated above brachial systolic blood pressure andthen slowly deflated until the return-to-flow pulse is detected by Finapres. Then the arm cuff israpidly deflated. This inflation and deflation has to be automated. The device has to be robust andnot too expensive so no complex control system and hardware should be used.Inflating a pressure cuff at the upper arm is an unpleasant feeling and most automatic bloodpressure devices make too little allowance for this inconvenience. In this chapter the requirementsfor inflation and deflation ofthe arm cuffwill be described and a hardware implementation, whichmeets the requirements and considers the patient feelings, will be discussed.Experiments with this implementation are described in the report of A Sellmeijer and AP.Brouwer "Vulsysteem voor bovenarm-bloeddrukmanchet" [23]. The results and discussion in thischapter is based in part on this work.

3.2 Requirements

3.2.1 Inflation height

When the arm pressure cuff is applied to the upper arm the device must be able to reach a cuffpressure up to 300 mmHg. Sphygmomanometers usually don't exceed a limit of250 mmHg andfor automatic devices maximum cuff pressure shall never exceed 330 mmHg because of safetyrequirements [31]. It is assumed that the cuff is wrapped tight around the upper arm.Arm cuff pressure must be inflated to approximately 30 mmHg above systolic brachial pressure[20]. The inflation height will be derived from the systolic finger blood pressure measured byFinapres. In the elderly, Finapres has the tendency to underestimate systolic pressure [22]. Thusarm cuff pressure has to be inflated higher above systolic (finger) pressure in the elderly.

3.2.2 Inflation time

The time necessary to reach full inflation must be short because the filling of the vascular treedistal to the arm cuffis an unpleasant feeling. The caused pain can raise the blood pressure of thepatient so systolic pressure may be measured too high. Besides it is desirable to do a RTFmeasurement as quickly as possible because during the RTF measurement the continuous fingerblood pressure is not obtainable.The inflation time has been chosen to be approximately within five seconds, even when the patienthas a thick arm.

31

3.2.3 Patients inconvenience

Most automatic non-invasive blood pressure devices make no allowance of the inconveniencecaused by the cuff inflation. The inflation curve is often slow initially but very steep just beforereaching the end pressure (figure 3.1a), which can be painful if this end pressure is high. Toconsider the patients feeling we chose to inflate the arm cuff according to a S-curve. This isshown in figure 3.1b.

a

b

mmHgl

III

I

I

e90mmn­I

I

I

>sFigure 3.1: Inflation curves for the arm cuff, a: linear and steep at the end

b: S-curve wich is less steep at the end.

When the last part of the inflation curve is flattened, the end pressure will be reached moregradual so the patient is less stressed nnd feels less disco!!1iort.With tllls S-curve the ltilear pa.-t in the rrJdd!e is steeper than in figure 3.1 a if the end pressure hasto be reached in the same time. This steepness is less problematic because arm cuff pressure islow.The non-linear part at the beginning of the inflation curve is inevitable. This non-linear part iscaused by the compliance of the arm cuff and arm. When the arm is thick and!or when the armcuffis not applied tight enough this non-linear part will last longer because more tissue has to becompressed and/or more air supply is necessary to built up the pressure in the arm cuff

32

3.2.4 Deflation rate

The deflation has to be nearly linear with a rate ofapproximately 2.5 mmHg/s [20]. This deflationrate is required to give adequate measurement precision. When heart rate is 60 beats/minute andsystolic blood pressure stays on the same level, the measuring error will be uniformly distributedaround 1.25 mmHg.

3.3 Implementation

A hardware implementation for the inflation and deflation of the arm cuff is shown in figure 3.2.

air pump

volumechamber

arm cuff

Finapres

computer

51

pressuretransducer

analog~gita1

converter

R2

flowrestrictor

hill.----.,.

