Frequency response of implantable blood pressure telemetry systems

TECHNICAL PAPER

Frequency response of implantable blood pressure telemetry systems

Daniel McCormick,* David M Budgett,*†¶ Matthew K Lim,†¶ Bruce Van Vliet,‡ Poul F Nielsen,*)

Aiguo P Hu§¶ and Simon C Malpas*†¶

*Auckland Bioengineering Institute, §Department of Electrical and Computer Engineering, )Department of Engineering

Science, The University of Auckland, †Telemetry Research Ltd, Auckland, New Zealand and ‡Faculty of Medicine, MemorialUniversity of Newfoundland, St John’s, Newfoundland, Canada

SUMMARY

1. Measurement of blood pressure via telemetry in a variety ofanimal species has become an indispensable part of cardiovascu-lar physiology, drug development and safety pharmacology.2. These telemetry systems use fluid-filled catheters, which dif-

fer from commonly encountered indwelling catheters by virtue oftheir short length, high durometer, variable inner diameter andthe use of a gel interface at the distal tip.3. Despite the widespread use of blood pressure telemetry,

there is little information describing the frequency response ofthese systems. The frequency response is of importance becauseit determines how well the waveform dynamics, such as pulsepressure, are captured.4. For this reason, we measured the frequency responses of

commonly used telemeters manufactured by Data SciencesInternational (St Paul, MN, USA; namely PA-C10, PA-C40 andPA-D70) and Telemetry Research (TR43P). The mean (± SEM) –3 dB frequencies measured for the PA-C10, PA-C40, PA-D70and TR43P telemeters were 57 ± 2 Hz, 40 ± 6 Hz, 32 ± 2 Hzand 173 ± 3 Hz, respectively.5. Simulation of the devices’ dynamic performance by apply-

ing their frequency responses to a high-fidelity recording ofarterial pressure demonstrated that the devices have sufficientbandwidth to accurately record arterial waveform dynamics.Experiments were also performed to determine how routine lab-oratory use and maintenance of the catheter affects the frequencyresponse of the telemeters. Provided no air bubbles were intro-duced, these showed that the telemeters’ frequency responseswere robust to the maintenance procedures of tip removal andgel application.

6. The frequency response measurements, combined withsimulation results, demonstrate that the systems tested haveadequate dynamic performance to record arterial pressure.Key words: arterial pressure, pulse pressure, telemetry.

INTRODUCTION

The measurement of arterial pressure is a fundamental aspect in theassessment of cardiovascular function. The broad concept of insertingfluid-filled catheters into an artery has remained essentially the samesince the first direct measurements of arterial pressure by StephanHales in 1733.1 The physics of fluid-filled catheter systems have beenwell described and practical design features that promote a high-fre-quency response (e.g. catheter stiffness, dimensions, fluid viscosity)have been identified.2 In recent years, measurement of blood pressurevia telemetry in a variety of laboratory animals has become an indis-pensable part of cardiovascular physiology, drug development andsafety pharmacology.3 The ability to record blood pressure in freelymoving animals in their home cages over extended periods of timehas opened new areas of research. Previous approaches had relied onchronic catheterization and associated catheter flushing to maintainpatency or tail-cuff measurements, each with their attendant prob-lems.3–5 The solution provided by Data Sciences International (DSI,St Paul, MN, USA) over 20 years ago is in the form of a catheterattached to a pressure sensor and a radio link to a remote receiver.6

This pioneering design features a catheter tip made of complianttubing filled with a viscous gel-like material. The gel prevents thepressure transmission fluid escaping from the catheter without intro-ducing a pressure offset and provides a biocompatible blood–fluidbarrier.7 However, the effect of the specialized tip on the dynamicresponse of the pressure sensor has not been reported. The frequencyresponse is of importance for the accurate recording of dynamicpressures, especially in small animals, in which heart rates may behigh. For example, the resting heart rate of mice is approximately8 Hz (480 b.p.m.) and has been recorded as high as 14 Hz(840 b.p.m.).8 Frequency response information is important not onlyfor validating the telemetric method for recording blood pressure, butalso for providing valuable information to the end user about thetreatment of the fluid-filled tip and its effect on the accuracy of thepressure waveform acquired.

Correspondence: Daniel McCormick, Auckland Bioengineering Institute,The University of Auckland, Private Bag 92019, Auckland, New Zealand.Email: [email protected]

¶These authors are employees and ⁄ or shareholders of Telemetry ResearchLtd.

Received 12 January 2010; revision 11 April 2010; accepted 14 April2010.

