Apparatus for determining an index representative of a fluid reactivity parameter - Patent Application 20070122997
Patent Information
- Application Number
- JP2023533256
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2020-12-02
- Filing Date
- 2021-11-29
- Publication Date
- 2024-12-04
AI Technical Summary
Existing methods for determining fluid responsiveness parameters, such as pulse pressure variation (PPV), systolic pressure variation (SPV), and stroke volume variation (SVV), are computationally intensive due to the use of the Levenberg-Marquardt algorithm, leading to prolonged calculation times.
A device and method that reduce the number of data values used for fitting by employing a multi-stage approach, using initial values of fitting parameters based on previous fittings and characteristics of the data, and adapting a function prototype in stages, thereby reducing computation time.
This approach significantly speeds up the determination of fluid responsiveness parameters like PPV, SPV, and SVV, maintaining accuracy by minimizing the number of iterations and reducing overall computation time.
Smart Images

Figure 00000000_0000_ABST
Abstract
Description
[Technical Field]
[0001] The present invention relates to an apparatus, method, and computer program for determining an index representing a fluid reactivity parameter. [Background technology]
[0002] U.S. Patent No. 5,704,362(A) discloses a method for measuring a person's blood pressure. Multiple oscillometric data values are obtained from complex amplitudes acquired at multiple pressure levels. A function curve having a physiologically expected shape for the oscillometric envelope is selected, and the function curve is defined by parameters. The function curve is fitted to the multiple data values by calculating the parameters of the function curve such that the error between the function curve and the data values is minimized at multiple pressure levels, and finally, the systolic and diastolic blood pressures are calculated based on the fitted function curve.
[0003] U.S. Patent No. 5,577,508 (A1) discloses an automated blood pressure monitor device having an inflatable and deflated pressure cuff and an inflation means operably coupled to the cuff for selectively applying a medium to the cuff under pressure to inflate and pressurize the cuff. The cuff further includes a cuff-coupled cuff pressure sensing means for sensing cuff pressure, including internal arterial blood pressure oscillations. The device also includes a deflation means operably coupled to the cuff for selectively releasing pressure from the cuff. A control means controls the inflation means to inflate the cuff to each cuff pressure level and the deflation means to deflate the cuff in order to detect arterial blood pressure oscillations at each cuff pressure level during blood pressure determination. An oscillometric envelope is determined, including pulse amplitude points corresponding to arterial blood pressure oscillations at each cuff pressure level during blood pressure determination. The oscillometric envelope is linearly approximated by a pair of best-fit lines, and the patient's mean arterial pressure is determined based on the pressure at the intersection of the pair of best-fit lines.
[0004] The paper "Improved Measurement of Blood Pressure by Extraction of Characteristic Features from the Cuff Oscillometric Waveform" by PKLim et al., SENSORS, Vol. 15, pp. 14142-14161 (2015), discloses a method for determining systolic and diastolic blood pressure from oscillometric waveform data, and CN105342590A discloses an oscillometric method for measuring blood pressure.
[0005] EP2759257B1 discloses a method for determining an index representing a patient's volume responsiveness. A sequence of the patient's pulse signal is measured over a time period corresponding to several subsequent respiratory cycles of the patient. The measured sequence of pulse signals is detected by a non-invasive pulse measurement method using a pressure cuff, where the pressure applied to the pressure cuff is continuously increased or decreased over the course of the measurement period. The measured pulse signal is represented as a pulse signal oscillating around its mean value, and an envelope signal curve of the pulse signal is determined. Furthermore, a fitted envelope signal function is determined based on the previously determined envelope signal curve. The fitted envelope signal function is determined based on a predetermined first function prototype and represents an idealized curve transition of the envelope signal curve over several respiratory cycles, excluding pulse variability caused by cardiopulmonary interactions induced by ventilation or respiration. Next, a respiratory pulse variability signal is determined, corresponding to pulse variability caused by cardiopulmonary interactions induced by ventilation or respiration over several respiratory cycles. The determined respiratory pulse variability signal corresponds to the difference between the envelope signal curve and the fitted envelope signal function. As the next step, an envelope breathing curve is determined for the previously determined respiratory pulse variability signal, and a fitted envelope breathing function is determined based on the previously determined envelope breathing curve. The fitted envelope breathing function is determined based on a predetermined second function prototype and represents the idealized curve transition of the envelope breathing curve over multiple respiratory cycles. Finally, an index representing the patient's volume responsiveness is determined based on the fitted envelope signal function and the fitted envelope breathing function.
[0006] The fitted envelope signal function is determined by fitting the Cauchy-Lorentz function to the envelope signal function using the Levenberg-Marquardt algorithm. Furthermore, the fitted envelope breathing function is determined by fitting the Cauchy-Lorentz function to the envelope breathing function, also using the Levenberg-Marquardt algorithm.
[0007] The Levenberg-Marquardt algorithm is computationally intensive and therefore time-consuming, which generally means that it takes a very long time to determine an index representing a patient's volume responsiveness. [Overview of the project] [Problems that the invention aims to solve]
[0008] The object of the present invention is to provide an apparatus, method, and computer program that enable faster determination of indices representing fluid reactivity parameters. [Means for solving the problem]
[0009] In a first aspect of the present invention, an apparatus is provided for determining an index representing a fluid reactivity parameter, the apparatus is - A pulse signal providing unit configured to provide a sequence of measured pulse signals of a patient over a time period corresponding to multiple subsequent respiratory cycles of the patient, wherein the sequence of measured pulse signals is detected by a non-invasive pulse measurement method, - An index determination unit configured to perform a determination procedure adapted to determine an index representing fluid reactivity parameters based on a given sequence of pulse signals, wherein the determination procedure includes determining data values based on a given sequence of pulse signals and fitting a given function prototype, which depends on fitting parameters modified during fitting, to the determined data values, wherein the index determination unit is configured such that a) before fitting, the number of data values for the entire fitting is reduced, b) initial values of fitting parameters are determined based on the characteristics of the data values related to the characteristics of the function prototype and / or based on the values of fitting parameters known from previous fittings, and / or c) the fitting is performed in several stages, the number of data values used for fitting increases in each stage, and in the current stage, the values of the fitting parameters determined in the previous stage are used as the initial values of the fitting parameters in the current stage. It is equipped with.
[0010] The computation time is significantly reduced by performing the fitting in several stages, where the number of data values used increases with each stage, and in the current stage, the values of the fitting parameters determined in the previous stage are used as the initial values of the fitting parameters. This allows for a faster determination of indicators representing fluid reactivity parameters.
[0011] The fluid responsiveness parameter is preferably the fluid responsiveness parameter of a patient. Alternatively, the fluid responsiveness parameter may be the fluid responsiveness parameter of a volunteer or animal. The fluid responsiveness parameter exhibits or corresponds to the volume responsiveness or fluid responsiveness of a patient; that is, fluid responsiveness is also called "volume responsiveness." Therefore, the determination of an index representing the fluid responsiveness parameter is preferably the determination of the fluid responsiveness parameter (FRP), particularly the pulse pressure variation (PPV), systolic pressure variation (SPV), and / or stroke volume variation (SVV). In these specific cases, the index to be determined is the respective FRP, i.e., PPV, SPV, and / or SVV.
[0012] The provision of a sequence of measured pulse signals can correspond to measuring a sequence of a patient's pulse signals, specifically, measuring a signal corresponding to blood pulsation by continuously or semi-continuously recording the pressure indicating blood pulsation. In one embodiment, the sequence of measured pulse signals is detected by a non-invasive pulse measurement method using a pressure cuff, where the pressure applied to the pressure cuff continuously increases or decreases over the course of the measurement period. The measured signal corresponds to a continuous or semi-continuous recording of the pressure in the pressure cuff. The provision of a sequence of measured pulse signals also includes, for example, the provision of a sequence of pulse signals previously measured by a physician. The measured pulse signals are then stored, and this stored sequence is provided. Generally, non-invasive pulse measurement methods include, for example, oscillometric methods. In oscillometric pulse measurement methods, a pressure cuff is typically inflated, resulting in an air-pressure coupling between the blood pulsation and a pressure sensor that records the pressure in the inflated volume. However, the pulse signal sequence is preferably measured using a pressure cuff adapted to provide a hydraulic coupling between the blood pulsation and a pressure sensor of the pressure cuff adapted to record the pressure indicating the blood pulsation. This type of pressure cuff is disclosed, for example, in WO2014 / 121945A1. By using such a pressure cuff, also hereafter referred to as a high-fidelity pressure cuff, it is possible to improve the quality of the recorded pressure signal compared to the case of oscillometric measurements that rely on pneumatic coupling. For example, the pressure cuff is adapted to bring the pressure sensor into direct contact with the patient's tissue so that the pressure sensor is hydraulically coupled to the blood pulsation via the tissue, and it is preferred that there is no air between the pressure sensor and the patient's tissue, i.e., the patient's skin, in order to ensure a hydraulic coupling between the blood in the blood vessel and the pressure sensor. The pressure cuff is, for example, a high-fidelity upper arm pressure cuff. The hydraulic coupling between the pressure sensor and the blood pulsation allows for more accurate detection of the blood pulsation and, therefore, an improved determination of fluid reactivity parameters based on the measured pulse signal.
