The present invention relates to a method of approximating a patient's pulse wave based on non-invasive blood pressure measurement. The invention also relates to a logic unit and a corresponding system for approximating a patient's pulse wave based on non-invasive blood pressure measurement.
A skilled practitioner, such as an experienced physician, can obtain useful information as to the health status of a patient from an analysis of the curve progression of the arterial blood pressure, i.e. pulse wave, of the patient. The pulse wave of the patient may be reliably measured in an invasive way, by introducing a catheter into one of the patient's blood vessels. However, invasive blood pressure measurement approaches are relatively complex, sometimes being accompanied by adverse side-effects for the patient, such as thrombo-embolic complications, bleedings and infections.
A well-known, less dangerous and more convenient way to determine the arterial blood pressure values of a patient is to use the so-called “oscillometric non-invasive blood pressure measurement method”. By that method, a pressure cuff is applied to one of the patient's extremities, preferably to his upper arm at the level of his heart, as schematically illustrated in
The distance between two subsequent lower (or upper) extreme values of the curve shown in
The above described oscillometric non-invasive blood pressure measurement method is relatively popular, because it enables a skilled practitioner to easily determine the systolic arterial blood pressure SAP and the diastolic arterial blood pressure DAP of a patient (by using an empirical approach). It is known that the oscillation amplitude is between 45-57%, usually 50%, of the maximum oscillation amplitude at a clamp pressure equal to the systolic arterial blood pressure SAP, whereas the oscillation amplitude is between 75-86%, usually 80%, of the maximum oscillation amplitude at a clamp pressure equal to the diastolic arterial blood pressure DAP. Thus, the absolute pressure values indicated by the manometer at corresponding moments correspond to the diastolic arterial blood pressure DAP and the systolic arterial blood pressure SAP. Notably, instead of a classic manometer, an electrical sensor may be equally applied. The above-described principle is conferrable to other physical values, such as acceleration, sound and optical reflection.
Disadvantageously, it is impossible with this known blood pressure measurement method to reliably determine the shape of the pulse wave of the patient's arterial blood pressure. As mentioned above, the exact shape of the patient's pulse wave, however, can provide important information (as to the health status of this patient) to a skilled practitioner, such as an experienced physician.
EP 0 078 090 A1 describes a non-invasive blood pressure measurement method that is—at least theoretically—capable of determining the arterial pulse wave of a patient. According to this method, a fluid-filled pressure cuff is attached to a patient's finger. A light source and a light detector are integrated in the pressure cuff, the light source and the light detector forming part of a photo-electric plethysmograph. The cuff pressure is controlled—via a fast-acting electric pressure valve—in closed-loop operation based on the plethymographic signal, so that the arterial volume in the finger is maintained at a predefined value. Measuring the pressure in the pressure cuff, thus, allows for determining the arterial blood pressure of the patient. This method is also known in literature as “volume-clamp-method”.
However, permanently correcting or re-adjusting the pressure in the pressure cuff in real time is technically difficult and prone to errors. Furthermore, till now, this method only works with a pressure cuff applied to a patient's finger. The finger is yet located relatively remote from the patient's heart, so that the diameter of arterial vessels in the finger is relatively small compared to the diameter of arterial vessels close to the heart. Due to interference effects caused by pressure reflections occurring when the diameter of arterial vessels (abruptly) changes, e.g. when arterial vessels branch, the blood pressure measurable at the finger only imprecisely corresponds to the arterial pulse wave at the patient's heart. To consider these interference effects, it was tried to reconstruct the pulse curve in the patient's aorta from the signals measured at the patient's finger, using transfer functions that are usually based on empirical and statistical parameters. However, since the parameters are not (or not sufficiently) adapted to the individual patient and situation, such an approach is not promising, potentially providing imprecise results.
It is therefore the object of the present invention to provide a method and a corresponding device for better approximating a patient's central arterial pulse wave based on non-invasive blood pressure measurement.
This object is achieved by the subject-matters of the independent claims. Preferred embodiments are the subject-matter of the dependent claims.