52

deflationvalve

pressuretransducer

Figure 3.2: Hardware implementation for inflating/deflating the arm cuff

A volume chamber provides the air supply necessary to inflate the arm cuff fast enough so a small(power and size) air pump can be used which makes little noise. Besides, this volume chamberalong with the flow restrictor (R2) provides an inflation according to a S-curve. The controlsequence of the valves is as follows: At first S1, S2 and S3 are closed. When S1 is opened thepressure in the volume chamber will rise. Ifthe right end pressure in the volume chamber has beenreached, 52 will be opened to inflate the arm cuff from the volume chamber. Ifnecessary 51 canstay open to support inflation. When S1 is closed the last part of the cuff inflation will be causedby the deflation ofthe volume chamber so the arm cuffwill reach its end pressure gradually as thepressure in the volume chamber will decrease. When the pressure in the arm cuff and in thevolume chamber are equal, 52 can be opened to deflate the arm cuff through the deflation valve.To provide a linear deflation rate of2.5 rnmHg/s a valve normally used for Riva-RocciIKorotkoffmeasurements is chosen with a fixed adjustment. With this valve the inflation will be accordingto a decreasing exponential curve so the first part of the deflation curve will be almost linear.When RTF is detected S3 will be opened to deflate the arm cuff fast.

33

To inflate to different end pressures in the arm cuff, the pressure in the volume chamber and inthe ann cuffare continuously measured and the valves are controlled depending on the measuredpressures. Experiments have been carried out with this hardware implementation to find the rightalgorithm for the control of the valves and to determine the influence of several parameters likethe size of the volume chamber, the flow supply of the air pump and the value of the flowrestrictor. These experiments are described in the report ofBrouwer and Sellmeijer [23]. In figure3.3 two measurement examples from this report are shown. To be able to reach the maximumrequired end pressure of300 rnmHg with the chosen air pump, a volume chamber ofone litre wasrequired. In the measurements offigure 3.3 a 0.33 litre volume chamber was used.

34

35.0030.0025.00

pressure involume cltamber

I

I,IIII,II

II,I

I

S2 closed I

(defl.)~ r~;,)~II,IIII,I

-->..,. s

III

III,

Slo":.--J Sl closed

~

0.00+---...1....-----.----'--....---------,20.00

200.00

100.00

400.00

mmHg

i ~oo

a

100.00

b

400.00

mmHg

i 30000

200.00S2 closed :(defl.) I ~2 open~ (mfl.)

:---7,I

I

I

IIII

O.oo-'----'''''''F''----------,.------~--

15.00 20.00-->..,. s

25.00

Figure 3.3: Two measurements with the hardware implementation of figure 3.2(volume chamber = 0.33 litre),a: arm cuff wrapped tight,b: arm cuffwrapped loose around the upper arm.

In both cases the control ofthe valves was the same with equal adjustments of the flow restrictor.The difference in end pressure and duration of the two inflation curves is caused by the differentmanner in applying the cuff around the upper arm. In figure 3.3a the cuffwas wrapped tight butin figure 3.3b the cuffwas wrapped loose around the upper arm (this simulates a thick arm). Fromthis example it is clear that the influence of thick or thin arm and tight or loose wrapping issubstantial.

35

3.4 Discussion

As shown in figure 3.2 no complex hardware is necessary to meet the requirements for inflatingand deflating the arm cuff With the use of a volume chamber it is possible to inflate the arm cufffast enough. Besides, the volume chamber along with the flow restrictor provides the rightinflation curve (S-curve). The needed room for the volume chamber (one litre) is not a problem(2.5 * 20 * 20 em).When a cuff is applied, one has to be sure that an arm cuffof the proper size according to therecommendations [20] is used and that this cuff is wrapped tight enough around the upper arm.To reach the right end pressure in the arm cuff (30 mmHg above systolic blood pressure) thevalves have to be opened or closed depending on the measured pressure in the volume chamberand arm cuff and/or certain time lags. First the air in the volume chamber has to be compressedto a certain starting pressure. Then the arm cuff is inflated (S2 open) and pumping may be stillnecessary for a certain time period or until a certain pressure in the arm cuff has been reached. Atthe last part ofthe inflation curve the pump must be switched off to level off to the end pressure.To reach a higher end pressure then 30 mmHg above systolic blood pressure, which is necessaryfor the elderly, the control of the valves can be made age dependent. The starting pressure in thevolume chamber has to be higher and/or extra pumping must last longer with increasing age ofthe patient.It is difficult to take full account of the compliance of the arm cuff and arm. One has to be surethat the planned end pressure in the arm cuffwill still be reached even when the arm is thick. Inthe experiments ofA.P. Brouwer and A. Sellmeijer measurements have been done with the cuffwrapped tight around the upper arm as well as measurements with a loose cuff to simulate a thickarm. The starting pressure in the volume chamber and extra pumping time have been set upbetween these two extremes for the different end pressures that have to be reached in the armcuff One could (additional) accomplish for this problem by making the starting pressure in thevolume chamber and/or the extra pumping time dependent on the arm circumference of thepatient.

