System for continuous estimation and display of cardiac ejection fraction and end diastolic volume

【0001】

(Background of the Invention) (Field of Invention) The present invention relates to in vivo determination and representation of cardiac ejection fraction and / or final expansion volume estimates.

【0002】

(Explanation of Related Techniques) Information about the output of the patient's heart is very useful for surgical teams operating the patient, or for doctors seeking to diagnose the disease or monitor the patient's condition. Therefore, few hospitals lack some conventional equipment to monitor cardiac output.

【0003】

The usual method of determining cardiac output is to install a catheter flow measuring device, then pass the catheter through the patient and operate it so that the device is in or near the patient's heart. It is in. Some of such devices inject either a bolus or heat in an upstream position, such as the right atrium, and determine the flow rate based on the properties of the injected material or the energy in the downstream position, such as the pulmonary artery.

【0004】

For example, U.S. Pat. No. 4,236,527 (Newbower et al., December 2, 1980) and U.S. Pat. No. 4,507,974 (Yelderman, April 2, 1985) use heat as an indicator for cardiac output. Describe the system for measurement. In such a thermal system, the balloon catheter is generally positioned proximal to the branch of the pulmonary artery via the right atrium and right ventricle. The catheter comprises a resistance heating element located in the atrium and / or ventricle and a thermistor located in an artery. Cardiac output is then calculated as a function of the detected downstream temperature profile.

【0005】

US Pat. No. 5,146,414 (McKown et al., September 8, 1992) provides a transfer function for channels (from the region where indicators such as heat are given upstream of blood to the downstream position where indicator concentrations such as temperature are detected). Is modeled, the approximate spectrum of noise is determined, and the output of the system is used in a feedback loop to adaptively update the parameters of the model and thus improve the estimation of heart rate output (CO). Is described. U.S. Pat. No. 5,687,733 (McKown et al., November 18, 1997) describes improvements to the early McKown '414 system that estimate both CO trends and instantaneous CO values. In addition, in the McKown system, the channel's zero frequency (dc or steady state) gain is required to obtain an estimate of cardiac output (CO).

【0006】

These known systems provide an estimate of cardiac output with varying accuracy, but they are defined as the ratio between the cardiac output of the heart (SV) and its final diastolic volume (EDV). Cannot provide an estimate of cardiac output (EF). Thus, ejection fraction is an indicator of how efficiently the heart drains the blood it can contain.

【0007】

Due to the importance of its diagnosis, there are several known methods of measuring EF. However, such systems are often based on the use of infused bolus and on the estimation of the drainage (heat dilution) curve in the blood vessel. For example, US Pat. No. 4,858,618 (Konno et al., Published August 22, 1989) describes a thermal dilution system for determining right ventricular ejection fraction. In this known system, the cold bolus indicator is injected into the right ventricle. Pre-bolus and post-bolus temperatures are detected in the pulmonary arteries. The temperature difference is used to determine the ejection fraction.

【0008】

One problem with using bolus to determine EF is that it is difficult to establish where to start the measurement on the detected bolus curve. This is because the anterior surface of the curve is highly dependent on mixing, heart rate, and even how quickly the administering nurse pushes the syringe plunger while the bolus is infused. Another problem faced by all such known systems is that if they generate EF estimates, they need to be synchronized with the cardiac cycle to reduce the effects of heart rate. Some systems synchronize based on the plateau of the runoff curve, which assumes a fast and very accurate thermistor. Other systems rely on EKG trigger synchronization. However, EKG synchronization is difficult because each requires slaves and precise adjustments at the timing of other devices that collect their own data.

【0009】

A further problem with existing systems that determine EF is due to the need to identify separate plateaus in the dilution profile produced by the heartbeat. This is necessary because these systems use plateaus as markers to fit exponential or ratio-based curves to the data. These are then used to estimate dilution decay. However, this approach is actually accurate only when the heart rate is relatively low and the thermistor response is significantly faster than the decay parameter τ.

【0010】

In short, these traditional systems assume a square wave dilution curve. However, this is usually an unrealistic assumption. First, many patients in hospitals who require EF measurements are not in optimal health. Rather, they tend to have relatively high and abnormal heart rates. Further, in systems that use a relatively cold fluid bolus, the detected heart rate can be inaccurate because the cold bolus itself tends to affect not only the heart rate but also its regularity. Second, the actual thermistor distorts the plateau, which distorts the exponential fit itself. Third, as EF rises, so does the plateau. This reduces the plateau used by the system and therefore its accuracy. This is because the signal-to-noise ratios of these systems are limited.

【0011】

Therefore, a system capable of generating continuous estimates of EF, EDV, or both is needed. The present invention provides such a system.

【0012】

(Gist of the Invention) According to the present invention, the cardiac ejection fraction EF is an indicator (heat, etc.) according to a predetermined indicator driver signal x (t) on the blood at an upstream position of the heart (preferably right atrium / right ventricle). Is estimated based on indicator dilution by injecting blood and by detecting the local indicator concentration of blood (such as temperature) at a downstream location (preferably the right branch of the pulmonary artery) using an indicator sensor such as a thermista. The indicator sensor, as the name implies, produces an indicator density signal y (t) corresponding to the locally detected density of the indicator. The region from upstream to downstream (including upstream and downstream) forms a channel for blood. Heart rate HR, or preferably average heart rate HR<u style="single"></u>avg is also measured by a heart rate monitor or subprocessor.

【0013】

A model of the channel is assembled and calculated by the processor as a given function of the indicator signal x (t) and the indicator concentration signal y (t). The model is preferably the delayed normal transfer function Hxy. It has an output parameter that includes both a zero frequency gain value dc and an indicator attenuation parameter τ for the modeled channel transfer function Hxy. The cardiac ejection fraction EF of the heart is then continuously estimated as a given function of heart rate and decay parameters. Preferably, EF is EF = 1-exp (-60 / (τ)^*Calculated as HR)).

【0014】

The present invention can also estimate the final dilated volume of the heart. To do this, the cardiac output CO value is also continuously estimated by the processor, preferably as a function of the zero frequency gain of the channel transfer function. The final expanded volume (EDV) value is estimated by the processor as a function of CO value, ejection fraction and heart rate.

【0015】

One embodiment of the invention uses a fast indicator sensor (such as a thermistor). This means that the step response is faster than the attenuation parameter τ. In this embodiment, the indicator concentration y (t) is sent to the processor through an open bandwidth front-end filter for inclusion in the model calculation before it is used in the channel model. It acts as a low-pass filter that removes only frequency components near the Nyquist sampling frequency, the Nyquist sampling frequency, or above the Nyquist sampling frequency. This provides substantially "raw" indicator data to the processor.

【0016】

In another embodiment of the invention, the indicator sensor is slower. To prevent this from affecting the accuracy of EF and other calculations, the transfer function Hs of the sensor can preferably be determined when manufacturing the catheter (eg) to which the sensor is mounted. The parameters featuring this transfer function Hs are then prestored in the memory device. The effect of slow sensor response on the EF calculation is then eliminated, or at least significantly reduced, by applying the "inverse function" of the sensor transfer function Hs to the channel transfer function Hxy. In essence, it "defilters" the indicator concentration signal y (t) to recreate the "raw" indicator data.

【0017】

(Detailed Description) In the broadest term, the components constituting the system according to the present invention include an indicator driver / sensor pair. The indicator driver / sensor pair injects the indicator into the upstream position of the heart (preferably the right atrium) and detects the indicator concentration signal at the downstream position (preferably the right branch of the pulmonary artery). After adjusting the signal detected by the indicator sensor, the cardiac output CO is estimated and the indicator spillage or attenuation included in the parameter set that characterizes the indicator response of the channel (the bloodway between the indicator driver and the sensor). Generate the parameter τ. It should be noted that the indicator driver signal can be generated in many different forms (continuous, impulse, deterministic, repetitive, pseudo-random, or even random).

【0018】

Also, the patient's average heart rate HR is detected and preferably HR at the same observation interval used for τ estimation.<u style="single"></u>Averaged as avg. Then the damping parameters τ and HR<u style="single"></u>Use avg to calculate a continuous estimate of the ejection fraction EF of the heart, especially the right heart.

【0019】

To estimate the final dilated volume of the heart, the present invention also requires an estimate of CO. In a preferred embodiment of the invention, CO is determined using the methods and systems described in McKown '733. The advantage of this choice is that the McKown '733 system provides continuous CO (as well as τ) estimates. Another advantage is that it is more accurate than other traditional alternatives. Yet another advantage is that when it is necessary to measure CO, it is more stable than other known systems in the presence of many noise sources typically found in that environment. These advantages are largely based on the fact that the delayed normal model uses the entire observation (data acquisition) interval, which makes it easier for the delayed normal model to reduce noise and allow the attenuation parameter τ to be determined more accurately.

【0020】

Moreover, unlike other systems, McKown'733 not only bases its estimation on its evaluation of the complete model of the channel, but also recursively updates its parameters. One of the advantages of this is that the complete channel model does not throw away the large amount of information lost in other traditional systems. In addition, recursion causes all of the detected data to be "used" at all times. Because even historical data is included in the current update.