[0013] A pressure cuff may include a fluid-based pressure sensor, such as a pressure transducer, which performs actual pressure measurement, in addition to a pressure-sensing element that performs actual pressure measurement. This sensor sensor may include, for example, a fluid-filled sensor pad on the patient's skin connected to the pressure-sensing element. The sensor pad may be filled with a water-based gel, silicone oil, or another fluid. The pressure-sensing element may be located within or on the sensor pad, or separately from the sensor pad, in which case the fluid is used to provide a hydraulic coupling between the sensor pad and the pressure-sensing element, for example, using a fluid-filled tube extending between the sensor pad and the pressure-sensing element. Instead of a fluid-filled sensor pad, another fluid-filled element, such as a fluid-filled pressure-sensitive film, may be used. Furthermore, when the pressure cuff uses a fluid-based pressure sensor with a pressure-sensing element, it is preferable that the pressure sensor is coupled to the outer shell without air. Additionally, it is preferable that the fluid is an incompressible fluid. Therefore, in one embodiment, the pressure cuff comprises a fluid-based pressure sensor having a pressure-sensitive film filled with an incompressible fluid or a fluid-filled sensor pad bonded to the tissue, i.e., the outer skin, without air, and the limbs are partially or completely enclosed.
[0014] The determination procedure adapted for the index determination unit to perform includes determining data values based on a provided sequence of measured pulse signals. For example, the pulse signal providing unit is configured to provide a sequence of measured pulse signals as a continuous or quasi-continuous digital signal, and the index determination unit is adapted to determine data values by discretizing the provided sequence of measured pulse signals and / or determining the start and end points of the provided sequence of measured pulse signals. The data values are still understood as a direct representation or sample of the sequence of measured pulse signals. However, determining data values based on a provided sequence of measured pulse signals also includes, as an addition or alternative, other types of preprocessing of the measured pulse signal, such as determining an envelope signal curve, determining a respiratory pulse variability signal, and / or determining signals or curves derived from respiratory pulse variability signals, such as an absolute respiratory pulse variability curve, or another curve showing the absolute value of the envelope respiratory curve or respiratory pulse variability signal for the measured pulse signal. In this case, the data values correspond to the envelope signal curve, or the signal or curve derived from the respiratory pulse variability signal, i.e., a discretized version thereof, respectively.
[0015] The determination procedure further includes fitting a given function prototype to determined data values. The function prototype may depend on one or more fitting parameters that are modified during the fitting process, and the initial values of each of the one or more fitting parameters may be determined. The initial values of one or more of the fitting parameters, or only some of them, may be determined based on the characteristics of the data values related to the characteristics of the function prototype and / or based on the values of each fitting parameter known from previous fittings. For one or more fitting parameters whose initial values are not determined based on the characteristics of the data values related to the characteristics of the function prototype and / or based on the values of each fitting parameter known from previous fittings, the initial values are either predefined or determined based on the characteristics of a measurement method, such as the rate of change of pressure applied to a compression cuff.
[0016] Depending on the expected format of the sequence of measured pulse signals, the function prototype can be predetermined, and the format is expected, for example, based on the measurement method. Furthermore, the function prototype can also be predetermined according to the method selected to determine the data values based on the provided sequence of measured pulse signals, and the selected method corresponds, for example, to a defined preprocessing.
[0017] In one embodiment, a pulse signal providing unit is configured to provide several sequences of measured pulse signals from a patient over time, each of which is detected during a different time period, and an index determination unit is configured to determine a set of data values for each of the provided sequences of measured pulse signals. The different time periods correspond, for example, to subsequent cycles of inflation and / or deflation of a pressure cuff.
[0018] Any known fitting algorithm is used to fit the provided function prototype to the determined data values. Preferably, a nonlinear least-squares algorithm, particularly the Levenberg-Marquardt algorithm, is used. However, it is understood that the above-mentioned effect of reducing computational complexity can also be achieved with different fitting algorithms.
[0019] In one embodiment, the index determination unit is configured to reduce the number of data values such that the reduced number of data values satisfies the Shannon-Nyquist criterion with respect to the bandwidth of the provided function prototype and / or with respect to the data values themselves. For example, the number of data values is reduced based on which data values are determined such that the reduced number of data values satisfies the Shannon-Nyquist criterion with respect to the provided sequence of pulse signals. This is particularly applicable when the data values are determined to represent the provided sequence of pulse signals. The bandwidth of the provided function prototype, specifically the frequencies included therein, is estimated based on, for example, the time elapsed while the pressure in the pressure cuff increases or decreases from a first pressure value to a second pressure value, where the first pressure value and the second pressure value each refer to, for example, a diastolic arterial pressure value and a systolic arterial pressure value. When the data values are determined to represent the patient's respiratory cycle, the number of data values is reduced such that the reduced number of data values satisfies the Shannon-Nyquist criterion with respect to the patient's respiratory cycle. For example, when data values having a sample every 0.004 seconds and a maximum frequency of 0.5 Hz of the subject are fitted to a bell-shaped curve with a full width at half maximum of 20..100 seconds, the reduction of the data values to be fitted can be performed by selecting one data value per second.
[0020] In one embodiment, the index determination unit is configured to reduce the number of data values by determining one data value at a predetermined frequency and / or by downsampling. In a further embodiment, the index determination unit is configured to low-pass filter the data values before reducing the number of data values.
[0021] The function prototype is preferably a bell function. This bell function is particularly preferred when a sequence of measured pulse signals is detected using a pressure cuff, and the pressure applied to the pressure cuff continuously increases or decreases over the course of the measurement period. The bell function is preferably a Cauchy-Lorentz function. The bell function may have a first fitting parameter indicating the height of the maximum value of the bell function and a second fitting parameter indicating the position of the maximum value of the bell function, and the index determination unit may be configured to determine the height of the maximum value of the data value and the position of the maximum value of the data value as the initial values of the first fitting parameter and the second fitting parameter.
[0022] Providing initial values for fitting parameters that are close to their final values can significantly reduce computation time. This prediction can be made by analyzing the data values of the characteristics that describe the function prototype. By pre-determining and pre-selecting characteristic patterns from the data values, the data values can be better fitted to the function prototype, thereby reducing the total number of iterations during fitting. For example, as described above, the height and position of the maximum value of the data values can be used as initial values for fitting parameters by fitting a bell-shaped function prototype on which at least two of the fitting parameters are the height and position of the maximum value. A third fitting parameter on which the bell-shaped function prototype may depend relates to the full width at half maximum. The initial value of this third fitting parameter can be determined without referring to data values, i.e., a) based on the time elapsed while the pressure in the pressure cuff increases from a first pressure value to a second pressure value, or the time elapsed while the pressure in the pressure cuff decreases from a second pressure value to a first pressure value, where the first pressure value and the second pressure value refer to, for example, the diastolic arterial pressure value and the systolic arterial pressure value, respectively; or the initial value of this third fitting parameter can be determined directly as b) the width of a pressure range including the first pressure value and the second pressure value, where the pressure range is, for example, from 0.7 times the diastolic arterial pressure value to 1.3 times the systolic arterial pressure value, or from a value 10 mmHg lower than the diastolic arterial pressure value to a value 20 mmHg higher than the systolic arterial pressure value. The diastolic and systolic arterial pressure values used to determine the third fitting parameter refer to, for example, predefined standard values or known mean values, or values previously measured for the patient.