� 2010 The AuthorsJournal compilation � 2010 Blackwell Publishing Asia Pty Ltd

Clinical and Experimental Pharmacology and Physiology (2010) 37, 862–869 doi: 10.1111/j.1440-1681.2010.05406.x

The aim of the present study was to determine the frequencyresponse of commercially available catheter-based blood pressuretelemetry devices manufactured by Telemetry Research (TR, Auck-land, New Zealand) and DSI. In addition, we sought to examine thefactors affecting device performance. In particular, we consideredwhether the viscous gel tip aids or hinders catheter performance anddoes catheter bandwidth ultimately determine the frequency responseof the telemeters under investigation. Further analysis was thenundertaken on practical aspects that may affect dynamic perfor-mance: for example, does re-gelling reduce bandwidth and whateffect does cutting the tip off the catheter to remove a clot have onperformance?

METHODS

Test system for measuring high bandwidth pressure sensoroperation

A new measurement system was designed and built specifically to evaluateimplantable manometers. The system uses a voice coil actuator coupled to alow-compliance chamber filled with fluid to generate sinusoidal pressurewaveforms with bandwidth approaching 10 kHz. Details of the system areavailable as Data S1 for this paper. The system produces a pressure signal intowhich the catheter of the telemeter to be tested is inserted along with areference transducer (3 Fr Mikro-Tip, catalogue no. SPR-249A; Millar Instru-ments, Houston, TX, USA). Through Fourier transformation, the signal com-ponents at the injected frequency of the applied pressure (measured by thereference transducer; PREF(jx)) and telemeter output (PT(jx)) are determined.The magnitude response of the telemeter was calculated using Eqn 1:

kMðjxÞkdB¼ 20LogPTðjxÞPref ðjxÞ

�� ð1Þ

A Bode plot of the response was constructed by stepping through the fre-quency range of interest. Data were normalized against the DC gain of thetelemeter (estimated from a 1 Hz recording) so that responses could becompared irrespective of device sensitivity.

Protocol 1: Frequency responses of commercial telemeters

The frequency response of four telemeters (i.e. PA-C10, PA-C40 and PA-D70from DSI and TR43P from TR) were measured. Results for the DSI deviceswere recorded using an RMC-1 receiver (DSI) and converted to analogueform for data acquisition using an R11CPA adapter (DSI). Data for the TRdevice were recorded using a TR131 analogue reconstruction unit (TR). Priorto measurement, the catheter tips of all devices were inspected for bubbles andtopped up with gel when appropriate. To mimic implanted operating condi-tions, measurements were performed in a temperature-controlled chamber at37�C.

Recordings were made by sweeping a 2.4 kPa (18 mmHg) peak-to-peaksinusoidal pressure signal in discrete steps between 1 and 10 kHz. In all tests,41 logarithmically spaced frequencies per decade were measured, resulting ina total of 161 measurements per frequency response recorded. The systemwas set to acquire five complete periods at each injected frequency or200 ms of data, which ever was greater. This reduces noise susceptibility athigh frequencies by acquiring more cycles of the injected frequency whileallowing a response to be measured in a reasonable amount of time. Whererequired, a dead time was introduced to allow for transmission delay (seeDiscussion).

Statistical analysis was performed on the –3 dB frequencies extracted fromthe Bode plots using Matlab (MathWorks, Natick, MA, USA). Owing to thesmall sample sizes and a lack of adherence to the underlying assumptions ofthe parametric analysis of variance (ANOVA) test, a non-parametric ANOVA wasperformed using the Matlab function kruskalwallis.m.

Protocol 2: Importance of gel volume

Four 150 mm catheters, with dimensions shown in Fig. 1, were prepared andtheir frequency responses recorded as the quantity of gel in the distal tip wasvaried. To eliminate telemetry system effects, the catheters were attacheddirectly to solid state pressure sensors (NPP-301-15A; GE NovaSensor, Fre-mont, CA, USA) equipped with a wideband amplifier. The catheters werefilled with low-viscosity paraffin oil and their frequency response recordedunder a variety of gel tip conditions. Initially, no gel was applied; subse-quently, gel was introduced in 1 mm increments (represented by the variable xin Fig. 1) and, after each addition, the frequency response was recorded.

Protocol 3: Testing consequences of tip removal

The frequency response of a TR43P device was recorded as increasingamounts of the distal tip were removed, ranging from 0 mm (catheter unal-tered) to 7 mm (entire thin-walled section removed). Because of the similarityof the dimensions of the TR43P and DSI PA-C40 catheters, tests were onlyperformed on the TR43P. After removing a section of the distal tip, gel wasreapplied such that the gel half-filled the remaining section of the tip. Forexample, if 3 mm tip was removed, half the remaining 4 mm distal tip wasfilled with gel. This would allow the gel to provide a blood–catheter fluid bar-rier, as required for chronic use.

