Blood pressure determination based on delay times between points on a heartbeat pulse

[0001] This application claims the benefit of provisional application 60/352,213 filed Jan. 29, 2002; 60/371,399 filed Apr. 11, 2002; 60/387,435 filed Jun. 11, 2002; and 60/410,349 filed Sep. 13, 2002, all of which are incorporated herein by reference.

[0002] This invention is concerned with the measurement of blood pressure (BP), and more particularly with non-occlusive, passive blood pressure measurements, using a sensor of heartbeat pulses at a single site, and with a resolution sufficient to resolve Pulsus Paradoxus (PP).

[0003] Previous attempts to measure blood pressure with resolution sufficient to resolve small changes in BP associated with PP have relied on pulse propagation delay times between brachial and radial artery sites, for example. In general, such two-site approaches have only been able to track substantial changes in BP using pulse transit time (PTT) but have failed to reliably resolve the small changes in BP associated with PP.

[0004] Two-site measurement approaches have been especially deficient in the measurement of systolic variations, because the heartbeat pressure pulse changes in shape and amplitude as it heads toward the arterial periphery. These changes are due to a number of factors, including changes in the arterial wall material composition that affect the wall's elastic behavior, the taper of the main arterial branch, the distribution of branch lines, and pulse reflections. The result is that the pulse steepens and contracts as it propagates, as linear pulse propagation models predict. Additionally, non-linear effects in the arterial tree can produce pulse steepening. More importantly, the non-linear elastic response of the arterial wall results in a distinctly non-linear relationship of pulse propagation velocity and pressure.

[0005] An object of the present invention is to avoid problems and disadvantages of multiple-site blood pressure measurements and to provide single-site measurement of blood pressure with less complexity and lower cost than has heretofore been possible.

[0006] More particularly, the present invention avoids the problems due to different pressure-induced pulse-shape modulations associated with different pulse detection sites, by detection of single heartbeat pulses at a single site and by analysis of individual pulses. In its preferred form, the present invention makes use of the fact that changes in time delay between different parts of a heartbeat pulse, subjected to different arterial pressures, reflect changes in blood pressure.

[0007] More particularly, it has been discovered that the well known pressure-velocity relationship that has been shown to hold for pressure-change induced pulse propagation changes also holds for the components of a single pulse, and are resolvable, since the pulse time evolution is comparatively slow. While the pressure pulse traverses the distance between brachial and radial pulse sites in time scales on the order of 40 to 50 milliseconds, the rise time of the pressure pulse, its “fastest” feature, is on the order of 100 milliseconds. As the pulse steepens and its components contract in time, as the pressure rises, time delay measurements on the rising edge of the pulse can be used to track pulse pressure changes. By measuring delay times between certain percentages of a given pulse's full amplitude (peak), automatic normalization is achieved, allowing comparison of all pulses, largely independent of coupling efficiency, which is the bane of prior pressure measurements, and which requires frequent occlusive re-calibration.

[0026] FIG. 1 illustrates an apparatus for implementing the invention, including a sensor 10 linked to a computer 12 (or other appropriate device) having a display 14. The sensor detects heartbeat pulses at a single site (such as the wrist) on a body through which blood flows. Analysis performed by the computer determines delay times between predetermined points on each heartbeat pulse, and the delay times are used by the computer to obtain a measurement of blood pressure. Various outputs from the computer (later described) can be viewed on the display and recorded.

[0027] In a preferred embodiment, systolic blood pressure is determined by measuring delay times between a reference point on the onset slope of the heartbeat pulse, located, for example, at 50% of the full amplitude (peak) of the heartbeat pulse, and points on the onset slope that are located at predetermined percentages of the pulse peak. FIG. 2 shows the profile of a heartbeat pulse and shows percentage points on the onset slope between the pulse onset and the pulse peak.

[0028] The present invention may use any appropriate sensor for detecting heartbeat pulses at a single site. For example, the sensor may be of the fiber-optic type disclosed in U.S. Pat. No. 6,463,187 granted Oct. 8, 2002, which responds to radial artery displacements at the wrist. In one embodiment of the invention, a completely integrated wireless sensor unit is employed featuring a fiber-optic sensor, 16-bit analog digital conversion at 512 Hz, amplification, a digital transmission system operating at 916.5 Mhz with a raw data rate of 250 Kbps using Manchester encoding, as well as an on-board Microchip PIC16C67 (OTP Version) operating at 20 MHz. This unit communicates with a PC computer via an antenna that is plugged into a serial port. It is apparent that the invention is not limited to such a sensor unit.

[0029] FIGS. 3A and 3B show a single-pulse BP determination flow chart in accordance with an embodiment of the invention. This embodiment is merely an example of the manner in which the invention can be implemented and is not intended to limit the invention.

[0030] Referring to FIG. 3A, a heartbeat pulse data stream is acquired for a given period of time, such as four to eight seconds, from a sensor (step 1). See FIG. 4. The acquired data stream is filtered by a fast Fourier transform filter, removing DC offsets and centering the data stream about zero (step 2). See FIG. 5. Peak/pulse detection of the filtered data is performed, in which the zero crossing of the onset slope of each heartbeat pulse is detected (step 3). See FIGS. 6 and 7. The onset slope of each heartbeat pulse is then fitted using a ninth order polynomial (step 4). See FIG. 8. While other functional forms can be used, the functional form of an arterial pulse propagation model would be preferred. The fitted function is used to determine the pulse onset and peak, and the time positions of a sequence of percentages of the full pulse amplitude (step 5). See FIG. 2. Steps 6 and 7 illustrate in detail the manner in which step 5 is performed. As shown in FIG. 3B, an effective delay time is then calculated (step 8) and is used to determine blood pressure (step 9). See FIG. 9.