【0021】

To fully understand the present invention, it is useful to understand at least some of the theories underlying the CO estimation routine used in McKown '733 (see the patent itself for a complete explanation). Therefore, a brief summary is given below.

【0022】

In relation to the estimation of cardiac output, the "delayed normal model" described by Bassingthwaighte et al. In "Application of Lagged Normal Density Curve as a Model for Arterial Dilution Curves", Circulation Research vol.18, 1966 is particularly accurate. And has been shown to be useful, therefore it is the model of cardiac output used in McKown '733. The lagging normal model is defined as a linear time-invariant system (LTIS) whose impulse response is a superposition of the unit region Gauss (normal distribution) function and the unit region decay exponential function. Gauss has two parameters. Mean μ and standard deviation σ. The exponential function has one parameter. The time decay parameter τ. Unit gain, lag normal transfer function H, sampled at each frequency ω (ω is the independent variable of this model)<u style="single"></u>LN therefore depends on μ, σ, and τ as follows.

【0023】

[Number 1]<img he="18" id="000002" wi="114" file="2_0003600527.tif" img-format="tif" img-content="drawing" /> Here, exp is an exponential function, and the physical meaning of the parameters is μ: pure time delay representing translational flow rate σ: index of random dispersion τ: related to mixing into the dispersed volume, which is a blood vessel in this example. The unit of the time constants μ, σ, and τ is time (seconds), and the unit of ω is radian per second.

【0024】

Other indicators can be used, but in a preferred embodiment of the McKown '733 system, heat is used as the indicator and the indicator driver signal is a pseudo-random binary sequence (PRBS). Therefore, the driver / sensor pair consists of a heater and a thermistor. H<u style="single"></u>LN is the composite value Hxy (ω)_nEstimated as the optimal fitting of the vector of), the composite value Hxy (ω)_n) Each represents an index of the transfer function between the heater power signal x and the thermistor temperature signal y. Each vector element has 10 frequencies ω_nIncludes parameters fitted to the temperature data measured at each of the (first 10 PRBS harmonics). If μ, σ, and τ are known, then the 10 composite measurements Hxy (ω)_nEach of) provides an estimate of cardiac output CO individually according to the following equation.

【0025】

CO (n) = K^*H<u style="single"></u>LN (ω)_n) / Hxy (ω_n), N = 1 to 10, where K is a known or experimentally determinable transformation constant.

【0026】

To apply this relationship, the McKown'733 system firstly calculates 10 cardiac output estimates CO (n) as well as what the values of μ, σ, τ, should be. Decide whether to combine with. It should be noted that cardiac output does not depend on the shape of H (ω) or Hxy (ω), but only on zero frequency gain, dc of Hxy. However, the experimental transfer function Hxy has 10 non-zero frequencies ω_nIn essence, the McKown'733 system estimates the Hxy (ω) measured for zero frequencies, as it is measured at. Then, using a (eg) single-optimal routine, 10 modeled transfer function values H<u style="single"></u>Provides the best fit of xy to the observed values. The relationship shown above for CO can then be reduced to CO = K / dc, where dc is the zero frequency (ω = 0) gain value in units of Celsius per watt, K. Is an experimentally determined constant, approximately equal to 0.0158, with (liters per minute) / (celsius per watt) units.

【0027】

Therefore, the dc value is mainly important for obtaining a CO estimate. This is the measurement that is most involved in the McKown'733 system, and tests, experiments and experience have shown that the optimal routines used, for example, τ, by constraining one or more parameters in many applications. By constraining to the linear function of, it can be speeded up with a negligible loss of accuracy. The inventors of the present invention can improve both speed and accuracy by removing the constraints, by constraining other parameters, or by changing the constraints on the sigma. Two-pass optimization can also be shown to be beneficial in some applications. For example, using the existing optimal routines of the McKown'733 system, we first quickly obtain the correct dc value by imposing constraints on σ in the first pass, and then in different routines, perhaps with different constraints. The dc value is used to calculate other parameters such as τ.

【0028】

However, it is important to understand the present invention that the McKown'733 system provides a continuous CO value (equivalently, a dc value) as well as an attenuation parameter τ. Here, "continuous" does not mean that the displayed value "changes continuously", but that it can be updated every processing cycle (preferably PRBS cycle) after the initialization period. Note what it means.

【0029】

The McKown'733 system is preferred for the reasons mentioned above, as it actually already exists. However, other systems that can provide a channel transfer function (impulse response) can be used instead. If the system can also generate values and determine CO (or dc) and τ from those values, these values are calculated in the present invention as described below for EF and EDV. Because it is used for.