36

LITERATURE REFERENCES

1. Altman D.G.Practical statisticsfor medical research.London: Chapman and Hall, 1991

2. Bailey R.H. & J.H. BauerA review ofcommon errors in the indirect measurement ofbloodpressure.Arch. Intern. Med., vol. 153 (1993), p. 2741-2748

3. Bergen F.H. van & D.S. Weatherhead, AE. Treloar, AB. Dobkin, J.1. BuckleyComparison ofindirect and direct methods ofmeasuring arterial bloodpressure.Circulation, vol. 10 (1954), no. 4, p. 481-490

4. Bland 1.M. & D.G. AltmanStatistical methodsfor assessing agreement between two methods ofclinicalmeasurement.The Lancet, vol. 1 (1986), p. 307-310

5. Block F.E. Ir. & M.V. Fletcher, T.1. Morris, R. DzwonczykA clinical evaluation ofrapid automatic noninvasive bloodpressure determination withthe Ohmeda 2120 "return-to-flow" method.1. Clin. Monit., vol. 7 (1991), p. 241-244

6. Bos W.IW. & J. van Goudoever, G.A van Montfrans, KH. WesselingInfluence ofshort-term bloodpressure variability on bloodpressure determinations.Hypertension, vol. 19 (1992), p. 606-609

7. Bos W.1.W & I van Goudoever, KH. Wesseling, G.AP.1.M. Rongen, G. Hoedemaker,IW.M. Lenders, G.A van MontfransPseudohypertension and the measurement ofbloodpressure.Hypertension, vol. 20 (1992), p. 26-31

8. Bos W.IW. & 1. van Goudoever, G.A van Montfrans, AH. van den Meiracker, KH.WesselingReconstruction ofbrachial artery pressure from noninvasive finger pressuremeasurement.submitted for publication 1995 and in:BosW.J.W.Measurement offinger and brachial artery pressure.Academisch proefschrift, Universiteit van Amsterdam, 1995

9. Caulkett N.A & T. Duke, 1.V. BaileyA comparison ofsystolic bloodpressure measurement obtained using a pulse oximeter,and direct systolic pressure measurement in anesthetized sows.Can. 1. Vet. Res., vol. 58 (1994), no. 2, p. 144-147

37

10. Chawla R. & V. Kumarvel, K.K. Girdhar, AK. Sethi, A. Indrayan, A BhattacharyaCan pulse Oximetry be used to measure systolic bloodpressure?Anesth. Analg., vol. 74 (1992), p. 196-200

11. Gardner R.M.Direct bloodpressure measurement - Dynamic response requirements.Anesthesiology, vol. 54 (1981), p. 227-236

12. Geddes L.AThe direct and indirect measurement ofbloodpressure.Chicago: Year Book Medical Publishers, 1970

13. Gizdulich P. & K.H. WesselingReconstruction ofbrachial arterial pulsationfrom finger arterial pressure.In: Proc. 12th lnt. Conf IEEE Eng. Med. BioI. Soc., vol. 12 (1990), p. 1046-1047

14. Irnholz B.P.M.Noninvasive finger arterial pressure waveform registration, evaluation ofFinapres andPortapres.Academisch proefschrift, Universiteit van Amsterdam, 1991

15. Kemmerer W.T. & R.W. Ware, HF. Stegall, W.E. EvansIndirect measurement ofhuman bloodpressure by the Doppler ultrasonic technique.Surg. Forum, vol. 18 (1967), p. 163-165