Protocol 4: Simulated dynamic performance

The average device transfer functions for the PA-C40 and TR43P (which areintended for use in rats) were extracted using the Matlab System IdentificationToolbox (MathWorks). Black box polynomial models using the model refer-ence output error method9 were fitted to the frequency response data, whichwere then converted to continuous time transfer functions of the form given inEqn 2 in terms of the polynomials B(s) and A(s).

GðsÞ ¼ BðsÞAðsÞ �

yðsÞuðsÞ ð2Þ

This is equivalent to the system output y(s) divided by its input u(s). Esti-mates of the suitability of the fit are given by Eqn 3 from the measured (ymeas)and simulated (ysim) responses.

fit ¼ 1001� kysim � ymeas kkymeas �meanðymeasÞk

� �ð3Þ

The transfer functions were then applied to a 1 min section of data recordedfrom the abdominal aorta of a 250 g Wistar rat using a Millar Instruments 3 FrMicrotip catheter sampled at 2 kHz. Experiments were performed with theapproval of the University of Auckland Ethics Committee. The self-resonantfrequency of the Microtip catheter was measured by gently tapping it on thebench while recording its output with an oscilloscope. The oscilloscope traceshowed a damped oscillation at approximately 9 kHz, indicating that thecatheter’s –3 dB frequency was greater than 9 kHz. Simulated waveformswere then analysed using ADInstrument’s Blood Pressure Module (ADInstru-

600 µm 300 µm i.d.

450 µm i.d.

Fig. 1 Dimensions of the catheters under investigation. The frequencyresponses of the catheters were tested as the amount of gel, x, was increased.

Frequency response of pressure telemetry 863


ments, Sydney, NSW, Australia) in order to estimate the effects of the deviceson pulse pressure. No statistical tests were applied to the simulated pressuredata because the models are deterministic.

RESULTS

Frequency response of commercial telemeters

Frequency response results for individual telemeters are plotted bydevice type in Fig. 2, along with a comparison of the mean frequencyresponses. In all cases, the frequency responses are flat in the passband and there is no evidence of resonance. This should be comparedwith the results of Protocol 3, which were measured directly at thepressure sensors’ output (without a telemetry system) where reso-nance is clearly visible. This suggests that the bandwidths of the cath-eters are higher than the bandwidths of the overall system becauseresonance would be visible if the catheters were the limitation. Atten-uation occurs in all devices at a rate much higher than the –40 dB ⁄decade that has been reported previously for indwelling catheters.2

At high frequencies (> 200 Hz) the traces level off, indicating thatmeasurement system noise is overwhelming the highly attenuatedpressure signal at the telemeter outputs. In the responses recorded forthe TR43P, artefacts of the sampling filter response of analogue todigital conversion are visible as a dip in the frequency response at thetelemeter’s sampling frequency of 500 Hz.The –3 dB frequencies for each telemeter were extracted from the

frequency responses and are summarized in Table 1. The resultsshow good agreement between the bandwidths of individual deviceswithin a telemeter type, except for the PA-C40 where one device (no.3) exhibited approximately half the bandwidth of its kin. This devicewas inspected for bubbles in its catheter; none were found and thecause of the reduced bandwidth is unknown.Figure 3 shows the measured bandwidth of 16 devices grouped

into the four model numbers from the data in Fig. 2. The mean band-width of TR43P (173 Hz) is approximately threefold higher thaneach of the other devices. The results of this test are given in Table 1and confirm the existence of a difference between the four groups(P = 0.005).The bandwidth for the DSI devices ranged from 32.5 Hz to

56.5 Hz, the –3 dB point being inversely proportional to device sizeand battery life. However, the bandwidths are proportional to theexpected heart rate of the types of animals they are intended to beused in, with the mouse-specific model (PA-C10) having a greaterresponse than the model commonly used in rats (PA-C40), which, inturn, is greater than the large animal model (PA-D70).

Telemeter re-gelling

With no gel, the frequency response has the form of a classicunderdamped second-order system, as published previously,2 withresonance clearly visible as a peak in the response. This resonancecan lead to distortion of the recorded waveform if the pressure signalcontains harmonics high enough to excite the resonance. As gel isincrementally added, resonance is increasingly damped and the band-width is reduced (Fig. 4a). However, this dampening will reduce thechances of resonance and the corresponding distortion of the wave-form. Figure 4b shows that adding gel to the tip has a small effect onbandwidth when moderate amounts are used (up to�4 mm). Beyondthis quantity, bandwidth falls sharply and is only a few Hz when thetip is completely filled.