【0030】

The following description of various embodiments of the invention assumes that heat is used as an indicator of injection into blood. Similarly, the upstream indicator driver is the heating element and the downstream indicator sensor is the thermistor. This technique is a well-established choice and is the choice in plot types and tests of the present invention. In addition, using the method described in McKown '733 and using heat as an indicator, a highly accurate CO estimate is shown. Nevertheless, heat is only one possible indicator that can be used in the present invention. As long as the indicator injectors and sensors used produce a measurable, sufficiently clear, noise-free signal (the signal can be determined by routine experimentation), then the signal is in the present invention. It can be used unmodified to the rest of the system or in an easily feasible modified state.

【0031】

As an example with different indicators that can be used in the present invention, a known fluorescent material may instead be injected into the patient's heart using a known device. Fluorescence can then be detected downstream, and by using known sensors, variations in fluorescence can function as indicator concentration signals. Weak radioactive dyes or reagents may be used as well.

【0032】

As another embodiment in a system of preferred embodiment, such as the Yelderman's or McKown'733 system, heat is injected according to a pseudo-random binary sequence (PRBS). However, it is also possible to inject the fluid to follow a similar injection pattern. As long as the infusion period is slow enough, for example, a small bolus to approximate the PRBS profile can relax the blood flow, and the concentration of the bolus material is detected downstream using the corresponding known sensor to provide an indicator concentration signal. Can be established. In short, the indicator injectors and sensors used produce a measurable, sufficiently clear, noise-free signal (the signal can be determined by routine experimentation), which is then the signal in the present invention. , Can be used in the unmodified state for the rest of the system, or in an easily feasible modified state.

【0033】

FIG. 1 is a block diagram of a first embodiment of the system according to the invention for continuous estimation of the ejection fraction of a patient's heart and / or the final dilated volume. For accurate measurement of the patient's cardiac output CO, the use of the McKown'733 system specifically injects an indicator into the patient's right atrium / right ventricle 100 or nearby blood and pulmonary artery 102. It has the advantage of detecting the indicator density at or near the branches of the heart. Therefore, these injection and detection positions are assumed below for the purpose of explaining preferred embodiments of the present invention. The flow of blood from the right atrium / right ventricle through the pulmonary artery is indicated by an arrow parallel to FIG.

【0034】

In order to improve accuracy, it is preferable to use a thermal signal as the basis for measuring cardiac output CO. However, as explained above, this is just one possible indicator that can be used. The indicator driver or injection device 104 is located in the right atrium 100. In a preferred embodiment where the indicator is thermal, the indicator driver is an electrical heating element 104. The heating element 104 is preferably an electrical resistance element whose temperature is determined by the current or voltage supplied to the element via the driver circuit 106. The driver circuit 106 drives the heating element 104, whereby its temperature follows a predetermined signal profile.

【0035】

In a preferred embodiment of the invention, the indicator signal (preferably temperature) profile x (t) followed by the indicator driver (preferably the heater) is described in the McKown '733 patent. In this system, as also described above in the Yelderman system, a thermal signal that can be efficiently detected at the downstream detection position is supplied with a high spectral content but low, and therefore with a reduced traumatic average of the heat applied. Therefore, the thermal signal is generated based on a pseudo-random binary sequence (PRBS). Moreover, although the thermal signal is pseudo-random, it has long been known to this system, so that the characteristics of the calculations based on it are well understood and well tuned.

【0036】

In practical applications, the injection device 104 cannot exactly follow the desired injection profile indicated by the driver 106. For example, the heating element does not exactly follow the square wave pattern due to the heat rise and cooling delay of the heating element. As a result, the invention preferably comprises the driver signal estimation subsystem 107, which comprises the estimated indicator drive signal x corresponding to the desired injection indicator profile x (t).^*Generate (t). In relation to thermal dilution, for example, the power output of the metal of the heating element is itself temperature dependent due to the properties of the metal used as the resistance element (eg nickel). Therefore, x^*One way to estimate (t) is to measure both voltage and current through the heating element. Multiplying the two gives a good estimate of the actual amount of indicator (here, heat) given to the blood.

【0037】

The indicator concentration sensor 108 is located downstream of the pulmonary artery 102. In a preferred embodiment where the indicator is heat, the sensor is a thermistor or some similar temperature sensing element 108. The heating element 104 and thermistor 108 are preferably spaced apart at or near the distal end of the catheter, then the catheter is injected into the patient's vein and the heating element and thermistor are in the operating position. It is passed intravenously and through the vein until it is reached. This technique is well known and therefore will not be described further.