[0031] The equation in step 9 has two parameters that are adjusted to fit the data, one being the exponential decay constant ξ, and the other a constant pressure offset C. In order to make the conversion from delay time to pressure using a velocity relationship, a distance has to be invoked, as indicated in the pressure expression. The distance, taken to be 1 meter, is not part of the fitting parameters, but simply a constant. Due to the nature of the logarithmic function, adjusting the distance accomplishes the same as adjusting the constant C; it provides a constant offset for the pressure. The scaling of the delay time to pressure is influenced solely by the choice of the exponential parameter. The values for the other constants are obtained from the literature. Specifically, E=0.94 10⁶ dyne/cm², h=0.08 cm, and α=0.18 cm.

[0032] FIG. 9 conveys a sense of the degree to which Pulsus Paradoxus (PP) can be resolved with the single-point delay-time method of the invention. FIG. 9 presents the PP modulations, converted into units of pressure from the measured delay times, as well as the interbeat interval, derived from the same data stream. The peak-to-peak amplitude of the PP modulations is 7.64 mmHg (error=0.49) while the corresponding measurement with a clinical blood pressure monitor (Colin Pilot) yielded 8.08 mmHg (error=1.17).

[0033] In the embodiment just described, delay times are obtained between points on the onset slope (front-end) of the heartbeat pulse, and more particularly, systolic blood pressure is determined from such delay times.

[0034] FIG. 10 is a single-pulse BP determination flowchart for measurement of both systolic and diastolic blood pressure by calculation of delay times in both systolic and diastolic regimes on the back-end of the heartbeat pulse. In this embodiment, the acquisition of a data stream, the filtering of data, and the max/min determination can be performed as in the first embodiment. Then time positions of predetermined points on the back end of the pulse are established (step 10). See FIG. 11. Data points between predetermined percentiles on the back side of the heartbeat pulse are linearly fitted, and based on these fits, a revised time determination of a predetermined number (e.g., 4) of percentiles is computed (Step 11). See FIGS. 12 and 13. Corresponding pressures are then calculated using the conversion expression used in the first flowchart (step 12). See FIG. 14. In FIG. 14, the delay time pressure obtained in accordance with the invention is the result of converting a sequence of delay time measurements between 55% and 85% of the full height of a given pulse on the falling side (back-end) of the pulse. These measurements were obtained at the wrist of a subject.

[0035] A data acquisition rate of 512 Hz, corresponding to a resolution of about 2 milliseconds, is sufficient to resolve delay time changes that are on the order of tens of milliseconds on the front-end (onset slope) of a heartbeat pulse and more by a factor of 4 to 5 on the back-end. The spectral content of the heartbeat pulse in the arterial periphery extends maximally to about 30 Hz, with the significant amplitude-carrying harmonics being in the range of up to 15–20. The heartbeat pressure is therefore bandwidth-limited, that is, except for noise interference, no spectral surprises lurk in higher frequency bands that could produce aliasing. Bearing in mind that a signal is over-sampled if the acquisition rate exceeds approximately three times the bandwidth limit, clearly then, 512 Hz over-samples the heartbeat pulse signal. With regard to establishing delay time changes that exceed the acquisition rate, it is possible to further “enhance” the acquisition rate by interpolating between measured data points during analysis.

[0036] As described earlier, one of the attributes of the invention is resolution sufficient to resolve Pulsus Paradoxus (PP). FIG. 15 shows evolution of averaged delay time for a series of percentiles on the pulse onset slope. FIG. 16 shows the results of converting the average delay time to pressure using the equation of step 9 in FIG. 3B, with an exponential factor ξ=0.039 and an offset of 30. The delay time pressure gives a peak-to-peak Pulsus of 6.37 mmHg with an error of 0.44 mmHg.

[0037] An interesting observation from FIG. 13 is that the artery's non-linear time-pressure relationship can be discerned. Visual comparison of the delay time modulations, presumably due to Pulsus during the first 30 seconds and the modulations during the 40–100 second period, reveals a significant difference in amplitude. This difference is essentially eliminated as the result of the non-linear conversion seen in FIG. 14. The specific values chosen to obtain the pressure conversion are ξ=0.027 with an offset of 72 mmHg.

[0038] With a correlation of the general BP curves established, it is interesting to compare the ability to determine PP. In order to isolate the PP modulations for comparison, both BP traces presented in FIG. 16 were Fourier filtered between 0.1 and 0.4 Hz. The results are presented in FIG. 17. In order to obtain a quantitative comparison of the magnitude of the modulations, and since the depth of the modulations did not exhibit a sustained increase or decrease, the peak-to-peak mean of all the positive and negative peak amplitudes was determined. In the case of the Colin unit, the average PP is 5.28 mmHg, with an error of 0.34 mmHg. For the invention, the corresponding values are 4.61 mmHg, with an error of 0.5 mmHg.

[0039] It is apparent that the single-site pulse delay-time method of the invention makes feasible the tracking of blood pressure changes at a resolution sufficient to determine Pulsus.

[0040] Annexed hereto are Appendices I and II showing examples of a single-point delay-time algorithm and single-point delay-time functions for implementing the invention pursuant to the above-described embodiments. It is apparent that other algorithms and functions can be used to implement the invention.

[0041] While preferred embodiments of the invention have been shown and described, it will be apparent to those skilled in the art that these embodiments are merely examples of the invention, and that various modifications can be made without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims.