Bloodpressure measurement, improvement of oscillometric method. Øystein Stene, Lars Jørgen Kristoffersen, Waldemar Sulkowski Høgskolen i Narvik, Sivilingeniørutdanningen, postboks 385, N-8501 Narvik Norway, Tlf. 76 92 21 90, Fax. 76 94 48 66, E-mail: wald@gaupe.hin.no ABSTRACT Article describes practical problems and solutions regarding bloodpressure measurements with the oscillometric method. This improved measurement method make use of a additional finger pulse sensor to determine puls frequency in advance of the measurement. Increased precision in the detection of the bloodpressure levels and better comfort for the patient are achieved. Unwanted pressure fluctations are suppressed by regulation of the cuff pressure (advanced identification of the cuff characteristic is applied) and further by individually tuned FIR filter. This allows for automatic measuring of bloodpressure even if the pulse amplitude is weak. Simulations and practical testresults are included. When the pressure in the cuff falls and reaches the systolic (upper) bloodpressure, small pressure variations appear in the cuff as the blood pulses through the partially occluded vein (figure 1). These pressure oscillations increase in amplitude with maximum at the mean value of the bloodpressure level(figure 2). The diastolic (lower) pressure is found in three different ways: 1) calculated from the systolicand mean pressure with some uncertainity [Ref.1], 2) by correlation to a certain level of the pressure envelope or 3) by correlation to a certain slope of the pressure envelope [ref. 3]. There is no definite answer to the question which method to use and how to use it, but the oscillometric principle is applied in a large number of commercial automatic bloodpressure meters. 1. THE OSCILLOMETRIC METHOD 1.1 Basic principle The oscillometric method is an indirect way of measuring the blood pressure. That means the bloodpressure is not measured in the vain, but by sensing the airpressure in a pressurized cuff usually wrapped around the upper arm. Figure 2. Pressure oscillation caused by the pulse during cuff deflation. 1.2 Problems experienced by users. Commercial available apparatus have revealed some problems: Figure 1. Cuff pressure and finger pulse after A/Dconversion (pulse is not related to mmhg). 1. Ending or repeating the measuring sequence without results. Probably because of noise, weak pulsation or insufficient cuff pressure.
2. Incorrect results. Probably causes: too fast deflation of the cuff, the persons condition is near the limit of the spesification for the equipment, noise. 3. The measuring sequence uses longer time than the manual measurement method. This is a general problem of automatic blood pressure meters (and unpleasant for the patient also). It gets worse if the initial inflation is insufficient, because measuring sequence must repeat. Naturally, if the inflation pressure is far higher than the systolic pressure, the measuring time gets longer. 2. WAYS OF IMPROVEMENT 2.1 Introduction To make the measurement adaptiv to individual conditions, the information from the finger pulseprobe before the measurement sequence (deflation) is used. The benefits are as follows: 1. The cuff is inflated to a suitable individual level above the systolic pressure. 2. The rate of cuff deflation can be adjusted to get the desired precision without prolonging the measuring sequence unnecessary. 3. The pressure oscillations in the cuff can be individually filtered for better rejection of noise and disturbances from arm movements and breath. 4. The bandwidth of the pressure regulator can be set to supress pressure fluctations up to the individual pulsefrequency. This is possible because the cuff characteristic is identified during inflation. 2.2 Pneumatic adaptation When the cuff is inflated to a pressure above the systolic, the finger pulse disappear because the vain is occluded. This indicate the value of the systolic pressure. By permitting the pump to increase the cuff pressure somewhat more, the initial values of the pressure envelope are included (figure 2). Setting the maximum cuff pressure in this way, avoids repeating of the measuring cycle and prolonged measuring time. 2.3 Cuff pressure regulation If the cuff pressure decrease too rapid during the measuring phase, only a few oscillations occur and the envelope gets coarse. This means less precise detection of the bloodpressure. On the other hand, if the pressure decrease too slowly, the measuring time increase together with the discomfort. The solution is to have a constant number of pulsebeat in the measuring phase by setting the deflation slope based on the pulse frequency. This results in short measuring time for persons with high pulse, and long measuring time when it is necessary by low pulse. Because the pulse frequency is known in advance from the finger probe, the deflation rate can be set correct from the beginning of the measurement. The regulation of the pressure involves an unknown component, the cuff. Because various cuffsizes are used, arms are different and tightness of the cuff when attached varies, the cuff characteristic changes from one measurement to another. Figure 3 shows the relationship between pressure and volume for some sizes. Figure 3. Cuff pressure as a function of volume. In the feedback loop for the pressure, flow is the input and pressure is the output for the cuff. Flow is the derivative of the volume with respect to time. Because the aim is linear decreasing pressure, the pressure axis can be considered as a time axis. The slope of the curves then shows that the flow out of the cuff must increase as the pressure decrease. Further that the gain of cuffs are different. The gain of the feedback loop must be set correct to avoid false oscillations and damping of the true oscillations from the pulse, and can occur if the bandwidth of the feedback loop is extended into the pulse frequency range. To find the optimum gain for the loop, the pressure during inflation of the cuff is recorded. Afterwards the cuff characteristic is identified and the gain for the feedback loop is calculated. This technique is able to supress unwanted pressure fluctations up to the pulse frequency but assumes that the specifications for the pump and valve are known. During deflation, cuff pressure is used as a parameter for regulator gain correction (see figure 7).
2.4 FIR filtering Digital filtering has been chosen because of : collectiv tuning by altering the sampling frequency, easy characteristic modification by altering coefficients, no phase distortion occur. Because the sampling frequency is in the range 5-60 Hz, no expensive hardware is required in implementation. The FIR filters are used to remove trends, baseline disturbances, overharmonics and noise to get clean pressure- and pulse signals for further analyses. To detect the period and amplitude correctly, the mean value of the signal should be close to zero. This demands for high rejection ratio for the frequency response of the filters towards zero (DC). Using standard FIR filter coefficient calculation ( i.e. fir1 in Matlab with Hamming window method) gives very long filter length for good rejection of lower frequency (N=250, H =-57dB). Because the mean value of the filter coefficients differs slightly from zero, the frequency response flattens at a finite level. To cancel DC components, coefficients adjustment is necesarry. This can be done by subtracting the mean value of all the coefficients from every coefficient. - The pressure range is 0-300 mmhg. - Normal cuff pressure amplitude is 2 mmhg p-p. - A weak pulse is assumed to be 10% of normal amplitude. - Minimum resolution is 10 steps. Then there is 0.02 mmhg/step, and required A/D converter resolution is log 2 (300/0.02)=13,9. If an analog HP filter with gain is used, and the AC and DC component of the pressure is sampled separately, the resolution can be reduced. 3. RESULTS 3.1 Introduction Parts of the improvements have been tested off line in Matlab with a oscillometric tester as reference. The pressure regulation have been simulated in Simulink, tested and implemented on a Siemens SAB80C167 microcontroller which was intended to run all bloodpressure measurement and control algorithm (shown in figure 6) communicating with a 16 bits A/D and pneumatic system. A plot of the pressure regulated by the controller is shown in figure 5. Figure 4. FIR BP-filter before and after adjustment of coefficients. Thereby the mean value of the coefficients gets zero. After applying this modification, filter lengths between 20 and 40 are usefull. The difference is obvious in figure 4. The price for this modification, is some ripple in the passband (0.4dB) and less rejection in the higher stopband (10 db worse). 2.5 Resolution requirements. The use of pure digital filtration of the pressure signal make it necessary with a 14 bits A/D converter according to the following consideration: Figure 5. Cuff pressure regulation. 3.2 Measured results on oscillometric tester. 29 tests were run on a Metron QA1280 oscillometric tester which generate artificial oscillations. Mean deviation Outside ±3mmHg Sys +0.72 mmhg 5 Mean +1.8 7 Dia -1.72 9 Bloodpressure with cuffpressure oscillations as low as 5% of normal amplitude, was measured.
Figure 6. Overview of the hardware.. 3.3 Pressure regulator simulation. The Simulink model for the pressure regulator is shown in figure 7, and the results from simulation in figure 8 and 9. In both cases, a pressure pulse is applied in the cuff to test the step response. In figure 7, a 0.2 Hz sinus is included to simulate arm movement etc. The feedback loop reject it. In figure 8, a 1 Hz sinus is included to simulate the pulse at 60 bpm and it is not rejected. Figure 7. Simulink model for simulation of the pressure regulation.
Figure 8. Simulation of the pressure regulation with 0.2Hz disturbance. 4. CONCLUSION Benefits by using a additional pulse probe in connection with the oscillometric method are: The maximum pressure in the cuff is related to the actual bloodpressure level to minimize measuring time and discomfort.. A certain number of pulsebeats are included in the measurement, independent of the pulse frequency, to make sure that sufficient data are availiable for detection of the bloodpressure. By using extensive signal processing for identification of the nonlinear cuff characteristic, the pressure regulation adapts to different cuffs (pressure regulator gain correction) and pulse frequency. The ramp pressure reference simplifies trend elemination. The bloodpressure is individually filtered to increase the rejection of noise and disturbances. 5. REFERENCES 1. Jensen Ø., Grimnes S. (1987). Elektromedisinsk Utstyr, chap. 4. 2. Ifeachor E.C., Jervis B.W. (1993). Digital Signal Processing, chap. 12.5.2. 3. Metron a/s: Users manual QA1280, oscillometric bloodpressure monitor tester. Figure 9. Simulation of the pressure regulation with 1Hz disturbance.