【0038】

Conventional power clock devices preferably include supplying power and timing signals to the driver circuit 106 and other configurations of the invention. Since these devices are well known, they are not illustrated or described.

【0039】

The electrical output signal from the thermistor 108, i.e., the indicator density signal y (t), is sent to the concentration estimation circuit or subprocessor 110, which is contained within the main processor 112 or electrically connected to the main processor 112. It is given as an input signal. In the first embodiment shown of the present invention, the assumption that the thermistor 108 has a fast response is that its instantaneous temperature signal closely matches the actual instantaneous temperature of the blood whose temperature is being measured. And it means that it is reflected in a predictable manner.

【0040】

The concentration estimation circuit 110 included in this embodiment of the present invention is preferably the same as that used in the McKown '733 system, but with one significant modification. In the McKown'733 system, the thermistor output signal y (t) is lowpass filtered at a cutoff frequency of approximately 1Hz before being sampled at 10Hz. This low sampling bandwidth not only helps reduce the effects of noise, but is also sufficient for CO calculations. This is because the raw temperature data is not needed to determine the parameters of the delayed normal model. This is because only zero frequency (dc) gain is needed and used to calculate CO.

【0041】

However, in the present invention, the actual shape of the thermal dilution curve, i.e., instantaneous values over the entire cycle, is used to provide continuous EF estimates. It requires raw temperature data. To provide this raw data, the concentration estimation circuit 110 in this embodiment of the present invention preferably has a much wider bandwidth than the McKown '733 system. This increased bandwidth is in the same magnitude range as the Nyquist bandwidth, which is half the sampling rate. In one plot type of the invention, the 3Hz sampling bandwidth (determined by thermistor limits) was used rather than the 1Hz bandwidth in the McKown '733 system. No other filtering is needed. Assuming that the thermistor 108's response is fast enough not to interfere with the fastest predicted possible τ value (minimum velocity can be determined by conventional experiments), then used in, for example, McKown '733. The delayed normal model to be done automatically provides the exact value of τ.

【0042】

However, regardless of the thermistor used, the sensor used (here, thermistor), the available signals included in the further processing are not accurate, but some estimation of the actual indicator concentration in the sensor. .. In many cases, the signal is filtered (in fact, whenever the signal is digitized, it is "filtered" by the very nature of the analog-to-digital conversion). For example, information about frequency components above the Nike straight, which is half the sampling frequency, is lost due to aliasing. As a result, the output signal from the indicator density signal estimator 110 is the estimated and detected density signal y.^*(t).

【0043】

Processor 112 includes a parameter modeling subprocessor 114, which is a delayed normal modeling system preferably described in McKown '733. As summarized above and described in McKown '733, this modeling method uses cross-correlation and optimal routines to calculate the estimation of the zero frequency (dc) gain of the channel transfer function and the estimation of the attenuation parameter τ. To do. The inputs of the modeling system 114 used in the present invention are the indicator driver signal x (t) and the thermistor signal y (t), or more precisely their estimation x.^*(t), y^*(t). Its output is the estimation of the zero frequency gain of the channel transfer function and the estimation of the attenuation parameter τ.

【0044】

Processor 112 also includes or is fitted with a heart rate monitor or estimation circuit 116. It preferably calculates the average heart rate of the patient's heart experimentally or otherwise at predetermined intervals. The heart rate monitor 116 according to the first embodiment of the present invention illustrated in FIG. 1 is preferably a subprocessor or subroutine incorporated in the processor 112.

【0045】

In this embodiment, the heart rate monitor 116 is an estimated indicator driver signal x.^*(t), Estimated indicator concentration signal y^*It has both (t) as input signals. Then, using standard signal processing techniques, the average heart rate HR from the temperature signal<u style="single"></u>You can get avg. For example, reliable HR measurements are obtained by calculating the power spectral density (PSD) of the zero average temperature signal in any conventional embodiment. It should be noted that the McKown'733 system includes a drift removal routine that includes these calculations and can therefore provide HR when used in a system according to the invention. Then, the position of the peak PSD in the normal heart rate range (known from experience) is HR = PSD<u style="single"></u>peak (Hz)^*Since it is 60 (beat per minute), the value HR is specified. For consistency, this should be done on a simultaneous observation window using the parameter modeling subprocessor 114. When a large number of current heart rate numbers are obtained, they are averaged and HR<u style="single"></u>Can generate avg. The average is preferably performed at the same observation interval, which is used to determine τ to minimize errors in calculating EF and EDV.