According to a first aspect of the present invention, there is provided a method of approximating a patient's pulse wave based on non-invasive blood pressure measurement, comprising the following steps:
A good approximation of the patient's pulse wave pulseapprox(t) can be obtained simply by weighting a corresponding number of measured pulse signals pulsen_measured(t) and then adding up the weighted pulse signals pulsen_weighted(t). The method according to the present invention allows—without difficulty—for measuring the pulse signals close to the patient's heart, e.g. at the patient's upper arm, so as to substantially avoid interference effects otherwise occurring when measuring the pulse signals at locations remote from the patient's heart, e.g. at a finger of the patient, as with the above-described volume-clamp-method. Furthermore, the method according to the present invention allows for determining an approximation of the patient's pulse wave without the need of permanently correcting or re-adjusting the pressure in the pressure cuff, like with the volume-clamp-method.
In method step (a) of the inventive method, a sequence n=1 . . . N of pulse signals pulsen_measured(t) of a patient is measured in an non-invasive fashion, using a (not-constant) clamp pressure clampn(t). N corresponds to the total number of individual heart beats of the patient within the detection period. In the exemplary diagram shown in
Preferably, the well-known and relatively comfortable “oscillometric non-invasive blood pressure measurement method” (as described above) is applied to measure the patient's pulse signals. In this method, the clamp pressure clampn(t) is applied to an extremity of the patient, preferably to an upper arm of the patient, as shown e.g. in
Thereby, the clamp pressure clampn(t) may be increased or decreased continuously, preferably at a substantially constant rate. Notably, the increase or decrease rate should be low enough so as to detect a sufficient number of pulses caused by individual heart beats of the patient (preferably at least 10). The clamp pressure clampn(t) may be continuously increased or decreased between the diastolic arterial blood pressure DAP and the systolic arterial blood pressure SAP within a detection period of e.g. about one minute. During this detection period, pulse signals pulsen_measured(t) associated to e.g. 60 individual heart beats of the patient may be measured, which represents a very good base for the further method steps. However, to avoid problems caused by blockage of blood circulation in the patient's extremity to which the pressure cuff is applied, the increase or decrease rate of the clamp pressure clampn(t) should not be too low, i.e. the detection time should preferably not exceed one minute.
If the increase or decrease rate of the clamp pressure clampn(t) is moderate (e.g. the detection time is about one minute), the clamp pressure clampn(t) associated with the time period of one individual heart beat might be considered—for the sake of simplicity—as being substantially constant. For example, the clamp pressure clampn(t) associated with the time period of one individual heart beat might be approximated so as to correspond to the actual clamp pressure at the beginning (t=tbeat_n) of the corresponding heart beat (clampn=clamp(tbeat_n)). In the example shown in
However, the clamp pressure clampn(t) associated with the time period of one individual heart beat might equally be approximated so as to correspond e.g. to the actual clamp pressure at the end or somewhere in the middle (preferably exactly in the middle) of the corresponding heart beat.
Notably, if the sequence of pulse signals pulsen_measured(t) has been previously measured and stored, the method step (a) might be skipped and the method according to the present invention may directly start with method step (b) based on the previously stored signal values.
In method step (b) of the inventive method, the measured pulse signals pulsen_measured(t) are weighted to obtain weighted pulse signals pulsen_weighted(t). A weighting function is applied in method step (b), as will be described in more detail below.
Finally, in method step (c), the weighted pulse signals pulsen_weighted(t) are added up to obtain the approximation of the patient's pulse wave pulseapprox(t). In method step (c), the approximation of the patient's pulse wave pulseapprox(t) might be simply calculated as follows:
As set forth above, the measured pulse signals pulsen_measured(t), i.e. the cyclic pressure variations corresponding to the individual heart beats, detected by the manometer (as shown in
The inventors have found out that at moments, when the actual internal pressure (i.e. arterial blood pressure) equals a predetermined difference, e.g. approximately zero, to the externally applied pressure (i.e. cuff pressure), there exists a substantially linear relationship between the measured pulse signals and the actual arterial blood pressure (for example, at moments when the applied cuff pressure substantially equals the actual internal arterial blood pressure, the body tissue between the artery, e.g. in the upper arm, and the pressure cuff is relaxed, i.e. not biased).
