A Blood Pressure Sensor for Long-Term Implantation*Edward Bullister, *Sanford Reich, *Peter d’Entremont, †Neil Silverman, and

*James Sluetz

*APEX Medical, Inc.; and †EDG, Inc., East Walpole, Massachusetts, U.S.A.

Abstract: An implantable flow-through blood pressuresensor prototype has been developed for use with an im-plantable left ventricular assist device (LVAD). This sen-sor incorporates a flat pressure-sensing diaphragm that isdesigned to be integral with the wall of a titanium tube thatmay be placed in the inlet or the outlet flow path of anyLVAD. The interior tube flow geometry is transitionedfrom a round to a D-shape such that flow separation iseliminated. Bench testing of 3 sensors was performed tocharacterize the sensor. The worst-case results showed amaximum nonlinearity of 0.64 mm Hg, a maximum hys-teresis of 0.87 mm Hg, and a maximum nonrepeatability of0.87 mm Hg. Long-term drift studies of 2 sensors at 193days and 112 days resulted in a projected annual drift rate

of 1.4 and 2.0 mm Hg, respectively. The APEX pressuresensors were evaluated in 5 ventricular assist acute calfexperiments in which the sensor outputs were comparedwith Millar pressure catheter sensors. Pressure outputcomparisons showed similar pressure tracings. No visibleevidence of thrombus formation was found on the APEXsensor compared with thrombus formation found on theMillar catheter at the entrance to the flow path. Testsdemonstrated that the blood pressure sensor can accu-rately measure blood pressure and indicate that it haslong-term stability. Key Words: Sensor—Blood pres-sure—Left ventricular assist device—Rotary pump—Physiologic control.

Recently, there have been significant advances inthe development of heart assist pumps. Second andthird generation rotary pumps are smaller andlighter, and have demonstrated viability for long-term support. These pumps are being designed to betotally implantable and to enable patients to even-tually leave the hospital and resume normal activi-ties.

These new generation rotary pumps cannot relyon the simple passive fill mechanism that gave aphysiological control input to the first generation ofpulsatile heart pumps.

APEX has developed a pressure sensor for physi-ologic control of heart assist devices. The direct pres-sure measurements are intended to provide a defini-tive input for a physiologic controller to preventexcessive negative pressures in the ventricle, to avoidretrograde flows in the pump, and to autoregulatethe pump impeller speed for patient exercise leveland perfusion needs.

The direct pressure measurements also are in-

tended to provide a definitive input for diagnosticmonitoring of the pump and the heart performances.The sensors also can detect and monitor abnormalleft ventricular assist device (LVAD) performance,abnormal ventricular function, and trends in ven-tricular function and LVAD performance.

MATERIALS AND METHODS

Sensor designThe prototype sensors used in these experiments

were developed and manufactured by APEX Medi-cal, Inc. (East Walpole, MA, U.S.A.). The prototypesensor is referred to as the APEX pressure sensor(APS). The U.S. Patent and Trademark Office hasissued patent coverage (US 6, 171, 253 B1)for theflat tube pressure sensor, counterpart foreign appli-cations are being pursued, and additional patent ap-plications relating to this technology are pending.

The APS uses a pressure-sensing diaphragm thatis integrally built into the wall of a titanium tube.The tube can be made with an inside diameter thesame size as any LVAD inlet cannula. Mounted onthe pressure-sensing diaphragm is a molecularlybonded, thin-film strain gauge, whose strain measur-ing elements are located at the points of highest ten-sile and compressive strain. These elements measurethe strain in the diaphragm as it deflects from theblood pressure inside the tube. The output of the

Received January 2001.Presented in part at the 8th Congress of the International So-

ciety for Rotary Blood Pumps, held September 6–9, 2000, inAachen, Germany.

Address correspondence and reprint requests to Dr. EdwardBullister, APEX Medical, 141 Washington Street, East Walpole,MA 02032.

Artificial Organs25(5):376–379, Blackwell Science, Inc.© 2001 International Society for Artificial Organs

376

strain measuring elements is converted to an outputvoltage signal by standard strain gauge electronics.

The size and thickness of the pressure-sensing dia-phragm were selected to keep the diaphragm in es-sentially a pure bending mode over its entire rangeof operation and to provide a stiffness that is manyorders of magnitude greater than any plaque or tis-sue build-up that might occur over the life of thesensor. This design creates a very linear sensor witha sensitivity between 0.50 and 1.00 mV/V/mm Hg.This results in voltages that can be read with stan-dard strain gauge electronics.