【0046】

Estimated driver signal x^*Note that in some cases it is possible to estimate heart rate using only (t). As mentioned above, the resistance of the heating element is generally temperature dependent. As a result, even if a constant voltage is applied to the internal resistance element (eg, consisting of nickel), due to the cooling, beating (and thereby non-uniform) effects of the blood surrounding the element, The core temperature of the element changes. Therefore, the HR signal is the signal x^*It is superimposed only on (t) and can be identified using conventional filtering techniques.

【0047】

Average heart rate HR using other devices<u style="single"></u>You may provide a value for avg. These include the heart rate output of a conventional dedicated heart monitor or an existing multi-parameter patient monitor. Although the present invention does not necessarily require averaging of heart rate, the smoothing effect of averaging facilitates irregular filtering without sacrificing system performance and provides continuous EF estimation. This is suitable. As long as the average is done with a sufficient number of heartbeats, the average also eliminates the need to synchronize with the heart cycle. In existing EF estimation systems, heart rate averages typically include 4-7 heart rates. However, in the present invention, when using the PRBS thermal signal, some additional beats may be included. This is because it usually occurs several times during one PRBS cycle. This further reduces the noise sensitivity of the present invention as compared to known systems.

【0048】

FIG. 2 shows an external heart rate monitor referenced in this embodiment, 117. The heart rate signal generated by the external device may be used directly in the present invention or may require filtering or other adjustments, thereby establishing the heart rate, other of the present invention. Adjusts the signals used by the components, or both.

【0049】

As shown in FIG. 2, in this case, the heart rate monitor circuit 116 is then modified in any conventional manner to appropriately adjust the heart rate signal provided by the external monitor 115. In addition, the heart rate signal is supplied by an external device, so the estimation circuit 116 is HR.<u style="single"></u>To get an accurate estimate of avg, x as input^*(t) and y^*Does not require (t).

【0050】

dc, τ, and HR<u style="single"></u>The calculated value of avg is given to the EF subprocessing system 118 as an input signal. The EF subprocessing system 118 is preferably built into the processor 112 and implemented in software. The EF subsystem 118 then further calculates an estimated EF value, an estimated EDV value, or both in the manner described below. Note that the values available for it or for HR, CO, and heart rate blood volume (SV) described below are calculated as well.

【0051】

Once the EF and / or EDV values have been calculated, they are displayed to the user on any conventional display 120. The display 120 includes any required conventional display driver. As desired, the display also displays the CO, HR, SV values available as output from the EF subprocessor 118. Of course, the EF / EDV value may be stored instead or even electronically in memory, or may be transmitted over the network or to other processing equipment. In the following description, some calculations will be described. It should be understood that all of the calculated values can be multiplied by appropriate scaling constants, which ensures, for example, the desired display range or is converted to different units.

【0052】

Now consider the impulse injection of heat into the right atrium / right ventricle. ΔT (i) is the pulmonary artery (eg, measured from one R wave to the next) from the baseline temperature when heat is first injected into the heart after i cardiac cycles (eg, measured from one R wave to the next). The change in blood temperature (or the concentration of any other indicator used) in PA). ΔT (in) is the temperature change before n R waves. At that time, it is well known that the following relationship approximates the indicator decay curve (also known as the physiological runoff curve).

【0053】

ΔT (i) = ΔT (in)^*exp (-t / τ), where τ is the decay constant.

【0054】

Physiological runoff attenuation is also (1-EF)ⁿCan be expressed by, where n is the number of cardiac events (eg, RR interval) during the observation period (often taken at about 80% to 30% of the peak value). For example, assume an ejection fraction (EF) of 0.6 (60%). At the end of an interval, indicators (eg, heated blood) remain in the heart at (1-0.6) = 0.4 (40%). After one more interval, only 40% of this 40% remains, i.e. (1-0.6)^*(1-0.6) = 0.16, that is, 16% of the total at the beginning. Therefore, the following relationships are also retained.

【0055】

ΔT (i) = ΔT (in)^*(1-EF)ⁿThe time t can then be expressed in terms of heart rate HR (beats per minute) and n. As a result, t = n^*60 / HR Combining these three equations to find EF, EF = 1-exp (-60 / (τ)^*HR)) According to the present invention, it is not always necessary for HR to be measured from the R wave of one heart to the next. Rather, the average HR is preferably used for the reasons mentioned above. At that time, the EF value obtained by using the mean value becomes smoother, the sensitivity of the heart rate to irregularities is reduced, and it can be continuously updated not only after the subsequent R wave or other triggered cardiac event.

【0056】

The present invention also includes τ and HR (ie, HR).<u style="single"></u>Observe that EF can be estimated as long as it has an estimate of avg). Of course, these are just the parameters provided by the parameter modeling subprocessor 114 and the heart rate monitor 116 respectively. Therefore, the EF subsystem 118 has the equation EF = 1-exp (-60 / (τ).^*EF is determined by calculating HR))).