[0023] In one embodiment, the index determination unit is configured such that the fitting is performed in three stages: in the first stage, a first percentage of the data value is used; in the next second stage, a higher second percentage of the data value is used; and in the next third stage, 100 percent of the data value is used. For example, the first percentage could be 1 / 10 percent and the second percentage could be 1 percent. The first and second percentages could be, for example, 1 / 250 and [Numerical] The corresponding percentages, that is, 0.4 percent and approximately 6.32 percent respectively, can also be used.
[0024] When applying this multi-stage adaptation, the total number of iterations may be more than when simply applying the adaptation algorithm to the initially determined data values. However, since most or all of the stages are executed with a reduced amount of data values used for calculation, the overall calculation time generally becomes shorter.
[0025] In a preferred embodiment, the index determination unit is configured such that the adaptation is executed in several stages. After the adaptation is executed at a certain stage, it is determined whether the adapted adaptation parameters meet the termination criteria. If they do, the adaptation ends.
[0026] For example, the index determination unit can be configured to use, as the termination criteria, the deviation between a) the adapted value of the adaptation parameters obtained as a result of each stage and b) the expected value of the adaptation parameters. For example, if the adaptation algorithm does not yield valid data because the underlying data values do not form a shape related to the function prototype, it may be concluded by analyzing the adaptation results of each stage before the final stage that it is not necessary to execute all the stages, thereby avoiding unnecessary calculation operations and thus shortening the execution time. As described above, when there is a certain range for the expected value of the final adaptation parameters and adaptation parameters that deviate significantly from these ranges are obtained at the stage before the final stage, the adaptation results can be analyzed.
[0027] Furthermore, the termination criterion can also indicate that the fit parameters obtained at a certain stage are already very good, for example, by determining whether the deviation between the fit parameters obtained at the current stage and those obtained at the previous stage is smaller than a predefined threshold. In this case, the fitting process can be stopped before the final stage is performed. This also allows for a reduction in calculation time.
[0028] When fitting is performed in several stages, i.e., in the case of multi-stage fitting, if the original number of data values are used in the final stage of fitting, the overall computation time can be reduced, but the accuracy of the final result, particularly the accuracy of the finally determined fitting parameters, may still be the same as when using a single-stage fitting with the original number of data values. To enable a further reduction in overall computation time, fitting is performed in several stages such that even in the final stage of fitting, fewer data values than the original number are used.
[0029] In a preferred embodiment, the pulse signal providing unit is configured to represent the measured pulse signal as a pulse signal that oscillates around its average value, and the index determination unit is configured - The first data value is determined by determining the envelope signal curve of the pulse signal. - In the first fit, a provided first function prototype, which depends on first fitting parameters modified during the first fit, is fitted to first data values such that the resulting fitted envelope signal function represents the idealized curve transition of the envelope signal curve over multiple respiratory cycles, excluding pulse rate variability caused by cardiopulmonary interactions induced by ventilation or respiration. - Respiratory pulse variability signals are determined that correspond to pulse variability caused by cardiopulmonary interactions induced by ventilation or respiration across multiple respiratory cycles, and the determined respiratory pulse variability signals correspond to the difference between the envelope signal curve and the fitted envelope signal function. - The second data value is determined by determining the absolute respiratory pulse variability curve of the respiratory pulse variability signal. - In the second fitting, the provided second function prototype, which depends on the second fitting parameters modified during the second fitting, is fitted to the second data values so that the resulting fitted respiratory function shows an idealized transition in the amplitude of the absolute respiratory pulse variability curve over multiple respiratory cycles. - Based on the fitted envelope signal function and fitted breathing function, an index representing the fluid reactivity parameter is determined, a) before the first fit, the number of first data values for the entire first fit is reduced, b) the initial value of the first fit parameter is determined based on the characteristics of the first data value related to the characteristics of the first function prototype and / or based on the value of the first fit parameter known from a previous first fit, and / or c) the first fit is performed in several stages, the number of first data values used in the first fit increases in each stage, and in the current stage, the value of the first fit parameter determined in the previous stage becomes the first fit parameter in the current stage The initial values of the second fitting parameters are used as the initial values of the second fitting parameters, and / or, a) before the second fitting, the number of second data values for the entire second fitting is reduced, b) based on the characteristics of the second data values related to the characteristics of the second function prototype, and / or based on the values of the second fitting parameters known from previous second fittings, the initial values of the second fitting parameters are determined, and / or, c) the second fitting is performed in several stages, the number of second data values used in the second fitting increases in each stage, and in the current stage, the values of the second fitting parameters determined in the previous stage are used as the initial values of the second fitting parameters in the current stage. This is particularly preferable when a sequence of measured pulse signals is detected using a pressure cuff, and the pressure applied to the pressure cuff increases or decreases continuously over the course of the measurement period.
[0030] The sequence of measured pulse signals is detected over a configurable time period. Preferably, the time period is configured to cover a predetermined number of subsequent respiratory or ventilatory cycles of the patient, where the predetermined number is any number from 1 to 10. For example, the time period is configured to be, for example, between 10 seconds and 4 minutes, preferably between 30 seconds and 2 minutes, and more preferably about 1 minute. The patient's pulse rate depends on various factors such as age and stress. An adult heart typically beats 50 to 90 times per minute. Therefore, a relatively large amount of pulse rate variation caused by the heartbeat can be detected within this detection time period. Such a method may be advantageous in terms of data quality, which forms the basis for calculating indices representing fluid reactivity parameters. However, when applying a non-invasive pressure measurement method using a high-fidelity pressure cuff, it is desirable that the detection period not exceed 3 minutes to avoid adverse effects caused by blood flow disturbances due to the pressure of the pressure cuff.
[0031] Patient respiration has a detectable effect on pulse variability. The pressure fluctuations measured are primarily derived from cardiac function, i.e., periodic contractions and relaxations, but other influential factors must also be considered. Thus, the two functions are superimposed; namely, the high-frequency fluctuations caused by cardiac function are superimposed on the low-frequency fluctuations caused by patient respiration or ventilation. In particular, such low-frequency fluctuations caused by patient respiration are detectable not only in patients undergoing mechanical ventilation but also in patients undergoing non-mechanical ventilation who are breathing spontaneously. With respect to pressure fluctuations, the effect of spontaneous respiration is somewhat similar to that of mechanical ventilation, but in reality, the two effects are not the same because in mechanical ventilation, air is forced into the lungs from the outside at high pressure during inspiration, while in spontaneous respiration, air is drawn into the lungs at low pressure during inspiration. Regardless of these phenomenological distinctions, in one embodiment, the device can be applied equally to patients undergoing mechanical ventilation and to non-ventilated patients who are breathing spontaneously, provided that a significant cardiopulmonary interaction occurs due to the ventilation or respiration-induced behavior.
[0032] The measured pulse signal is expressed as a function of time, or alternatively, as a function of the tightening pressure applied by the pressure cuff, where the pressure applied to the pressure cuff continuously increases or decreases over the course of the measurement period.
[0033] The envelope signal curve corresponding to the first data value is determined based on the sequence of measured pulse signals. Typically, the envelope or function of a rapidly changing signal is considered to be a smooth curve that outlines the extrema in the amplitude of the rapidly changing signal. For example, the envelope or function is determined by simply connecting the maximum or minimum values of the rapidly changing signal. However, preferably, the envelope of the pressure signal, i.e., the envelope signal curve, is determined by continuously determining the distance dimension of the measured pulse signal from its mean, and then preferably by applying a low-pass filter to the distance dimension. The low-pass filter preferably has a cutoff frequency below the patient's pulse rate. For example, portions of the oscillation curve of the measured pulse signal below the mean are preferably folded into portions above it. The resulting curve is then preferably flattened by using a low-pass filter with a cutoff frequency below the patient's pulse rate. The resulting curve may be flattened such that the region below the flattened curve is no different from the region below the unflattened curve. Optionally, the flattened curve is further multiplied by a predetermined value. If the flattened curve is multiplied by, for example, the square root of 2, the resulting envelope signal curve effectively represents the upper limit level of the amplitude of the measured pulse signal.