Therefore, it is advantageous—in order to obtain an improved approximation of the patient's pulse wave—to use the clamp pressure clampn(t) as an input parameter of the weighting function, wherein the weighting function is preferably a differential pressure function. That is, the measured pulse signals pulsen_measured(t) may be weighted in such a way that those portions of the curve of the measured pulse signals pulsen_measured(t) are more “emphasised” that have been measured during moments at which the actual internal arterial blood pressure equals a predetermined percentage of the externally applied cuff pressure.
However, a problem resides in that usually the moments, at which the actual internal arterial blood pressure equals a predetermined percentage of the externally applied cuff pressure, are unknown, because the actual internal arterial blood pressure (i.e. the patient's pulse wave) is unknown. In fact, an approximation of the patient's pulse wave is the pursued result of the method.
To overcome this problem, method steps (b) and (c) of the method according to the present invention are preferably iteratively repeated at least one more time. The outcome of the first iteration loop, i.e. the approximation of the patient's pulse wave, can then be used as approximation of the actual internal arterial blood pressure in the second iteration loop. Thus, the moments, at which the actual internal arterial blood pressure equals a predetermined percentage of the externally applied cuff pressure, can be (at least approximately) determined. In the second iteration loop, the measured pulse signals pulsen_measured(t) can then be weighted accordingly (in step (b) of the second iteration loop) before adding up the weighted pulse signals pulsen_weighted(t) (in step (c) of the second iteration loop) so as to obtain an improved approximation of the patient's pulse wave pulseapprox(t).
The outcome of the method can be even further improved by iteratively repeating method steps (b) and (c) more than two times, wherein the outcome of method step (c) of the previously iteration loop is used as input value for the present iteration loop. Preferably, the weighting function of the present iteration loop is a differential pressure function comprising, as an input parameter, the externally applied cuff pressure and, as another input parameter, the approximation of the actual internal arterial blood pressure, i.e. the approximated patient's pulse wave pulseapprox(t) determined in the previous iteration loop.
Of course, it is not possible to determine the weighting function to be applied in the very first iteration loop that way, since there is no result of a previous iteration loop available as input. Consequently, the weighting function applied in the first iteration loop preferably differs from the weighting function applied in the second and/or higher iteration loop. The first iteration loop, thus, provides a more roughly approximated pulse wave pulseapprox(t) of the patient compared to following iteration loops. For example, the weighting function of the first iteration loop might simply be determined as follows:
weight1n=1 if DAP<clampn(t)<SAP,
weight1n=0 otherwise,
wherein DAP corresponds to the diastolic arterial blood pressure and SAP corresponds to the systolic arterial blood pressure of the patient.
In such a case, in method step (c) of the first iteration loop, the following formula might be applied for calculating the approximated pulse wave pulseapprox(t);
As mentioned above, the weighting function applied in the second and/or higher iteration loop is preferably a differential pressure function having the clamp pressure clampn(t) as an input parameter and having the approximated pulse wave pulseapprox(t) obtained in the previous iteration loop as another input parameter.
For example, the weighting function applied in the second and/or higher iteration loop can be a triangular function, preferably having its maximum when the clamp pressure clampn(t) equals a predetermined difference, preferably zero, to the approximated pulse wave pulseapprox(t) obtained in a previous iteration loop.
If a triangular function is applied as weighting function weightn(t) in the second and/or higher iteration loop, the weighting function weightn(t) might be calculated as follows:
As mentioned above, the index n refers to the number of the heart beat of the corresponding measured pulse signal pulsen_measured(t). As mentioned above, even though, the clamp pressure clampn may not be constant over the detection period of one pulse wave, for the sake of simplicity, the clamp pressure clampn might be considered as being substantially constant during this detection period, e.g. corresponding to the clamp pressure clampn=clampn(t=tbeat_n,) at the beginning of the corresponding detection period. pulseapprox_prev(t) corresponds to the result, i.e. the approximated pulse wave of the patient, of the previous iteration loop.
As an alternative to a triangular weighting function, the weighting function applied in the second and/or higher iteration loop may be a bell-shaped function, preferably having its maximum when the clamp pressure clampn(t) equals a predetermined difference, preferably zero, to the approximated pulse wave pulseapprox(t) obtained in a previous iteration loop.