Figure 1 shows the geometry and strain level ofthe pressure-sensing diaphragm under approxi-mately 300 mm Hg blood pressure in a view cut by its2 planes of symmetry. This finite element model wasgenerated using the ANSYS finite element analysispackage (ANSYS, Inc., Canonsburg, PA, U.S.A.).The small outward deflection of the diaphragm atthis blood pressure is exaggerated for visibility. Thestrain map on the flat diaphragm shown in Fig. 1represents the surface strain (the legend indicatesstrain levels in mm/mm). The highest strains are onthe thin, nominally flat, pressure-sensing diaphragm,which is built into the tube wall. The strain plot ofthe outside surface of the diaphragm indicates thatthe applied pressure creates a peak compressivestrain at the diaphragm edge and a peak tensilestrain at the diaphragm center. Based on the resultsof this finite element analysis, and a supportiveanalysis of a square diaphragm with built-in edges,using the principle of minimum strain energy de-tailed in Timoshenko (1), the optimum locations ofthe strain gauges are at 2 outside edges and the cen-ter of the diaphragm. The strain sensing elements ofthe APS are at these locations.

The flow path of the APS design is shown in thecutaway views of Figs. 2 and 3. The side view of Fig.2 shows the gradual transition along the flow pathbetween the fully circular tube and flat chord sec-tion. The end view of Fig. 3 shows how the flat dia-phragm forms a chord with the circular tube. Thissmooth transition is designed to eliminate flow sepa-ration. Flow separation results in flow stagnation atthe downstream reattachment point. In the region offlow stagnation at the reattachment point, blood isexposed to conditions of both low shear and longresidence times. This flow separation condition hasbeen shown to result in thrombus formation (2). TheAPS design avoids this mechanism of thrombus for-mation.

Because the diaphragm is integral with the wall ofthe tube, the APS presents no new material surfacesto the blood and introduces no new seams or jointsthat could be potential sites for initiation of throm-bus formation.

Test methods: Bench tests

Accuracy characterization testsEach prototype APS was characterized with a pre-

cision pressure calibrator. Compressed dry nitrogenwas used to pressurize the sensors from 325 to 1,100mm Hg. This range of gauge pressures was used tosimulate an absolute pressure sensor at various at-mospheric pressure conditions. The APS outputswere recorded in a computer using LabVIEW soft-ware (National Instruments, Inc., Austin, TX,U.S.A.). The APSs were kept at a constant tempera-ture of 37°C by a Model 5DM Incubator (Thelco,Winchester, VA, U.S.A.). Temperatures were re-corded by a 5,000 V Thermistor (Model 44034,Omega Engineering, Inc., Stamford, CT, U.S.A.).The APSs were exposed to this temperature-controlled environment for a minimum of 3 h priorto the testing in order to ensure that they hadreached thermal equilibrium.

FIG. 1. The geometry, strain, and deflection of the pressurizedAPS is shown. FIG. 2. A cutaway side view of the APS is depicted.

IMPLANTABLE BLOOD PRESSURE SENSOR 377

Artif Organs, Vol. 25, No. 5, 2001

A Ruska 7215i pressure calibrator with a range of50 psig (Ruska, Inc., Houston, TX, U.S.A.) was usedto pressurize tube sensors with a precision of ±0.15mm Hg. An HBM Model MVD2555 measurementamplifier (HBM Amplifier, Hottinger BaldwinMesstechnik, Darmstadt, Germany) amplified theoutput of the strain gauge bridge on each sensor andsent it to the LabVIEW computer over an RS232interface. The strain gauge signals and applied pres-sures were collected by an automatic data acquisi-tion system using PC-based LabVIEW (National In-struments, Inc.). A LabVIEW device driver waswritten for the Ruska calibrator to enable fully au-tomated control of the pressurization as well as au-tomated data collection and analysis. With this au-tomatic control, the experiments could be repeatedcontinuously to measure repeatability as well aslong-term drift.

For each test, the HBM amplifier was pro-grammed for a 2.5 V excitation, a 4 mV/V outputrange, a 0.2 Hz Bessel output filter, and the trans-ducer type in the full bridge mode. The pressure wasapplied in 25 mm Hg increments over the 325 to1,100 mm Hg range and then decremented from1,100 to 325 mm Hg in 25 mm Hg decrements.Within these pressurization and depressurizationcycles, the pressure was maintained at each incre-mental pressure level for a minimum of 30 s beforethe readout data were recorded. The data then wereused to compute the hysteresis, linearity, and accu-racy at each applied pressure. After each sensorcharacterization, a summary data report was auto-matically printed, and the raw data were recorded intab-delimited data files.

Longer term drift testsA series of longer term tests were used to measure

drift. In these tests, the pressurization and depres-surization cycles were run continuously. The dura-tion and levels of the incremental pressures were

modified as described below to give the maximumaccuracy for measurement of small amounts of drift.