【0057】

CO = HR^*SV, where SV is the heart rate blood volume, and it should be further noted that CO is measured in units of volume (liters) per minute. This means that the amount of blood that the heart drains in one minute is equal to the number of beats per minute multiplied by the amount that the heart drains per contraction (beat). Finally, it should be noted that the final expansion volume (EDV) and ejection fraction (EF) are related as follows.

【0058】

[Number 2]<img he="18" id="000003" wi="41" file="3_0003600527.tif" img-format="tif" img-content="drawing" />Again, the intuitive relationship that the heart's drainage efficiency (EF) is the ratio between the amount of blood that the heart drains with each beat (contraction) and the amount of blood in the heart chamber just before the beat. To express. By reorganizing this expression, EDV = SV / EF is understood.

【0059】

Therefore, the EF subprocessing system 118 either calculates CO based on the value dc received from the parameter modeling subprocessor 114 and the given conversion K, (CO = K / dc), or the parameter modeling subprocessor 114. If already calculated in, receive the value CO. By dividing CO by heart rate HR (obtained from heart rate monitor 112), then EF subprocessing system 118 calculates SV = CO / HR, and once SV is known, EF subprocessing system 118 is 1-exp. (-60 / (τ^*Since we have an estimated EF by calculating HR)), we can calculate EDV as SV / EF.

【0060】

The EF subprocessing system 118 and the parameter modeling subprocessor 114 do not have to be separate units. Rather, they can both be realized as a single processing device. In fact, they can also be implemented as simply different software modules of processor 112. As such, CO and EDV calculations can be performed on either subprocessor 114 or 118 without affecting the resulting or final displayed value.

【0061】

FIG. 3 is a block diagram showing a second embodiment of the present invention. In FIG. 3, essentially the same components as those shown in FIGS. 1, 2 and above have the same reference numerals. This embodiment is suitable when the indicator sensor, such as the thermistor 108, is slow compared to the thermistor attenuation parameters. For this reason, the measurement of the exponential decay parameter τ is influenced by the thermistor's slow response time, especially at fast heart rates.

【0062】

To compensate for this, the present invention predetermines the transfer function (equivalently step response) of each sensor (here, thermistor) used as sensor 108, and the "inverse function" of this transfer function. Apply to Hxy to "defilter" or compensate for the effects of slow sensor response times. There are several known methods of characterizing the step response of a transfer function, the beginning of which is simply applying a series of impulse input signals to the transfer function, measuring each response, and then averaging the results.

【0063】

One practical way to predetermine the transfer function (Hs) of some thermistors is to have several catheters (each equipped with a thermistor) immersed (for example) from air to a temperature control bath. Is to establish a catheter measuring unit. Any of the many known curve fitting techniques are then used to calculate the parameters of each mathematical model of Hs that best fits the recorded thermistor data. These parameters can then be stored in a permanent memory device, such as an EEPROM for each thermistor. By using this procedure, unwanted things can be removed, or at least significantly reduced. In this embodiment of the invention, the values characterizing Hs are stored in such a device, shown as component 200 in FIG.

【0064】

The pre-computed value contained in the transfer function storage device 200 can be used as an input signal to the modeling subprocessor 114 or the indicator concentration estimation circuit 110. In this embodiment, the indicator concentration estimation circuit also has a thermistor signal y (t) and a heater signal x (t) (preferably its estimation x) as input signals.^*Following (t)), these are also conferred on the heart rate averaging circuit 116.

【0065】

The estimation circuit 110 for the fast thermistor in the first and second embodiments of the present invention (FIGS. 1 and 2) basically acts as an open bandwidth front-end lowpass filter and is near the Nyquist frequency. Filter only frequencies at or above the Nyquist frequency to pass "raw" thermistor data. However, in this third embodiment (Fig. 3), the slower thermistor itself imposes a frequency-dependent attenuation profile on the data and effectively functions as a lowpass filter itself. Therefore, the system needs to "unfilter" the data and use EF calculations to get the "raw" data. The McKown'733 system has a composite value of Hxy (ω)_n) Vector as the optimal fitting to the observed value H<u style="single"></u>It should be recalled to estimate the LN. This vector Hxy is still available in the present invention when the McKown '733 modeled subprocessor is used as the subprocessor 114. Otherwise, like McKown '733, the modeling subprocessor 114 should calculate each Hxy (measured transfer function of the channel).