[0034] The distance dimension is also defined as the squared signal, or the extreme value within a certain region of, for example, about one pulse width or any other metric function. More generally, the distance dimension is also defined as, for example, the nth power of the absolute value of the difference between the signal and its mean value, where n is 1 or any other number. When the absolute value of the difference is used as the distance dimension, the results correlate well with SVV and PPV. When the square of the absolute value of the difference is used as the distance dimension, the results also correlate well with SVV and PPV. When SVV is determined, it is preferable to use the square of the absolute value of the difference as the distance dimension, and when PPV is determined, it is preferable to use the absolute value of the difference as the distance dimension. Increasing n further increases the correlation. When the maximum metric, i.e., an infinite n, is used as the distance dimension, the results correlate very well with PPV. For example, the maximum metric is achieved by finding the value obtained by subtracting the minimum value from the maximum value of the difference within a moving window equal to the duration of the heartbeat.
[0035] The mean value of the measured pulse signal used as the basis for calculating distance dimensions can be determined as a moving average over the duration of a single pulse cycle in a patient. Given a numerical series, which is the measured pulse signal in this example, and a fixed subset size, which is the duration of a single pulse cycle in a patient in this example, the first element of the moving average is obtained by taking the average of the initial fixed subset of the numerical series. The subset is then corrected by a "forward shift," i.e., the first number in the series is removed and the next number following the original subset in the series is included. This creates a new, averaged subset of numbers. This process is repeated over the entire data series.
[0036] In some cases, it is advantageous to apply a window function to the measured pulse signal before determining the envelope signal curve in order to obtain a well-fittable envelope signal curve. However, since the pulse signal is measured using a non-invasive pressure measurement method employing a high-fidelity pressure cuff, the application of a window function is not necessary. Under these circumstances, the measured pulse signal usually already exhibits a bell shape and is therefore well-fittable on its own to a function prototype that also exhibits a bell shape. If a window function is applied, it is preferable that the window function be a non-negative smooth bell curve, such as the Cauchy-Lorentz function.
[0037] The determination of the fitted envelope signal function represents the idealized curve transition of the envelope signal curve in question, excluding pulse rate variability caused by cardiopulmonary interactions induced by ventilation or respiration.
[0038] The difference between the envelope signal curve and the fitted envelope signal function reflects modulation due to the patient's respiration or ventilation. Therefore, a respiratory pulse variability signal corresponding to pulse variability caused by the patient's respiration is determined. Preferably, the respiratory pulse variability signal is determined so that it oscillates around its mean value. Preferably, the region defined by the lower part of the respiratory pulse variability signal curve, i.e., the region below the mean value, substantially corresponds to the region defined by the upper part of the respiratory pulse variability signal curve, i.e., the region above the mean value.
[0039] The absolute respiratory pulse variability curve corresponds to the envelope breathing curve, in which case the fitted breathing function is also considered the fitted envelope breathing function. The fitted breathing function represents the idealized curve transition of the envelope breathing curve over multiple respiratory cycles. In this case, the determination of the envelope breathing curve corresponding to the second data value is based on the previously determined respiratory pulse variability signal. The envelope breathing curve is preferably calculated by continuously determining the distance dimension of the respiratory pulse variability signal from its mean, and then preferably by applying a low-pass filter to the distance dimension. The low-pass filter preferably has a cutoff frequency below the patient's breathing frequency. In other words, portions of the oscillation curve of the respiratory pulse variability signal below the mean are preferably folded into portions above it. The resulting curve is then preferably flattened by using a low-pass filter with a cutoff frequency below the patient's breathing frequency. The curve may be flattened such that the region below the flattened curve is unchanged compared to the region below the unflattened curve. Optionally, the flattened curve may be further multiplied by a predetermined value. Multiplying the flattened curve by, for example, the square root of 2, the resulting envelope respiratory curve effectively represents the upper limit level in the amplitude of the respiratory pulse variability signal. The mean value of the respiratory pulse variability signal used as the basis for distance dimension calculation is preferably determined as a moving average over the duration of a single respiratory cycle in the patient.
[0040] Preferably, the absolute respiratory pulse variability curve does not correspond to the envelope breathing curve, but to a different curve that shows the absolute value of the respiratory pulse variability signal. That is, it is preferable that the fitted breathing function is determined without first determining the envelope breathing curve of the respiratory pulse variability signal that was previously determined, and then determining the fitted breathing function based on that envelope breathing curve. Instead, it is preferable to determine the fitted breathing function by fitting a second function prototype to data values corresponding to a different type of absolute respiratory pulse variability signal, i.e., second data values, i.e., data values derived in a different way from the respiratory pulse variability signal. Nevertheless, such different types of preprocessing still include one or more of the steps described above with respect to determining the envelope breathing curve from the respiratory pulse variability signal. In one embodiment, the absolute respiratory pulse variability curve corresponding to the second data value is determined by considering the absolute value of the difference between the respiratory pulse variability signal and its mean. In this case, these absolute values correspond to the second data value. Therefore, specifically, the fitted respiratory function is determined by considering the absolute value of the difference between the respiratory pulse variability signal and its mean value, thereby folding the portion of the respiratory pulse variability signal below its mean value into the portion above it, and fitting the second function prototype to the value thus obtained.
[0041] When the fitted breathing function is determined by first determining the envelope breathing curve of the respiratory pulse variability signal that was previously determined, and then determining the fitted breathing function based on that envelope breathing curve, it is preferable that the envelope breathing curve is calculated in the same way as the envelope signal curve, but associated with the respiratory pulse variability signal instead of the measured pulse signal, in order to avoid bias from the calculation steps involved. More generally, it is preferable that the fitted breathing function is calculated in the same way as the fitted envelope signal function, based on the data value to which the fitted breathing function is fitted, i.e., a second data value, i.e., an envelope breathing curve or a data value derived in a different way from the respiratory pulse variability signal, except that the data value to which the fitted breathing function is fitted is different, i.e., the data value to which the fitted breathing function is fitted is associated with the respiratory pulse variability signal instead of the measured pulse signal. That is, if the Cauchy-Lorentz function is used as a function prototype for determining the fitted envelope signal function, it is preferable that the Cauchy-Lorentz function is also used as a function prototype for determining the fitted breathing function. However, for example, how the number of data values is reduced for the entire fit before fitting is determined based on the characteristics of the data values, i.e., whether the data values correspond to an envelope or to a curve derived from the oscillatory signal by considering the absolute value of the oscillatory signal. Thus, for example, how the number of data values is reduced before fitting differs between the first data values and the second data values, i.e., between the first fit and the second fit. In one embodiment, the index determination unit is configured to determine an absolute respiratory pulse variability curve by considering the absolute value of the difference between the respiratory pulse variability signal and its mean, and to reduce the number of second data values by determining one second data value for each maximal of the absolute respiratory pulse variability curve.
[0042] An index representing the fluid reactivity parameter can be determined based on the fitted envelope signal function and the fitted respiratory function. Preferably, the index representing the fluid reactivity parameter is determined based on at least one parameter of the fitted envelope signal function and at least one parameter of the fitted respiratory function. For example, an index representing a patient's volume reactivity can be determined based on the ratio between the maximum value of the fitted envelope signal function and the maximum value of the fitted respiratory function. Such a ratio, or a function of this ratio, in particular its reciprocal, represents a suitable index representing the patient's volume reactivity.
[0043] In one embodiment, further data values are determined by determining a further absolute respiratory pulse variability curve of the respiratory pulse variability signal; in further fitting, a further function prototype that depends on further fitting parameters modified during fitting is fitted to the further data values so that a further fitted respiratory function is obtained; and an index representing the fluid reactivity parameter is determined further based on the further fitted respiratory function. Furthermore, further fitting is performed by a) reducing the number of further data values for the entire further fitting before further fitting; b) determining initial values of further fitting parameters based on the characteristics of the further data values related to the characteristics of the further function prototype and / or based on the values of further fitting parameters known from previous further fittings; and / or c) further fitting is performed in several stages, with the number of further data values used for further fitting increasing in each stage, so that in the current stage, the values of the further fitting parameters determined in the previous stage are used as initial values of the further fitting parameters in the current stage. Preferably, the further function prototype is the same as the second function prototype. However, preferably, the further absolute respiratory pulse variability curve is different from the second absolute respiratory pulse variability curve. It is also possible to perform several such further modifications, namely two, three, or more further modifications, in which case the index representing the fluid reactivity parameter may be determined based on several further modified breathing functions.