100 101 102 103

100 101 102 103

100 101 102 103

100 101 102 103

100 101 102 103

–60

–40

–20

0PA-C10

–60

–40

–20

0PA-C40

–60

–40

–20

0PA-D70

–60

–40

–20

0TR43P

–60

–40

–20

0

Frequency (Hz)

Mag

nitu

de (

dB)

Comparison

Fig. 2 Average frequency responses of the PA-C10, PA-C40 and PA-D70(Data Sciences International) and TR43P (Telemetry Research; all n = 4)telemeters with raw data from individual telemeters in grey. A comparison(bottom), reveals that their frequency response is only slightly inverselyproportional to the size of the device or catheter dimensions.

864 D McCormick et al.


Tip removal

Transfer functions for various amounts of removed tip are shown inFig. 5. This figure shows that removing the tip has negligible impacton dynamic performance unless the tip is almost completelyremoved, at which point a small drop in bandwidth is visible. Thereis a slight increase in bandwidth when small amounts (0–3 mm) ofthe tip are removed. This is consistent with the results shown inFig. 4 and is equivalent to a reduction in the length of the gel plug,which has been shown to increase bandwidth. These results suggestthat cutting a portion of the tip off to remove a stubborn clot will notcompromise the dynamic performance of the telemeter.

Simulated dynamic performance

The identified transfer functions for the PA-C40 and TR43P areplotted in Fig. 6, along with the average measured responses. Thefitted models accurately capture the real and imaginary compo-

nents of the measurements up to the point where the noise floordominates. At these frequencies, the magnitude of the response atthe injected frequency is reduced by several orders of magnitudeand is assumed not to contribute to measurements. The PA-C40was fitted with a zero-order numerator and fifth-order denominator,which gave a goodness of fit (see Eqn 3) of 96%. The resultingtransfer function was:

GðsÞ ¼ BðsÞAðsÞ

¼ e�0:007s4:406� 1010

s4þ 1237s3þ 7:924� 105s2þ 3:303� 108sþ 4:409� 1010

For the TR43P, the transfer function was estimated using a zero-order numerator and fifth-order denominator, which gave a goodnessof fit value of 86%.

100

10Frequency (Hz)

Quantity of gel (mm)

110

210

3–20

–15

–10

–5

0

5(a)

−3 dB

0 1 2 3 4 5 6 70

100

200

300

400

500(b)

Fig. 4 (a) Transfer function from catheter tip to pressure sensor output asgel is added to the distal tip. The recordings clearly show the dampeningeffect of adding gel to the distal tip for n = 4 catheters as described in Proto-col 2. 0 mm, no gel was used and the tip was filled with pressure transmis-sion fluid; 1–7 mm, 1–7 mm gel was applied, respectively, with theremainder of the tipped filled with fluid. (b) Mean (±SEM) –3 dB frequencyversus quantity of gel for n = 4 catheters as described in Protocol 2. Initiallythe gel has little effect on bandwidth, but rolls off steadily after the tip ishalf-full. The –3 dB frequency is still adequate for measurements with up to6 mm gel.

PA−C10 PA−C40 PA−D70 TR43P0

50

100

–3 d

B (

Hz)

150

200

Fig. 3 Comparison of the –3 dB responses of the individual telemeters whensorted by device type. The PA-C10, PA-C40 and PA-D70 are produced byData Sciences International, whereas the TR43P is produced by TelemetryResearch.

Table 1 Average frequency responses of telemetry devices and results ofnon-parametric analyses

Device type

PA-C10 PA-C40 PA-D70 TR43P

Catheter dimensions*(length · o.d.)

50 ·0.4 mm

80 ·0.7 mm

250 ·1.2 mm

80 ·0.7 mm

–3 dB (Hz) of Device 1 59 40 38 177–3 dB (Hz) of Device 2 59 39 27 180–3 dB (Hz) of Device 3 56 25 31 165–3 dB (Hz) of Device 4 52 55 32 170Mean (±SEM) –3 dB (Hz)

56.5 ± 1.7 39.8 ± 6.1 32.5 ± 2.3 173 ± 3

Non-parametric analysisMean rank 10.25 5.75 3.5 14.5

The PA-C10, PA-C40 and PA-D70 devices are produced by Data SciencesInternational, whereas the TR43P device is produced by Telemetry Research.

*Length of the catheter on the devices tested (other lengths are available).o.d., outer diameter.