When using a bell-shaped function as weighting function weightapprox_n(t) in the second and/or higher iteration loop, the weighting function weightn(t) might be calculated as follows:
wherein parameter pw corresponds to an empirically determined parameter being decisive for the width at half maximum of the bell-shaped weighting function. The parameter pw is preferably chosen in accordance with the particular circumstances of the blood pressure measurement that have an influence on the distortion of the measured pulse curves. If distortion of the measured pulse curves increases (e.g. due to the use of another blood pressure measurement device), the increase or decrease rate of the cuff pressure should be decreased so as to measure more pulses of the patient within the detection time. In such a case, a smaller value for the parameter pw may be chosen. Generally, the parameter pw may preferably be chosen according to the following equation:
wherein N is the total number of pulses measured during the detection period, i.e. measured substantially during the time needed by the cuff pressure to change from the diastolic arterial blood pressure DAP to the systolic arterial blood pressure SAP of the patient, or the other way around.
Preferably, method step (c) of the second and/or higher iteration loop further comprises: scaling the approximated pulse wave pulseapprox(t) to the difference between the diastolic blood pressure value DAP and the systolic blood pressure value SAP of the patient. An example of such a scaling is provided below.
Scaling the approximated pulse wave pulseapprox(t) ensures that the amplitude of the (scaled) approximated pulse wave correctly corresponds to the amplitude of the actual pulse wave of the patient. That is, the approximated pulse wave pulseapprox(t) is scaled in such a way that its lower extreme value substantially corresponds to the diastolic blood pressure value DAP of the patient, whereas its upper extreme value substantially corresponds to the systolic blood pressure value SAP of the patient. As mentioned before, it is well-known to those skilled in the art, how to determine diastolic and systolic arterial blood pressure values DAP and SAP based on the so-called “oscillometric non-invasive blood pressure measurement method”.
If the approximated pulse wave pulseapprox(t) is scaled in method step (c) of the second and/or higher iteration loop, the scaled approximated pulse wave pulseapprox_scaled(t) (instead of the approximated pulse wave pulseapprox(t)) is applied in method step (b) of the subsequent iteration loop.
Similarly, method step (a) may further comprise: scaling the measured pulse signals pulsen_measured(t) to the difference between the diastolic blood pressure value DAP and the systolic blood pressure value SAP of the patient.
Scaling of the measured pulse signals pulsen_measured(t) in method step (a) might be performed by applying the following formula:
pulsen_measured_scaled(t)=offsetn+scalen×pulsen_measured(t),
wherein the parameter offsetn is preferably calculated as follows:
offsetn=DAP−min(pulsen_measured(t)), and
wherein the parameter scalen is preferably calculated as follows:
max(pulsen_measured(t)) corresponds the maximum value of the measured pulse wave corresponding to the heart beat with the number n. Similarly, min(pulsen_measured(t)) corresponds the minimum value of the measured pulse wave corresponding to the heart beat with the number n.
Of course, as will be apparent to those skilled in the art, other formulas may be applied to calculate the parameters offsetn and scalen. For example the formulas might equally be based on the mean arterial pressure MAP of the patient, which can also be determined based on the so-called “oscillometric non-invasive blood pressure measurement method”.
If the measured pulse signals pulsen_measured(t) are scaled in method step (a), the scaled measured pulse signals pulsen_measured_scaled(t) (instead of the measured pulse signals pulsen_measured(t)) are applied in method step (b) to determine the weighted pulse signals pulsen_weighted(t).
In such a case, in method step (c) of the first iteration loop, the following formula might be applied for calculating the approximated pulse wave pulseapprox(t):
wherein the weighting function weightn(t) applied in the first iteration loop is preferably a function of a difference between scaled measured pulse signals pulsen_measured_scaled(t) and the clamp pressure clampn(t).
According to another aspect, the invention refers to a logic unit for approximating a patient's pulse wave based on a non-invasive blood pressure measurement, configured to carry out the following steps:
The logic unit according to the present invention is configured to carry out the above described method, wherein the sequence of pulse signals pulsen_measured(t) has been previously measured and stored, so that method step (a) can be skipped and the logic unit according to the present invention directly starts with method step (b), based on the previously stored signal values.