16. Kirby R.R. & W.T. Kemmerer, J.L. MorganTranscutaneous Doppler measurement ofbloodpressure.Anesthesiology, vol. 31 (1969), p. 86-89

17. Korbon G.A & M.H. Wills, F. D'Lauro, D. LawsonSystolic bloodpressure measurement: Doppler vs. pulse oximeter.Anesthesiology, vol. 67 (1987), no. 3A, p. A188

18. Langbaum M. & F.G. Eya!A practical and reliable method ofmeasuring bloodpressure in the neonate by pulseoximetry.J. Pediatr., vol. 125 (1994), p. 591-595

19. Lee 1. & D. Koh, C.N. Dng('/,..I;~/;,.,..I "",.I.,,.ti,.. .. ,..1' ,.rno""..." ..1 I.olu/oo'" hA/n nJolhrvlf: (nr mPflf:lIrincr fl n1Jnntitntivp..UIW.....,.'''"'w.. ~ ........." ......v,. VJ ....6' .......... '·· ..... ,·· _ ..... , .. _-, ..... - ..._ ........-.",J" • ..• - ............. ~ .. ··0 -- -JO---------- •

•• ~_;~J..ln...."""UU.. c:..

Comput. BioI. Med., vol. 19 (1989), no. 1, p. 61-70

20. van Montfrans G.A & W.H. BirkenhagerBloeddruk meten.Hart Bulletin, vol. 21 (1990), p. 164-167

38

21. Riva-Rocci S.Un nuovo sfigmomanometro.Gaz. med. Torino, vol. 47 (1896), p. 981-996

22. Rongen G.A & W.1.W. Bos, J.W.M. Lenders, G.A van Montfrans, H.I.J. van Lier, 1.van Goudoever, KH. Wesseling, T. ThienComparison ofintrabrachial andfinger bloodpressure in healthy elderly volunteers.Am. J. Hyp., vol. 8 (1995), p. 237-248

23. Sellmeijer A & AP. BrouwerVulsysteem voor bovenarm-bloeddrukmanchet.report of 'Eerste Vrije Onderzoek natuurkunde practicum voor N-studenten'TV Delft, 1995

24. Talke P.O.Measurement ofsystolic bloodpressure usingpulse oximetry during helicopterflight.Crit. Care Med., vol. 19 (1991), no. 7, p. 934-937

25. Talke P.O. & R.J. Nichols, D.L. TraberDoes measurement ofsystolic bloodpressure with a pulse oximeter correlate withconventional methods?I Clin. Monit., vol. 6 (1990), p. 5-9

26. Wallace C.T. & ID. Baker, C.C. Alpert, SJ. Tankersley, IM. Conroy, R.E. KernsComparison ofbloodpressure measurement by Doppler and by pulse oximetrytechniques.Anesth. Analg., vol. 66 (1987), p. 1018-1019

27. WareR.W.New approaches to the indirect measurement ofhuman bloodpressure.presented at the Third National Biomedical Sciences Instrumentation Symposium, 1965

28. Wesseling KH.Finapres™ model 5.User manual, TNO BioMedical Instrumentation, Amsterdam, 1987

29. Wesseling KH.The FAST-mfl-cZ system.User manual, TNO BioMedical Instrumentation, Amsterdam, 1993

30. Wesseling KH. & J.1. Settels, WJ.A GoedhardNieuwe ontwikkelingen in het bloeddruk meten, in het bijzonder bij oudere mensen.In: Goedhard WJ.A & D.L. Knook

De bloeddruk op oudere leeftijdAlphen aan den Rijn: Gerontologische Reeks, 1984

39

31. White W.B. & A.S. Berson, C. Robbins, M.l Jamieson, L.M. Prisant, E. Roccella, S.G.ShepsNational standardfor measurement ofresting and ambulatory bloodpressures withautomated sphygmomanometers.Hypertension, vol. 21 (1993), p. 504-509

32. WirthN.Programming in Modula-2.New York: Springer-Verlag, 1985

33. Wong D.T. & G.A. Volgyesi, B. BissonnetteSystolic arterialpressure determination by a new pulse monitor technique.Can. l Anaesth., vol. 36 (1992), no. 6, p. 596-599

40