GðsÞ ¼ BðsÞAðsÞ

¼ e�0:035s1:004� 1012

s4þ 1972s3þ 3:41� 106s2þ 2:677� 109sþ 1:004� 1012

The exponential terms in the numerators relate to the system delays,which were 7 and 35 ms for the PA-C40 and TR43P, respectively.As can be seen in Fig. 7, all telemeters accurately capture the

waveform recorded by the Millar transducer, aside from a slightreduction in pulse pressure and a phase shift due to the radio trans-mission of the acquired signals. The TR43P exhibits a longer timedelay than the PA-C40 (35 vs 7 ms) owing to the digital radio sys-tem, which catches data before transmission. These delays are fixedand therefore can be compensated for where required by subtractingan appropriate amount of time from the recordings. This delay wasmeasured by applying a burst of a 1 Hz sine wave to the telemeterusing the pressure generator and simultaneously recording the recon-structed pressure signal and the reference waveform from the MillarMikro-Tip transducer. The point when the pressure was first applied(beginning of the burst as measured by the Mikro-Tip sensor) wastaken as the reference time. The same point was then extracted fromthe reconstructed telemeter output. Subtracting these times gave thetransmission delay of the systems.The lower plot in Fig. 7 is a close up comparison of the waveforms

with the time stamps adjusted to remove the delay, which furtherillustrates the fidelity with which the devices are able to measurepulse pressure. The only discernable distortion is a slight reduction inpulse pressure. These were quantified using the Blood Pressure Mod-ule (ADInstruments) by calculating the average pressure and extract-ing the cyclic systolic and diastolic pressures, which were thenaveraged over the 15 s recording. For the PA-C40, the mean, systolic,diastolic and pulse pressures were 12.2 kPa (91.7 mmHg), 15.3 kPa(114.8 mmHg), 10.1 kPa (75.6 mmHg) and 5.2 kPa (39.2 mmHg),

respectively. For the TR43P, the mean, systolic, diastolic and pulsepressures were 12.2 kPa (91.8 mmHg), 15.5 kPa (116.0 mmHg),10.1 kPa (75.1 mmHg) and 5.5 kPa (40.9 mmHg), respectively.These compare very well with values extracted from the raw MillarProbe data of 12.2 kPa (91.8 mmHg), 15.5 kPa (116.4 mmHg),10.0 kPa (74.7 mmHg) and 5.6 kPa (41.7 mmHg), respectively.These reductions in pulse pressure are far less than variations causedby catheter tip placement in the artery;10 thus, in vitro testing hasdemonstrated that both telemetry devices possess sufficient band-width to record arterial pressure in the rat.

DISCUSSION

In the present study, we determined the frequency response ofimplantable telemetry devices commonly used to record blood pres-sure in physiological research. The frequency response and dynamicperformance of these devices have not been reported previously.Previous studies have tested the frequency response of indwelling

pressure transducer catheters for acute surgical use.11–13 However,with long lengths, compliant catheter materials and pressure sensors

100

10Frequency (Hz)

Frequency (Hz)

Pha

se (

°)M

agni

tude

(dB

)

110

210

3−40

−30

−20

−10

0

10(a)

(b)

100

101

102

103

−500

−400

−300

−200

−100

0

100

Fig. 6 Comparison of the measured and fitted transfer functions for theTR43P (Telemetry Research) and PA-C40 (Data Sciences International)devices showing magnitude response (a) and phase response (b). (s), TR43Pmeasured; (·), TR43P fitted; ( ), PA-C40 measured; ()), PA-C40 fitted.Transfer functions were fitted with an emphasis on the lower frequencies,which are most important for accurate simulations. Simulated and measuredresponses begin to diverge at higher frequencies, where the attenuation islarge.

100

101

102

103

–20

–18

–16

–14

–12

–10

–8

–6

–4

–2

2

0

Frequency (Hz)

Mag

nitu

de (

dB)

0 mm

1 mm

2 mm

3 mm

4 mm

5 mm

6 mm

7 mm

Fig. 5 Frequency response for one TR43P device (Telemetry Research) plot-ted as a function of how much of its tip has been removed. Cutting the tip offhas very little effect on the frequency response until the entire thin-walledportion of the catheter has been removed (7 mm).



and small diameters (relative to length) limiting frequency response,these systems have only evaluated frequencies to approximately100 Hz.2,14,15 Implantable manometers for use in small animal physi-ology possess catheters that have short lengths, comparatively largediameters and non-compliant catheter materials. For example, theDSI PA-C10 device for use in mice has a 50 mm long catheter. Thissuggests that the frequency response of these devices will be muchhigher than previously published pressure generators can evaluate.2