According to yet another aspect, the present invention also refers to a system for approximating a patient's pulse wave based on a non-invasive blood pressure measurement, comprising the logic unit described above and a blood pressure measurement device, the blood pressure measurement device being configured for non-invasively measuring a sequence n=1 . . . N of pulse signals of a patient to obtain measured pulse signals pulsen_measured(t), wherein the system is configured for providing the measured pulse signals pulsen_measured(t) as input values to the logic unit. Thus, the system is also configured to obtain the measured pulse signals pulsen_measured(t) according to step (a) of the above described method.
Preferably, the blood pressure measurement device comprises a pressure cuff, and even more preferably, the pressure cuff is configured for being disposed around a patient's arm so as to measure the patient's arterial blood pressure in a non-invasive way. Thus, the system is configured to obtain the measured pulse signals pulsen_measured(t) using the above-described “oscillometric non-invasive blood pressure measurement method”. Since the pressure cuff is configured for being attached around a patient's arm, preferably an upper arm of the patient, substantially no interference effects caused by pressure reflections adversely affect the measurement—contrary to the above described “volume-clamp-method”.
Even though, the “oscillometric non-invasive blood pressure measurement method” exhibits the advantage that substantially no interference effects caused by pressure reflections adversely affect the measurement (in contrast to known methods for measuring peripheral blood pressure waveform data, such as the above described “volume-clamp-method” utilized on a finger or the so called “applanation-tonometry-method” utilized at the patient's wrist), the “oscillometric non-invasive blood pressure measurement method” does not allow for continuous measurements without blocking blood flow in an unallowable manner. However, continuous measurement can be performed with the “volume-clamp-method” or with the “applanation-tonometry-method”.
As described above, it has already been tried in the past to overcome the disadvantages of the known methods for measuring peripheral blood pressure waveform data, such as the “volume-clamp-method” and the “applanation-tonometry-method”, by using transfer functions in order to reconstruct the central blood pressure waveforms from peripherally measured signals, e.g. signals measured at a patient's finger. However, since the applied transfer functions are usually based on statistical and empirical parameters that are not (or at least not sufficiently) adapted to the individual patient and situation, such an approach bears the likelihood to provide imprecise results.
Using pulse signals measured e.g. with the above described “oscillometric non-invasive blood pressure measurement method” to calibrate the transfer function to an individual patient did not represent a promising approach, either, since, in the past, it was not possible to approximate (with sufficient quality) the patient's pulse wave based on the measured pulse signals. However, with the method according to the present invention, an approximation of good quality of the patient's pulse wave based on non-invasive blood pressure measurement becomes possible. Therefore, the system described above may be combined with a device for measuring peripheral blood pressure waveform data, and a transfer function may be applied, wherein the transfer function is calibrated to an individual patient based on the patient's pulse wave that has been previously approximated according to the method of the present invention. That way, the central arterial blood pressure waveforms can be continuously determined with high quality.
Thus, the previously described system for approximating a patient's pulse wave based on a non-invasive blood pressure measurement preferably further comprises a second blood pressure measurement device that is adapted for non-invasively measuring peripheral blood pressure waveform data of the patient in a continuous way, wherein the system is adapted to apply a transfer function to reconstruct central blood pressure waveforms from the measured peripheral blood pressure waveform data based on the approximated pulse wave pulseapprox(t).
The patient's pulse wave may be approximated according to the inventive method only once, preferably just before the continuous measurement of the peripheral blood pressure waveform data.
More preferably, the approximated pulse wave pulseapprox(t) of the patient is yet determined at substantially regular intervals, wherein the transfer function is regularly recalibrated based on the regularly determined approximated pulse wave pulseapprox(t) of the patient. For example, the pulse wave may be approximated according to the method of the present invention every two minutes. This way, it is possible to continuously obtain central blood pressure waveforms of very good quality.
For example, applying the transfer function may comprise the following steps: In a first step, both time-varying signals, i.e. the intermittent determined approximated pulse waves pulseapprox(t) and the continuously measured peripheral blood pressure waveforms, are transformed into the frequency domain. Then, in a second step, the transfer function is determined. In a third step, the transfer function is applied to the peripherally measured blood pressure waveforms so as to calibrate the peripheral blood pressure waveforms. Finally, in a fourth step, the calibrated peripheral blood pressure waveforms are re-transformed into the time domain.