[0044] In a further aspect of the present invention, a method is provided for determining an index representing a fluid reactivity parameter, the method being - A step of providing a sequence of measured pulse signals of a patient over a time period corresponding to multiple subsequent respiratory cycles of the patient, wherein the sequence of measured pulse signals is detected by a non-invasive pulse measurement method. - A step of performing a determination procedure adapted to determine an index representing a fluid reactivity parameter based on a given sequence of pulse signals by an index determination unit, the determination procedure comprising determining data values based on a given sequence of pulse signals and fitting a given function prototype, which depends on fitting parameters modified during fitting, to the determined data values, wherein the index determination unit is configured to a) reduce the number of data values for the entire fitting before fitting, b) determine initial values for fitting parameters based on the characteristics of the data values related to the characteristics of the function prototype and / or based on the values of fitting parameters known from previous fittings, and / or c) perform fitting in several stages, with the number of data values used for fitting increasing in each stage, and in the current stage, the values of the fitting parameters determined in the previous stage are used as initial values for the fitting parameters in the current stage. It holds.
[0045] In another aspect of the present invention, a computer program is provided for determining an index representing a fluid reactivity parameter, the computer program including program code means for causing the apparatus according to any one of claims 1 to 12 to perform the steps of the method according to claim 13.
[0046] It should be understood that the apparatus of claim 1, the method of claim 13, and the computer program of claim 14 have preferred embodiments similar to and / or identical to those described in the dependent claims.
[0047] It should be understood that preferred embodiments of the present invention may also be dependent claims or any combination of the above embodiments and their respective independent claims.
[0048] These and other aspects of the present invention will become clear and clarified by referring to the embodiments described below. [Brief explanation of the drawing]
[0049] [Figure 1] This figure schematically and illustratively shows a device for determining an index representing fluid reactivity parameters. [Figure 2] This figure schematically and illustratively shows several curves that depend on the applied pressure using a pressure cuff. [Figure 3] This is a curved diagram illustrating several decision-making steps in a schematic and illustrative manner. [Figure 4] This is a curved diagram illustrating several decision-making steps in a schematic and illustrative manner. [Figure 5] This is a curved diagram illustrating several decision-making steps in a schematic and illustrative manner. [Figure 6] This is a curved diagram illustrating several decision-making steps in a schematic and illustrative manner. [Figure 7] This is a curved diagram illustrating several decision-making steps in a schematic and illustrative manner. [Figure 8] This is a curved diagram illustrating several decision-making steps in a schematic and illustrative manner. [Figure 9] This is a curved diagram illustrating several decision-making steps in a schematic and illustrative manner. [Figure 10] This is a curved diagram illustrating several decision-making steps in a schematic and illustrative manner. [Figure 11] This is a curved diagram illustrating several decision-making steps in a schematic and illustrative manner. [Figure 12] This diagram schematically and illustratively illustrates the steps of a determination procedure for determining an index representing a fluid reactivity parameter. [Figure 13] This flowchart exemplifies a method for determining an index representing fluid reactivity parameters. [Modes for carrying out the invention]
[0050] Figure 1 schematically and illustratively shows an apparatus 100 for determining an index representing a fluid reactivity parameter. Apparatus 100 comprises a pulse signal providing unit 101 configured to provide a sequence of patient measured pulse signals over a time period corresponding to multiple subsequent respiratory cycles of the patient. The sequence of measured pulse signals is detected by a non-invasive pulse measurement method. Apparatus 100 further comprises an index determination unit 102 configured to perform a determination procedure adapted to determine an index representing a fluid reactivity parameter based on the provided sequence of pulse signals. The determination procedure includes determining data values s0, S0, S0', S0'' based on the provided sequence of pulse signals and fitting a provided function prototype, which depends on fitting parameters modified during fitting, to the determined data values s0, S0, S0', S0''. The index determination unit 102 is configured such that a) before fitting, the number of data values s0, S0, S0', S0'' for the entire fitting is reduced; b) initial values of the fitting parameters are determined based on the characteristics of the data values s0, S0, S0', S0'' related to the characteristics of the function prototype and / or based on the values of the fitting parameters known from previous fittings; and / or c) the fitting is performed in several stages, the number of data values used for fitting increases in each stage, and in the current stage, the values of the fitting parameters determined in the previous stage are used as the initial values of the fitting parameters in the current stage. As shown in Figure 1, the device 100 includes a connection between the pulse signal providing unit 101 and the index determination unit 102, and the pulse signal providing unit 101 can provide the index determination unit 102 with a sequence of measured pulse signals via this connection.
[0051] Figure 2 schematically and illustratively illustrates how an index representing a patient's volume responsiveness can be determined based on a given sequence of measured pulse signals detected using a pressure cuff, where the pressure applied to the pressure cuff increases continuously over the course of the measurement period. The horizontal axis of the graph shown in Figure 2 represents the pressure applied to the compression cuff and can therefore also be viewed as a time axis. The vertical axis represents the pressure measured by the pressure sensor of the compression cuff. The measured pulse signal is represented as a pulse signal p0 that oscillates around its mean value. The data value s0 is determined by determining the envelope signal curve of the pulse signal p0 based on the pulse signal p0. The data value s0 is considered to be either the envelope signal curve or the value of the envelope signal curve. In this example, the function prototype provided for fitting is:
number
number
[0052] Finally, an index VR representing the patient's volume responsiveness is the parameter g of the fitted respiration function g0 amp with the value g0 amp and the parameter f of the fitted envelope signal function f0 amp with the value f0 amp and the ratio between them, that is,
Number
Number
[0053] Referring to Figure 2, the decision procedure described above is an example of a decision procedure used to determine fluid reactivity parameters, i.e., indicators representing the volume reactivity of a patient. This decision procedure involves a fitting algorithm and is therefore computationally intensive and can be time-consuming. This problem is particularly likely to occur in the procedure described above because it involves two fittings. Figures 3 to 10 show how this problem can be mitigated using apparatus 100.
[0054] Figure 3 schematically and illustratively shows a graph in which the axes of the graph correspond to the axes of the graph shown in Figure 2, similar to Figures 4 through 11. The graph in Figure 3 includes data values S0 determined based on a given sequence of pulse signals and a fitted function F0 corresponding to the fitting of a given function prototype to the data values S0. In the illustrated embodiment, the data values S0 are determined based on a function derived from a sequence of measured pulse signals, not the values of the measured pulse signals, i.e., sample values, and this function is considered to correspond to an envelope signal curve such as the envelope signal curve s0 in Figure 2. The form of the fitted function F0 in Figure 3 corresponds to the form of the provided function prototype, and in this case, the function prototype is the Cauchy-Lorentz function. The fitted function F0 is fitted to the data values S0, and the values of the fitted parameters are determined during the fitting process, i.e., modified starting from their respective initial values. All of the data values S0 are taken into consideration in order to fit the fitted function F0 to the data values S0.
[0055] Figure 4 corresponds to Figure 3, but instead of the complete set of data values S0 and the corresponding fitting function F0, a reduced number of data values SR and the corresponding fitting function FR fitted to the reduced number of data values are plotted. The data values SR were obtained by selecting every 1 / 1000th of the data values S0. This reduced the number of data values S0 used for fitting to 0.1 percent. The fitting was performed using the same function prototype, namely the Cauchy-Lorentz function, that was used to arrive at the graph in Figure 3.
[0056] In Figure 5, related to Figures 3 and 4, the data values are not plotted; only the fitted functions F0 and FR obtained as a result of fitting to the entire set of data values S0 and the reduced set of data values SR, respectively, are plotted. From Figure 5, it can be seen that even though the number of data values used to arrive at the reduced fitted function FR was reduced by 99.9 percent compared to the total number of data values S0, the reduced fitted function FR deviates only by a relatively small amount from the fitted function F0. However, the time required to fit the function prototype to the reduced number of data values was significantly shorter than the time required to fit the function prototype to the entire number of data values S0. Since the fluid reactivity parameter, i.e., an index representing the patient's volume reactivity, is determined from the final fitted parameter obtained as a result of the fitting, such as the maximum value of the fitted function in equation (3) above, the index can be determined much more quickly based on the fitting to the reduced number of data values that yield the fitted function FR. Furthermore, since the fitted function FR deviates only relatively slightly from the fitted function F0, the rapid determination of the index does not necessarily negatively affect the accuracy of the determined index.