Thus, we developed a new frequency response test system with suffi-cient bandwidth to test blood pressure telemeters.The system was used to test the frequency response of commer-

cially available telemetry devices. The results show that there is a sig-nificant difference between the types tested. This raised the questionof whether the device frequency responses would lead to noticeabledifferences in recorded pressure dynamics. There have been manyprevious studies describing the frequency requirements for recordingarterial pressure.2,14–18 However, these are almost exclusively aimedat indwelling fluid-filled catheters, which possess bandwidths muchlower than the devices described here and typically display under-damped second-order transfer functions that are determined by themechanical properties of the catheter and fluid system. The measure-ments presented herein show much higher order phenomena due tosampling and filtering within the telemetry systems that have notbeen described in the literature. This makes it difficult to assess theperformance of implantable manometers using the usual mathemati-cal relationships.A theoretically simple means to understand the effect would be to

simultaneously insert catheters of the devices of interest into the exactsame position in the artery, such that the tips were exposed to identi-cal pressure waveforms; this would allow a direct comparison of therecordings. Unfortunately, it is very difficult to perform such a proce-dure in small animals with a high heart rate (where having a high-

frequency response is arguably the most important) owing to relativedimensions of the artery and the catheters.A practical means to understand these effects is to compare the

impact of the documented frequency responses of different telemetersystems to blood pressure data originally recorded with a transducerof much higher bandwidth. By simulating the responses of thedevices to a high-fidelity recording of arterial pressure, we were ableto clearly demonstrate that the dynamic performance of the PA-C40and TR43P are sufficient for the purposes of recording the dynamicsof arterial blood pressure. The Bode plots shown in Fig. 2 for thetelemeters manufactured by DSI are very similar in shape, with thedifferences between types largely determined by their respective –3 dB frequencies. Thus, the PA-C40 was chosen to represent thetypes manufactured by DSI because it could be compared withthe TR43P, which is intended for use in the same species. Because ofthe similarity of the DSI frequency response curves, similar conclu-sions about the ability of the PA-C10 and PA-D70 to record thedynamics of arterial blood pressure can be drawn based on their mea-sured bandwidths and the heart rate that can reasonably be expectedfor the species in which they are commonly used (mice, rabbits, pigs,dogs etc.).If a clot is present at the tip of the catheter after explantation, one

procedure to remove the clot and prepare the catheter for reuse is tocut off a small section of the tip of the catheter. This procedure savestime when removing difficult clots compared with using gel to washthe clot from the tip. However, its effect on frequency response isunknown. In order to investigate the effects of this procedure, mea-surements were made to determine what effect the length of the distaltip has on frequency response. The measurements showed that tiplength can be reduced without affecting bandwidth. Based on band-width alone, it seems reasonable to operate the telemeters with as lit-tle as 1 mm of distal tip remaining. However, this assessment doesnot test the washout effects or long-term consequences of having avery small gel barrier. One of the reasons for using gel at the end ofthe catheter is to seal the transmission fluid in the catheter, withoutintroducing a pressure offset that could eventuate in a sealed systemdue to temperature-dependent fluid expansion. The thin-walled sec-tion of the catheter has been suggested to contribute to the frequencyresponse.19 However, significant reasons for its use are to allow easyre-gelling and to increase the volume into which the catheter fluidcan expand. Therefore, using a telemeter with more than a fewmillimeters of tip removed is not recommended. However, when astubborn clot is encountered, it may be possible to remove a sectionof the tip (including the clot) to enable the telemeter to be reusedwithout refurbishment if one is willing to accept the increased likeli-hood of gel washout and blood entering the catheter.A telemeter may be implanted multiple times during the course of

its life. After explantation, the distal tip is checked for bubbles orclots, cleaned and re-gelled. Although the re-gelling operation maybe performed many times, there appears to be no information regard-ing the effect of this process on the telemeter’s ability to subsequentlyrecord arterial pressure dynamics. In order to investigate these affects,Protocol 3 measured the frequency response of a rat-sized catheter asthe amount of gel in the distal tip was varied. We observed that mod-erate amounts of gel (up to 3–4 mm) will not result in an observablereduction in the –3 dB frequency. Even with 6 mm gel, the band-width is still approximately 100 Hz; enough to capture the first 10harmonics of a 10 Hz (600 b.p.m.) mouse heart beat. Statistical anal-ysis of the results showed that bandwidth reduction due to gel

5.0 5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9 6.070

80

90

100

110

120(a)

(b)

mm

Hg

5.00 5.02 5.04 5.06 5.08 5.10 5.12 5.14 5.16 5.18 5.2070

80

90

100

110

120

Time (s)

mm

Hg

Fig. 7 High-fidelity original signal (solid line) passed through the model ofthe PA-C40 (dashed; Data Sciences International) filter response and passedthrough the model of the TR43P (dotted; Telemetry Research) filter response.(a) Original output from simulation including transmission delay. (b) Close upof aligned waveforms demonstrating that both devices capture the waveformwith good fidelity.