Exemplary embodiments of the present invention are described in more detail below based on the figures, in which:
pressure oscillations caused by the patient's heart beats, indicated by the manometer over the time, thereby omitting the pressure variation caused by the continuously increasing cuff pressure;
As described above,
In the following, the present invention will be described in more detail in view of
The measured pulse signals pulsen_measured(t) of a patient are scaled so as to make the amplitudes of the (scaled) measured pulse signals all correspond to the amplitude of the actual pulse wave of the patient. That is, the measured pulse signals pulsen_measured(t) are scaled in such a way that the lower extreme value of each measured pulse signal substantially corresponds to the diastolic blood pressure value DAP of the patient, whereas its upper extreme value substantially corresponds to the systolic blood pressure value SAP of the patient. As mentioned before, it is well-known to those skilled in the art, how to determine diastolic and systolic arterial blood pressure values DAP and SAP, e.g. by applying the “oscillometric non-invasive blood pressure measurement method”.
Even though the measured and scaled pulse signals pulsen_measured_scaled(t) all have the same amplitude,
Furthermore, a curve named “first pulseapprox(t)” is illustrated in
In other words, the following formula has been applied for calculating the approximated pulse wave first pulseapprox(t) of the first iteration loop:
However, instead of obtaining the approximated pulse wave first pulseapprox(t) of the first iteration loop by simply averaging the curves of the measured and scaled pulse signals pulsen_measured_scaled(t), the measured and scaled pulse signals pulsen_measured_scaled(t) may be weighted in a more sophisticated way before adding them up. For example, a bell-shaped weighting function may be applied is schematically illustrated in
Calculation of the bell-shaped weighting function weightn(t) for the first iteration loop works substantially analogue to the calculation of the bell-shaped weighting function weightn(t) for the second or higher iteration loop, which has been described in detail above.
Generally, the approximated pulse wave pulseapprox(t) lasts on the axis of the abscissas t-tbeat_n from 0 seconds till a mean pulse duration time tmean of the measured and scaled pulse signals pulsen_measured_scaled(t). In the example shown in
Moreover,
The same is repeated for all measured and scaled pulse signals pulsen_measured_scaled(t), wherein always those portions of the signal curves are particularly accentuated which correspond to moments at which the approximated pulse wave first pulseapprox(t) determined in the first iteration loop substantially equals the corresponding clamp pressure clampn. Then, the weighted pulse signals pulsen_weighted(t) are added up to obtain a better approximation of the patient's pulse wave second pulseapprox(t) as result of the second iteration loop.
Notably, it is not necessary to accentuate the portions of the signal curves which correspond to moments at which the approximated pulse wave first pulseapprox(t) determined in the first iteration loop substantially equals the corresponding clamp pressure clampn (i.e. the difference is zero). Instead, it is also possible to apply another difference, as long as the same difference is applied for weighting all of the measured and scaled pulse signals pulsen_measured_scaled(t).
Finally,
It should be noted that in
With such a system, the method shown in
E.g. the intermittent central blood pressure curve pc(t) and the peripheral blood pressure signal pp(t) are then both transformed into the frequency domain, so as to obtain a central blood pressure signal curve in the frequency domain Pc(f) and a peripheral blood pressure signal curve in the frequency domain Pp(f).
Next, a transfer function G(f) is calculated, based on the central blood pressure signal curve in the frequency domain Pc(f) and a peripheral blood pressure signal curve in the frequency domain Pp(f).
A calibrated blood pressure signal curve Pc(f)* can then be simply obtained by multiplying the transfer function G(f) with the a peripheral blood pressure signal curve in the frequency domain Pp(f).
Finally, a calibrated blood pressure curve signal pc(t)* is determined by transforming the calibrated blood pressure signal curve Pc(f)* again into the time domain.
Number | Date | Country | Kind |
---|---|---|---|
13000376.7 | Jan 2013 | EP | regional |
Number | Date | Country | |
---|---|---|---|
61756895 | Jan 2013 | US |
Number | Date | Country | |
---|---|---|---|
Parent | 14761649 | Jul 2015 | US |
Child | 17474668 | US |