Sensors 4 and 30 were continually cycled on theRuska precision calibrator through the following lev-els of pressures (in mm Hg): 0, 700, 900, 1,100, 900,and 700. Each of the 6 pressure levels in the se-quence was maintained for 0.5 h for a total cycle timeof 3 h. At the end of each cycle, a new cycle repeatedthe sequence starting with the first pressure level.

Test methods: Calf testsTo date, prototypes of the sensors have been used

in 5 calf experiments in which the pressure sensorswere used in conjunction with a pulsatile LVAD.The APS sensors were connected inline with theLVAD inlet and outlet cannulas along with an inlineconnector that permitted insertion of a catheterpressure sensor. The catheter pressure sensor (MPC-500 MikroTip, Millar Instruments, Houston, TX,U.S.A.) was inserted into the inline connector so thatthe Millar pressure sensor tip was placed adjacent tothe APS inline pressure sensor. The APS sensorswere zero balanced and calibrated at 37°C prior tothe calf study. Custom designed amplifiers (with abandpass of 60 Hz) were used for the calf studies.

RESULTS

Bench testsThe measured sensitivities for the diaphragm ge-

ometries tested were in approximate agreement withthe corresponding finite element predictions. Theexperimental accuracy results are shown in Table 1.

Longer term drift studies have been performedand are ongoing, and current results suggest that theoutputs appear to be stable. The target range for long-term drift is less than 3 mm Hg per year (Table 2).

At the time of this writing, Sensor 4 has beentested for drift under these conditions for 193 days

TABLE 1. Summary of accuracy test results

Specification(mm Hg)

Sensor002

Sensor006

Sensor007

Nonlinearity (max) <1.6 0.59 0.64 0.40Hysteresis (max) <0.4 0.87 0.86 0.64Nonrepeatability (max) <0.4 0.45 0.87 0.49Accuracy <3.0 1.91 2.34 1.52

TABLE 2. Summary of drift test results

Specification(mm Hg) Sensor 004 Sensor 030

1 year drift <3.0a 1.4b 2.0c

a <0.5% of full scale (325–1,100 mm Hg).b Drift data extrapolated from 193 days of life testing.c Drift data extrapolated from 100 days of life testing.

FIG. 3. A cutaway end view of APS is depicted.

E. BULLISTER ET AL.378

Artif Organs, Vol. 25, No. 5, 2001

(0.53 years). The slope of the best-fit straight line of theoutput drift is 1.4 mm Hg per year. Sensor 30 has beentested for 112 days (0.3 years). The slope of the best-fitstraight line of the output is 2.0 mm Hg per year.

Acute calf testsA goal of the acute calf tests was to demonstrate

that the APS generated blood pressure signal is com-parable to the Millar generated blood pressure sig-nal. Figure 4 shows a comparison between the APSand the Millar MikroTip catheter based blood pres-sure sensor mounted at the inlet of a pulsatileLVAD. The results show excellent agreement be-tween the 2 sensors.

Another goal of the calf tests was to determine ifany gross visible thrombus formed on the pressure-sensing surface that was in contact with the blood.After the tests, the titanium blood contacting sur-faces of the sensor were examined for evidence of

thrombus formation. No visible thrombus formationwas found on the tube sensing surfaces (Fig. 5).

DISCUSSION

The bench testing showed that the accuracy of theAPS is in the range of accuracy of the target require-ments for an implantable blood pressure sensor. Ex-trapolation of the drift data to 1 year predicts valuesthat are close to meeting the requirements. The va-lidity of this extrapolation is supported by our expe-riences that the drift rate in these sensors decreaseswith time. We are continuing with further testing toincrease the amount of data on these sensors.

CONCLUSIONS

The APS was able to track physiological bloodpressure signals in calf tests without apparent grossthrombus formation. Bench tests showed that theAPS can measure blood pressure in an acute, short-range time frame with an accuracy of approximately2 to 3 mm Hg. Extrapolation of initial longer termresults suggests that long-term drift will be within thetarget range of 3.0 mm Hg per year.

Acknowledgment: Partial financial support for this re-search and development effort was provided through theU.S. Department of Commerce’s National Institute ofStandards and Technology Advanced Technology Programunder Cooperative Agreement No. 70NANB7H3059.

REFERENCES1. Timoshenko S. Theory of Plates and Shells. New York: Mc-

Graw-Hill, 1959, reissued 1987.2. Friedrich P, Reininger A. Occlusive thrombosis formation on

indwelling catheters: In vitro investigation and computationalanalysis. Thromb Haemost 1995;73:66–72.

FIG. 4. Shown is a comparison of theAPS signal (top) with the Millar MikroTipsignal (bottom) .

FIG. 5. The photograph shows a posttest inspection of the APS.

IMPLANTABLE BLOOD PRESSURE SENSOR 379

Artif Organs, Vol. 25, No. 5, 2001