becomes significant when 5 and 6 mm gel is applied to the tip of theTR43P and PA-C40 devices, respectively. This suggests that the gel-ling process is relatively robust, unless the gel reaches the neck of thecatheter. If this happens, the telemeter should not be used because ofthe marked reduction in frequency response that occurs, presumablyas a consequence of the viscous gel entering the segment of the cathe-ter with a relatively narrow lumen. With this limitation in mind, there-gelling process appears to be robust with regard to frequencyresponse because a relatively large amount of gel would not lead to anoticeable reduction in the frequency content of the recordings.Although the PA-C10 and PA-D70 catheters have different dimen-sions to the catheters that the gel was applied to, it is expected thatthe responses of these devices will be similar to the TR43P and PA-C40 as far as the quantity of gel is concerned. Therefore, whenapproximately half the tip is filled with gel, it is expected that noobservable reduction in bandwidth will be observed for the PA-C10and PA-D70.We occasionally encountered a frequency response that was much

lower than other devices of the same type. The problem was alwaysfound to be a small, visually undetected air bubble in the gel or gel–fluid interface. After re-gelling, the response of the device under testwas restored to its expected value. The one exception to the normwas device no. 3 of the PA-C40 type in Table 1, where the responsecould not be returned to normal after careful inspection and re-gel-ling. Air bubbles in the distal tip would appear to be the greatest riskin using these gel-filled catheters during normal laboratory use. Reli-able re-gelling clearly requires the use of a dissecting microscope anda light source, where the direction the light applied can be manipu-lated independently of the microscope such that the tip can beinspected under a range of lighting directions.Although we could find no publications detailing the frequency

responses for the devices measured here, a frequency response for thePA-C20 produced by DSI has been reported previously to be145 Hz.19 This shows a much higher frequency response than themeasurements of the PA-C10 as given in the present study. The PA-C10 is smaller than the old PA-C20 and one of the consequences maybe a reduction in device bandwidth to conserve battery power. Millset al.19 recorded the responses relative to a Millar Mikro-Tip catheterreference, which has been shown to have a flat response to severalkHz, making it unlikely that the response of the pressure generatorused affected the results. However, no method is given by the authorsdetailing how the response was calculated and what dB scaling wasused (see Eqn 1 in the present study). When evaluating the responseof a potential function such as pressure, a multiplier of 20 is usuallyused. With this scaling, at –3 dB the power in the signal is reduced tohalf the reference value. When a factor of 10 is used in Eqn 1, aswould be the case if signal power was being recorded directly, the–3 dB frequency of the PA-C10 increases to approximately 100 Hz.This is equivalent to reading the –6 dB frequency off the responsesgiven here. Thus, the large differences in the reported responses forthe PA-C10 and PA-C20 may be a matter of convention.As part of Protocol 4, the respective transmission delays of the

TR43P and PA-C40 were measured. This raises the question of syn-chronization of the recordings with respect to other recordings. If therecordings are made by the same telemeter, they will already be syn-chronized. If the telemeterized pressure recordings are to be synchro-nized to external signals, reference signals for all the devices to besynchronized will be required such that the delay can be measured asis given in the Results section for Protocol 4. A simple means of

generating a reference pressure change is a pop test,2 which includesa reference transducer such as the Millar Mikro-Tip. However, inmost situations this will not be required because the delay is smallcompared with the physiological response to be investigated.The frequency responses of the PA-C40 and TR43P are lower than

the measurements taken directly from the catheter, as shown Fig. 4.These measurements were obtained using catheters that are veryclose in dimensions to the PA-C40 and TR43P. It seems likely thatthe bandwidth of the telemetry devices is being limited not by thefluid-filled catheters, but by the data transmission systems. TheTR43P is able to have a high bandwidth because it has a continuouspower delivery system20 and does not need to conserve energy toextend device operating time. It would appear that a consequence ofachieving long battery life for the PA-C40 (and other DSI telemeters)is that the telemetry system limits bandwidth to well below thepotential of its catheter system.The present study has shown that the in vitro responses of the tele-

meters evaluated are sufficient to capture arterial dynamics. It doesnot consider the effect of long-term implantation on the frequencyresponse, which could be compromised by encapsulation and clotformation. These effects were not investigated owing to the largenumber of variables to be accounted for, including implantation time,species and implantation method, as well as user inputs, such ashandling and sterilization methods.

CONCLUSIONS

The frequency responses of commercially available telemeters, whichuse a fluid-filled catheter, have been measured in vitro using a spe-cially developed test system. The wide bandwidth of the system givesconfidence to the accuracy of the measurement of the bandwidth ofthe telemeters. The system was used to determine the frequencyresponse of commonly available implantable manometers for use insmall animals. Measurement of the TR43P and PA-C40 transfer func-tions followed by simulation of their response to high bandwidthrecordings of arterial pressure has shown that their frequencyresponse is adequate to represent the arterial pressure waveform of arat. User practices of re-gelling have demonstrated that the frequencyresponses of these devices are robust with catheter gel re-fillingproviding that the gel–oil interface is within the catheter tip.