【0066】

Here, Hxy<u style="single"></u>Let p be the desired complex transfer function of channel (cardiac / vascular) physiologic drainage, i.e., that transfer function is observed without the distortion effect of slow thermistors. Suppose we calculate Hs-thermistor's predetermined transfer function-at the same frequency as Hxy. For example, in McKown '733, Hxy has 10 component elements that correspond to the power measured against the temperature transfer function of the channel in the first 10 PRBS harmonics. This is Hxy = Hs^*Hxy<u style="single"></u>Means p, thereby the "raw" transfer function Hxy<u style="single"></u>p is calculated by the modeling subprocessor 114 as 10 scalar vector (component-component) divisions in Hs of Hxy.

【0067】

Therefore, Hxy<u style="single"></u>p = Hxy / Hs Then this Hxy<u style="single"></u>The p data, that is, the Hxy data "scaled" by Hs, is either transmitted to the modeling subprocessor 114 (if scaling is done on the estimation circuit 110) or used by the modeling subprocessor 114 as described above. Will be done. This is a mathematically delayed normal transfer function H<u style="single"></u>Hxy with LN<u style="single"></u>Corresponds to modeling p-data. Not only does this provide the same dc value that is inversely proportional to CO, but it also provides an accurate τ. Because the transfer function Hxy<u style="single"></u>The higher frequency components of p are adjusted to compensate for the slow response of the thermistor. Therefore, the EF subprocessor 114 can calculate EF and EDV as in the first embodiment of the present invention. Also, as in the first embodiment, the calculated EF / EDV values are also continuous in the same sense as McKown'733, i.e. they have the same initial capture time and update interval.

【0068】

A further problem with slow thermistors is that when the HR is large, thermistor data alone are not useful in determining heart rate HR. For this reason, in a third embodiment of the invention (FIG. 3), the heater signal x (t) and the "raw", i.e., "scaled" or "defiltered" thermistor signal y (t) (preferably). , These estimates x^*(t), y^*Both (t)) are preferably supplied to the heart rate averaging circuit 116. Correlation and other known techniques are then applied, along with both of these available signals, to obtain more accurate estimates even at high heart rates.

【0069】

Of course, as mentioned above, the average HR value (HR) is not related to the thermistor characteristics.<u style="single"></u>A conventional external heart rate meter may be used instead of the heat rate circuit 116 to provide avg). Thus, FIG. 4 shows an embodiment of the invention in which the external heart monitor 117 supplies a heart rate signal and stores the sensor's "reverse transfer function" Hs parameters in advance.

【0070】

The main difference between the first and third embodiments of the present invention (FIGS. 1 and 3) and the second and fourth embodiments (FIGS. 2 and 4) is whether the thermistor is "fast" or "fast". It depends on whether it is "slow" or not. In a practical application, the boundary between the two does not have to be clear. Of course, the "slow-sensor" embodiment of the present invention may always be used. Because they also include "fast-sensor" embodiments (the transfer function Hs is simply "flat" with respect to the entire sampled bandwidth). However, slow sensor embodiments require extra manufacturing steps, the cost of individual sensor calibration, the addition of a memory device 200 to store sensor response parameters, programming (or vice versa).

【0071】

The determination of which embodiment to use can be made using conventional simulation and experimental techniques. For example, the heater signal x (t) and thermistor signal y (t) can be generated, if possible, through a simulation that incorporates actual data from the patient. The parameter modeling subprocessor can then estimate τ. Alternatively, the τ value can be estimated from actual measurements obtained using very fast known thermistors. The transfer function Hs of a representative thermistor of the type used in the present invention can then be determined as described above. The calculated EF values can then be compared with and without compensation for the Hs profile. The uncompensated EF value differs from the compensated EF value by less than a predetermined amount, and then the thermistor can be assumed to be fast enough to use the first or third embodiment of the invention.

[Simple explanation of drawings]

FIG. 1 is a block diagram of a first embodiment of the system according to the invention for continuous estimation of the ejection fraction of a patient's heart and / or the final dilated volume. Here, a quick answer indicator sensor is used to measure the blood indicator response, and the patient's heart rate is estimated by the system itself.

FIG. 2 is a block diagram according to a second embodiment of the present invention in which a quick answer indicator sensor is used, but the patient's heart rate is detected by an external device.

FIG. 3 is a block diagram according to a third embodiment of the invention in which a slow answer indicator sensor is used, but the response steps of that sensor are characterized and available for the rest of the system. .. In this embodiment, the patient's heart rate is further estimated by the system itself.

FIG. 4 shows a block according to a fourth embodiment of the invention in which a slow answer indicator sensor is used, the response step of which is available, and the patient's heart rate is detected by an external device. It is a figure.