[0057] Reducing the number of data values to be considered for fitting is not the only way to enable faster fitting and, therefore, to enable faster determination of an index representing the patient's volume or fluid responsiveness based on the fitted function. Specifically, as an addition or alternative, the values of one or more fitting parameters, which the provided function prototype depends on and whose values will be determined by fitting, can be estimated before fitting, and the estimated values are used as initial values for the fitting parameters for fitting. This is schematically and illustratively illustrated by Figures 6 and 7.
[0058] Figures 6 and 7 relate to embodiments in which the function prototype is also the Cauchy-Lorentz function. As already explained in relation to equations (1) and (2) above, the Cauchy-Lorentz function has a first fitting parameter that indicates the height of its maximum value and a second fitting parameter that indicates the position of its maximum value.
[0059] Figure 6 shows an embodiment in which the index determination unit 102 is configured such that initial values of the fitting parameters are determined based on the characteristics of the data value S0 related to the characteristics of the function prototype. In the particular embodiment shown, the index determination unit 102 is configured to determine the height and position of the maximum value of the data value S0, and to determine these values as the initial values of the first fitting parameter and the second fitting parameter, respectively. The initially assumed fitting function, i.e., the fitting function assumed at the start of fitting, is thus sufficiently adapted to the data value S0 so that the amount of fitting to be performed can be reduced.
[0060] In general, the amount of fitting required can be reduced by considering only the fitting data values that capture the characteristics of the entire set of data values relevant to the function prototype. Thus, not only can the initially assumed fitting function be adapted to the data values to which that fitting function is adapted, but a subset of the data values to which that fitting function is adapted can also be adapted to the type of fitting function. This is schematically and illustratively shown in Figure 7.
[0061] Figure 7 shows a graph plotting the data value S0', which was determined based on a sequence of measured pulse signals using different types of preprocessing, compared to the data value S0. The data value S0' is called the second data value and is considered to be derived from the signal corresponding to the respiratory pulse variability signal r0 shown in Figure 2, and therefore is considered to be derived from the data value corresponding to the data value S0, which is called the first data value. In this case, the data value S0' is derived by considering the absolute difference between the signal corresponding to the respiratory pulse variability signal r0 shown in Figure 2 and its mean value, that is, by folding the portion of the corresponding signal that is below its mean value into the portion that is above its mean value, without further determining the envelope for these signals. Thus, the data value S0' corresponds to an absolute respiratory pulse variability curve, not an envelope respiratory curve. The data value S0' includes a series of peaks whose amplitude, i.e., height, changes. In this embodiment, the target characteristic of the data value S0' is still the transition of the peak amplitude, i.e., not the shape of any one of the peaks. Therefore, the number of data values considered for fitting can be reduced to the data value that captures this target characteristic, i.e., the data value that captures the transition of the peak amplitude, i.e., the height.
[0062] In the illustrated example, the data values considered for fitting are reduced to the data values corresponding to the local maximums. In Figure 7, the resulting reduced set of data values is annotated with SR'. The fitting function G0' is obtained as a result of fitting the entire set of data values S0', and the fitting function GR' is obtained as a result of fitting the reduced set of data values SR'. In this example, the deviation of the fitting function GR' from the fitting function G0' is significant because the deviation of the original data values S0' is somewhat large. Nevertheless, GR' allows for a good prediction of the fitting parameters corresponding to G0'. Specifically, the position of the maximum value of GR' is close to the position of the maximum value of G0', and the full width at half maximum of GR' is a good estimate of the full width at half maximum of G0'. Furthermore, it is preferable to apply a predefined scaling factor to the data values S0' or the fitting function GR' before fitting. In this way, by considering only the local maximum SR' of S0' in the fitting, it can be taken into account that the height of the fitted maximum is always greater than when fitting all data values S0'.
[0063] Further additional or alternative methods for reducing the time required for fitting, and therefore for determining an index representing the fluid reactivity parameter based on the fitting function obtained as a result of fitting, are schematically and illustratively shown in Figures 8 to 11.
[0064] Figures 8 to 11 relate to one embodiment in which the fitting is performed in several stages, the number of data values used for fitting increasing with each stage, and in the current stage, the value of the fitting parameter determined in the previous stage is used as the initial value of the fitting parameter in the current stage. Specifically, in the embodiment shown by Figures 8 to 11, the index determination unit 102 is configured so that the fitting is performed in three stages, in the first stage a first percentage of the data values is used, in the next second stage a higher second percentage of the data values is used, and in the next third stage 100 percent of the data values is used.
[0065] The graph in Figure 8 plots the data value SR1'', which corresponds to the first proportion of the entire set of data values S0'', along with the fitting function FR1'' that is fitted to the data value SR1''. In the illustrated embodiment, the original data value S0'' corresponds to the original data value S0 from the embodiment described with respect to Figures 3 to 5. Furthermore, the reduced set of data values SR1'' corresponds to the reduced set of data values from the embodiment described with respect to Figures 3 and 5. That is, in this case, the first proportion is 0.1.
[0066] In the graph of Figure 9, the set of data values SR2'', corresponding to a higher second proportion of the entire set of data values S0'', is plotted along with the fitted function FR2'' that is adapted to the data values SR2''. In this example, the second proportion is selected to be 1 percent.
[0067] Figures 8 and 9 show the first and second stages of the overall adaptation process, respectively, while the third and final stages of the overall adaptation process in this embodiment are shown in Figure 10.
[0068] In the graph of Figure 10, the data value SR3'' corresponding to the entire set of data values S0'' is plotted along with the fitting function FR3'' that fits the data value SR3''=S0''.
[0069] In the illustrated embodiment, part of the fitting process performed in the second stage included fitting the provided function prototype to the data value SR2'' by using the final values obtained as a result of the first stage, i.e., the fitting parameters describing the fitting function FR1'', as initial values for one or more of the fitting parameters of the function prototype. Therefore, the time required for fitting in the second stage can be performed in less time than the time required if the function prototype were fitted to the data value SR2'' based on general initial values of the fitting parameters. Similarly, the fitting performed in the third stage includes fitting the function prototype to the data value SR3''=S0'' by starting with the values of one or more of the fitting parameters corresponding to the final values of the fitting parameters obtained as a result of fitting in the second stage, i.e., based on the fitting function FR2'' as the initial fitting function. Therefore, fitting in the third stage also required less time than the time required to fit the function prototype to the data value SR3'' corresponding to the entire set of data values S0'' based on general initial values of the fitting parameters.
[0070] The graph in Figure 11 plots the final fitting functions FR1'', FR2'', and FR3'', obtained as a result of the three stages of fitting, for comparison. Fitting function FR1'', still deviates from fitting function FR3'', albeit by a relatively small amount, while the deviation of fitting function FR2'', from fitting function FR3'', is not noteworthy. Therefore, fitting function FR2'', obtained as a result of the second stage, can be considered an exact fit for the entire set of data values S0'',
[0071] For these reasons, in some embodiments, the index determination unit 102 is configured so that the fitting is performed in several stages, and after the fitting is performed in one stage, it is determined whether the fitted fitting parameters meet the termination criteria, and if they do, the fitting is terminated. For example, in the embodiments shown in Figures 8 to 11, after the fitting is performed in the second stage, one or more of the fitting parameters have already met the termination criteria, and as a result, the entire fitting is terminated. In some embodiments, the fitting is terminated only if two or more or all of the fitting parameters meet their respective termination criteria.
[0072] The termination criterion is used to end the fitting process if the current fitting function is already sufficiently accurate. However, the termination criterion is also used to end the fitting process if, for example, after one stage of a multi-stage fitting process, it is concluded that the fitting result does not lead to a valid fitting function at all.
[0073] For example, when fitting a bell-shaped Cauchy-Lorentz function, the fitting parameter corresponding to the position of the function's maximum value is expected to be within a specific range for a given set of data values. The range is defined by one or both of an upper and lower limit, the lower limit corresponding to the position of the first data value, and the upper limit corresponding to a multiple of the signal length, which is, for example, any value from 1 to 5.