ACKNOWLEDGEMENT

This research was supported by a Tertiary Education CommissionTop Achiever PhD scholarship (to DMcC).

REFERENCES

1. Hales S. Statical Essays, Containing Haemastaticks. Hafner Publica-tions, New York, 1964.

2. Gabe I. Pressure measurement in experimental physiology. In: Bergel D(ed). Cardiovascular Fluid Dynamics. Academic Press, London, 1972;11–50.

3. Kurtz TW, Griffin KA, Bidani AK, Davisson RL, Hall JE. Recom-mendations for blood pressure measurement in humans and experi-mental animals: Part 2. Blood pressure measurement in experimentalanimals: A statement for professionals from the Subcommittee ofProfessional and Public Education of the American Heart Associa-tion Council on High Blood Pressure Research. Hypertension 2005;45: 299–310.



4. O’Neil PJ, Kaufman LN. Effects of indwelling arterial catheters or phys-ical restraint on food consumption and growth patterns of rats: Advanta-ges of noninvasive blood pressure measurements techniques. Lab. Anim.Sci. 1990; 40: 641–3.

5. Irvine RJ, White J, Chan R. The influence of restraint on blood pressurein the rat. J. Pharmacol. Toxicol. Methods 1997; 38: 157–62.

6. Brockway BP, Mills PA, Azar SH. A new method for continuouschronic measurement and recording of blood pressure, heart rate andactivity in the rat via radio-telemetry. Clin. Exp. Hypertens. 1991; 13:885–95.

7. Brockway BP, Mills PA, Miller JT, inventors. Data Sciences Inc.,assignee. Device for chronic measurement of internal body pressure. USpatent no. 4846191. 11 July 1989.

8. Foster HL, Small JD, Fox JG. The Mouse in Biomedical Research.Academic Press, New York, 1981.

9. Ljung L. System Identification: Theory for the User, 2nd edn. PrenticeHall, New Jersey, 1998.

10. Potkay J. Long term, implantable blood pressure monitoring systems.Biomed. Microdevices 2008; 10: 379–92.

11. Foreman JE, Hutchison KJ. Generation of sinusoidal fluid pressures ofrelatively high frequency. J. Appl. Physiol. 1970; 29: 511–16.

12. Ball G, Gabe I. Sinusoidal pressure generator for testing differentialmanometers.Med. Biol. Eng. Comput. 1963; 1: 237–41.

13. Levasseur JE, Funk FC, Patterson Jr JL. Square-wave liquid pressuregenerator for testing blood pressure transducers. J. Appl. Physiol. 1969;27: 426–30.

14. Yeomanson CW, Evans DH. The frequency response of external trans-ducer blood pressure measurement systems: A theoretical and experi-mental study. Clin. Phys. Physiol. Meas. 1983; 4: 435–49.

15. Schwid HA. Frequency response evaluation of radial artery catheter-manometer systems: Sinusoidal frequency analysis versus flush method.J. Clin. Monit. Comput. 1987; 4: 181–5.

16. Ernst W, Volker C, Helmut O, Eckart F. High fidelity correction ofpressure signals from fluid-filled systems by harmonic analysis. J. Clin.Monit. Comput. 1999; 15: 307–15.

17. Hok B. Dynamic calibration of manometer systems. Med. Biol. Eng.Comput. 1976; 14: 193–8.

18. Kleinman B, Powell S, Gardner R. Equivalence of fast flush and squarewave testing of blood pressure monitoring systems. J. Clin. Monit. Com-put. 1996; 12: 149–54.

19. Mills PA, Huetteman DA, Brockway BP et al. A new method formeasurement of blood pressure, heart rate, and activity in the mouse byradiotelemetry. J. Appl. Physiol. 2000; 88: 1537–44.

20. Budgett DM, Hu AP, Si P et al. Novel technology for the provision ofpower to implantable physiological devices. J. Appl. Physiol. 2007; 102:1658–63.

SUPPORTING INFORMATION

Additional Supporting Information may be found in the online version of this article:

Data S1 Technical Supplement (PDF) covering the design and performance of the frequency response measurement system.

Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting materials supplied by the authors. Anyqueries (other than missing material) should be directed to the corresponding author for the article.



Frequency response of implantable blood pressure telemetry systems

Documents

Transcript of Frequency response of implantable blood pressure telemetry systems