[0074] For example, the fitting is performed in three stages, in the first stage 1 / 250 of the data value is used, and in the second stage the data value
number
[0075] Figure 12 schematically and illustratively shows the determination procedure 1100 by combining the embodiments described above.
[0076] In the first step 1101, data values s0, S0, S0', and S0'' are determined based on the sequence of measured pulse signals provided.
[0077] In the second step 1102, initial values are determined for each of the one or more fitting parameters of the function prototype to be fitted with the data values s0, S0, S0', S0''. The initial values are determined based on the characteristics of the data values s0, S0, S0', S0'' that correspond to the characteristics of the function prototype that each fitting parameter represents. For example, if the function prototype is a bell function, and the first parameter represents the height of the maximum value of the bell function, and the second fitting parameter represents the position of the maximum value of the bell function, then in step 1102, the height of the maximum value of the data values s0, S0, S0', S0'' and the position of the maximum value of the data values s0, S0, S0', S0'' may be determined as the initial values for the first and second fitting parameters.
[0078] In the next step, 1103, a multi-stage fitting is performed using the initial values determined in step 1102 as the initial values for the fitting parameters in the first stage. The multi-stage fitting performed in step 1103 may be carried out as described, for example, with reference to Figures 7 to 10.
[0079] In the subsequent step 1104, it is determined whether an additional set of data values needs to be considered for further fitting, and the additional set of data values is determined based on a further sequence of measured pulse signals. This determination is made separately for the first set of data values determined for the first fitting and for the second set of data values determined for the second fitting performed on the same sequence of measured pulse signals. The determination of whether further fitting is necessary is made according to termination criteria similar to those that may be used to determine whether the multi-stage fitting should be terminated after any non-final stage of the multi-stage fitting.
[0080] If it is determined in step 1104 that further fitting is needed, in step 1105, initial fitting parameters for further fitting are estimated based on the final fitting parameters obtained as a result of the previous fitting performed in step 1103. The multi-stage fitting performed in step 1103 is then repeated for a further set of data values. The exception is that, in the first stage of the multi-stage fitting performed in the repeated step 1105, values that describe the characteristics of the further data values corresponding to the characteristics of the function prototype represented by the fitting parameters are not used as initial values for the fitting parameters; instead, the final values of the fitting parameters obtained as a result of the multi-stage fitting for the previous set of data values are used. Thus, in the embodiment shown in Figure 12, the index determination unit 102 determines the initial values of the fitting parameters based on the values of the fitting parameters known from the previous fitting.
[0081] The cycle defined by steps 1103, 1104, and 1105 is repeated as long as it is determined that further conformance is required in step 1104 or its iteration, i.e., unless the termination criteria are met, and not all of the provided sets of data values are used for conformance. For example, if the pulse signal providing unit provides four sequences of measured pulse signals or a single sequence of measured pulse signals corresponding to four inflations of the pressure cuff, and if four sets of data values are determined based on the provided sequences or separate sequences, with each set corresponding to one sequence or separate inflations, the cycle defined by steps 1103, 1104, and 1105 is performed up to four times.
[0082] In other words, whenever it is determined that no further fitting is needed in step 1104 or its iteration, step 1106 proceeds. In step 1106, an index representing the fluid reactivity parameter is determined based on the final fitting function obtained as a result of the previous, i.e., last fitting.
[0083] For example, if the apparatus 100 configured as described above with reference to Figure 12 is used for a set of 540 representative records to calculate PPV, SVV, or SPV, and for each record, 11 individual fits are calculated for two different types of input samples, and a bell-shaped Cauchy-Lorentz function is used as the function prototype, the following performance optimizations can be observed.
[0084] In the unoptimized case, i.e., when all data values are used for fitting and fitting is performed in a single stage, the total time required to fit all records is 64 seconds, and the fitting time per record is 0.119 seconds. For each record, an average of 49 fitting iterations can be observed for all approximately 13,000 input samples.
[0085] In the optimized example, for each of the two types of input samples, the final fit parameters of the previous fit were used as the initial fit parameters for the intended fit. In this example, the time required to fit all records was 24 seconds, with an average fitting time of 0.044 seconds per record, which is approximately 38 percent of the time required in the unoptimized example. For each record, we can observe an average of 38 fit iterations for the approximately 50 reduced input samples, an average of 23 fit iterations for the approximately 800 reduced input samples, and an average of 11 fit iterations for all approximately 13,000 input samples.
[0086] Figure 13 schematically and illustratively illustrates one embodiment of Method 200 for determining an index representing a fluid reactivity parameter. In a first part 201, Method 200 has the step of having a pulse signal providing unit 101 provide a sequence of patient measured pulse signals over a time period corresponding to several subsequent respiratory cycles of the patient, the sequence of measured pulse signals being detected by a non-invasive pulse measurement method. The measurement is performed using a pressure cuff, the pressure applied to the pressure cuff is continuously increased or decreased over the course of the measurement period. In a second part 202, Method 200 has the step of having an index determination unit 102 perform a determination procedure adapted to determine an index representing a fluid reactivity parameter based on the provided sequence of pulse signals, the determination procedure including determining data values s0, S0, S0', S0'' based on the provided sequence of pulse signals and fitting a provided function prototype, which depends on fitting parameters modified during fitting, to the determined data values s0, S0, S0', S0''. The index determination unit 102 is configured such that a) before fitting, the number of data values s0, S0, S0', S0'' for the entire fitting is reduced; b) initial values of the fitting parameters are determined based on the characteristics of the data values s0, S0, S0', S0'' related to the characteristics of the function prototype and / or based on the values of the fitting parameters known from previous fittings; and / or c) the fitting is performed in several stages, the number of data values used for fitting increases in each stage, and in the current stage, the values of the fitting parameters determined in the previous stage are used as the initial values of the fitting parameters in the current stage. The determination procedure performed in the second part 202 of Method 200 can be of the type described above, for example, with reference to Figure 12.
[0087] In the embodiments described above, the function prototype was a bell-shaped function, but other function prototypes can also be used. For example, the execution time of the fitting algorithm can be reduced compared to simply applying the fitting algorithm to the entire set of data values by reducing the total number of iterations in the fitting algorithm or the number of data values actually used for fitting.
[0088] If the number of data values actually used in the fitting is reduced, it is preferable that the reduction be carried out in a manner that preserves the main characteristics of the data values with respect to the function prototype.
[0089] For example, the total number of iterations required and therefore the total time required for fitting can be reduced by first predicting the parameters of a function prototype that are close to the final parameters determined by the fitting, preferably. When successive fittings are performed on several sets of data values having similar characteristics, the final fitting parameters of a previous fitting can be used as the initial fitting parameters for the fitting to be performed. In the context of determining an index representing a patient's volume responsiveness, the set of data values corresponds to the patient's subsequent respiratory or ventilatory cycles, and in the more general context of reducing the execution time of the fitting algorithm, different sets of data values may have similar characteristics for reasons other than the cyclical nature of respiration or ventilation.
[0090] The execution time of a general fitting algorithm can also be reduced by iteratively, or stepwise, applying the fitting algorithm, starting with a reduced number of data values and increasing the number of data values in each subsequent step, preferably using the entire original set of data values in the final step. In such multi-step fitting, the preferred number of steps can be found depending on the intended application, for example, by experimentally finding the minimum total execution time. In this way, the optimal combination of the number of data values used in fitting at each step and the number of steps can be found.
[0091] By studying the drawings, disclosures, and appended claims in carrying out the claimed invention, other variations of the disclosed embodiments can be understood and implemented by those skilled in the art.
[0092] In the claims, the word “including” does not preclude other elements or steps, and a singular element does not preclude the existence of multiple elements.
[0093] A single unit or device may perform the functions of several items described in the claims. The mere fact that certain means are described in different dependent claims does not mean that combinations of these means cannot be used advantageously.
[0094] Procedures such as providing a sequence of measured pulse signals, determining data values based on the provided sequence of pulse signals, fitting a provided function prototype to the determined data values, reducing and / or increasing the number of data values, determining initial values for fitting parameters, determination procedures, and any other procedures involved in determining indicators based on the provided sequence of pulse signals, may be performed by any number of other units or devices. These procedures may be implemented as program code means in a computer program and / or as dedicated hardware, particularly as in an embedded system.
[0095] Computer programs are stored / distributed on suitable media such as optical storage media or solid-state media supplied together with or as part of other hardware, but may also be distributed in other forms, such as via the Internet or other wired or wireless communication systems.
[0096] None of the reference numerals in the claims should be construed as limiting the scope.
[0097] The present invention relates to an apparatus for determining an index representing a fluid reactivity parameter, such as a fluid reactivity parameter. The index is determined based on fitting a function prototype to data values determined from measured pulse signals over a subsequent respiratory cycle. The fitting process is accelerated by a) reducing the number of data values before fitting, b) determining initial fit parameter values for the function prototype based on the characteristics of the data values and / or fit parameter values known from previous fittings, and / or c) performing the fitting in several stages, wherein the number of data values used increases in each stage, and the fit parameter values determined in the previous stage are used as the initial fit parameter values in the current stage. This enables a faster determination of the fluid reactivity parameter.
Claims
1. 1. An apparatus for determining an index representative of a fluid reactivity parameter, said apparatus comprising: a pulse signal providing unit for providing a sequence of measured pulse signals of the patient over a period of time corresponding to a number of subsequent respiratory cycles of the patient, the sequence of measured pulse signals having been detected by a non-invasive pulse measurement method, the measured pulse signals being represented as a pulse signal oscillating around an average value of the measured pulse signals; and an index determination unit for executing a determination procedure for determining an index representative of a fluid responsiveness parameter based on the provided sequence of measured pulse signals, the determination procedure comprising: determining a first data value based on the provided sequence of measured pulse signals; and a first fitting of a provided first function prototype to the determined first data value, the first fitting parameter being modified during a first fitting; Equipped with the first data values are determined by determining an envelope signal curve of the measured pulse signal, and in the first fitting, the first function prototype is fitted to the first data values such that a resulting fitted envelope signal function represents an idealized curve progression of the envelope signal curve over multiple respiratory cycles that does not include pulse variation caused by ventilation or respiration induced cardiopulmonary interactions; a respiratory pulse variability signal corresponding to the pulse variability caused by ventilation or respiration induced cardiopulmonary interactions over a number of respiratory cycles is determined, the determined respiratory pulse variability signal corresponds to a difference between the envelope signal curve and the fitted envelope signal function, and a second data value is determined by determining an absolute respiratory pulse variability curve of the respiratory pulse variability signal; In a second fitting, a provided second function prototype is fitted to the second data values, the second function prototype depending on second fitting parameters which are modified during the second fitting, such that a resulting fitted respiratory function shows an idealized progression in the amplitude of the absolute respiratory pulse variability curve over a number of respiratory cycles; determining the index representative of the fluid responsiveness parameter based on the fitted envelope signal function and the fitted respiratory function; In the index determination unit, a) before the first fitting, the number of the first data values for the entire first fitting is reduced; b) initial values of the first fitting parameters are determined based on characteristics of the first data values related to characteristics of the first function prototype and / or based on values of the first fitting parameters known from a previous first fitting; and / or c) the first fitting is performed in several stages, the number of the first data values used for the first fitting increasing from stage to stage, and in a current stage, the values of the first fitting parameters determined in the previous stage are used as initial values of the first fitting parameters in the current stage.
2. 2. The apparatus of claim 1, wherein the index determination unit reduces the number of the first data values such that a reduced number of the first data values satisfies a Shannon-Nyquist criterion with respect to a bandwidth of the provided first function prototype and / or with respect to the first data values themselves.
3. 3. The apparatus according to claim 1 or 2, wherein the index determination unit reduces the number of the first data values by determining one data value at a predetermined frequency and / or by downsampling.
4. The apparatus according to claim 1 , wherein the index determination unit is adapted to low-pass filter the first data values before reducing the number of the first data values.
5. The apparatus of claim 1 , wherein the first function prototype is a bell-shaped function.
6. 6. The apparatus of claim 5, wherein the bell-shaped function has a first fitting parameter indicative of a height of a maximum of the bell-shaped function and a second fitting parameter indicative of a position of a maximum of the bell-shaped function, and the index determination unit determines a height of a maximum of the first data value and a position of the maximum of the first data value as an initial value of the first fitting parameter and an initial value of the second fitting parameter.
7. 7. The apparatus according to claim 1, wherein in the index determination unit, the fitting is performed in three stages, in a first stage a first percentage of the first data values is used, in a second stage a higher second percentage of the first data values is used, and in a third stage 100 percent of the first data values are used.
8. 8. The apparatus according to claim 1, wherein in the index determination unit, the first adaptation is performed in several stages, and after the first adaptation is performed in a stage, it is determined whether the adapted first adaptation parameters meet a stopping criterion, and if so, the first adaptation is terminated.
9. The apparatus according to claim 8, wherein the index determination unit uses as the stopping criterion a deviation between a) the fitted value of the first fitting parameter obtained as a result of each stage and b) an expected value of the first fitting parameter.
10. a) prior to the second adaptation, the number of the second data values for the entire second adaptation is reduced; b) initial values of the second adaptation parameters are determined based on characteristics of the second data values related to characteristics of the second function prototype and / or based on values of the second adaptation parameters known from a previous second adaptation; and / or c) the second adaptation is performed in several stages, the number of the second data values used for the second adaptation increasing from stage to stage, and in a current stage, the value of the second adaptation parameter determined in a previous stage is used as the initial value of the second adaptation parameter in the current stage.
10. Apparatus according to any one of claims 1 to 9.
11. 11. The apparatus of claim 10, wherein the index determination unit determines the absolute respiratory pulse variability curve by considering the absolute value of each difference between the respiratory pulse variability signal and an average of the respiratory pulse variability signal, and reduces the number of the second data values by determining one second data value for each local maximum of the absolute respiratory pulse variability curve.
12. 12. The apparatus of claim 1, wherein the non-invasive pulse measurement method uses a pressure cuff that brings a pressure sensor for measuring a pulse signal into direct contact with the patient's tissue such that the pressure sensor is hydraulically coupled to blood pulsations through the patient's tissue.
13. 1. A method for determining an index representative of a fluid reactivity parameter, the method comprising: providing, by a pulse signal providing unit, a sequence of measured pulse signals of the patient over a time period corresponding to a number of subsequent respiratory cycles of the patient, the sequence of measured pulse signals having been detected by a non-invasive pulse measurement method, the measured pulse signals being represented as a pulse signal oscillating around an average value of the measured pulse signals; executing, by an index determination unit, a determination procedure for determining an index representative of a fluid responsiveness parameter based on the provided sequence of measured pulse signals, said determination procedure comprising: determining a first data value based on the provided sequence of measured pulse signals; and a first fitting of a provided first function prototype to the determined first data value, the first fitting parameter being modified during a first fitting. having the first data values are determined by determining an envelope signal curve of the measured pulse signal, and in the first fitting, the first function prototype is fitted to the first data values such that a resulting fitted envelope signal function represents an idealized curve progression of the envelope signal curve over multiple respiratory cycles that does not include pulse variation caused by ventilation or respiration induced cardiopulmonary interactions; a respiratory pulse variability signal corresponding to the pulse variability caused by ventilation or respiration induced cardiopulmonary interactions over a number of respiratory cycles is determined, the determined respiratory pulse variability signal corresponds to a difference between the envelope signal curve and the fitted envelope signal function, and a second data value is determined by determining an absolute respiratory pulse variability curve of the respiratory pulse variability signal; In a second fitting, a provided second function prototype is fitted to the second data values, the second function prototype depending on second fitting parameters which are modified during the second fitting, such that a resulting fitted respiratory function shows an idealized progression in the amplitude of the absolute respiratory pulse variability curve over a number of respiratory cycles; determining the index representative of the fluid responsiveness parameter based on the fitted envelope signal function and the fitted respiratory function; A method in which, in the indicator determination unit, a) before the first fitting, the number of the first data values for the entire first fitting is reduced; b) initial values of the first fitting parameters are determined based on values of the first fitting parameters known from a previous first fitting; and / or c) the first fitting is performed in several stages, the number of the first data values used for the first fitting increasing in each stage, and in a current stage, the value of the first fitting parameter determined in the previous stage is used as the initial value of the first fitting parameter in the current stage.
14. 14. A computer program for determining an index representative of a fluid reactivity parameter, said computer program comprising program code means for causing an apparatus according to any one of claims 1 to 12 to carry out the steps of the method according to claim 13.
