System and method for estimating cardiac output

• US 5,687,733 A
• Filed: 10/26/1995
• Issued: 11/18/1997
• Est. Priority Date: 10/26/1995
• Status: Expired due to Term

• Current Assignee: Edwards Lifesciences Corporation
• Original Assignee: Baxter International Inc.
• Inventors: McKown, Russell
• Primary Examiner(s): Bahr, Jennifer
• Assistant Examiner(s): WINAKUR, ERIC FRANK
• Application Number:
  US08/548,937
• Time in Patent Office: 754 Days
• Field of Search: 128/691, 128/692, 128/713, 128/736, 128/668
• US Class Current: 600/505
• CPC Class Codes: A61B 5/028 by thermo-dilution
• CPC Class Codes: A61B 5/029 Measuring or recording blood output from the heart, e.g. minute volume A61B8/065 takes precedence
• CPC Class Codes: A61B 5/7257 using Fourier transforms

Abstract

Cardiac output CO is separately estimated in a local estimator and a trend estimator to provide CO values. A thermal signal is injected preferably according to a pseudo-random binary sequence signal profile at an upstream position in a blood flow region of a patient'"'"'s body and is sensed as an indicator output signal at a downstream position. Low-frequency noise is preferably removed from the sensed indicator output signal. The local and trend estimations are based on measured frequency-domain transfer function values between the sensed and input indicator signal in relation to a pre-determined transfer function model, which is preferably a lagged normal model. The trend estimator is preferably a Kalman filter. The local estimator preferably forms its estimate based on non-recursive optimization of a cost function. The local estimator preferably provides initial values to start the trend estimator.

Claims

1. A method for estimating blood output through a flow region of a patient'"'"'s body, comprising the following steps:

A) injecting an indicator as an indicator input signal having an input signal profile and a period into an upstream position in the flow region;

B) sensing the presence of the indicator at a downstream position in the flow region to determine an indicator output signal;

C) estimating as a predetermined function of the input signal profile and the indicator output signal both a local blood output value over a local estimation time, and a trend blood output value over a trend estimation time, in which the trend estimation time is longer than the local estimation time, thereby providing estimated blood output values corresponding to both relatively fast and slow changes in blood output.

2. A method as in claim 1, in which the step of estimating the trend output value comprises the step of recursively estimating the trend output value.

3. A method as in claim 2, in which the step of recursively estimating comprises Kalman filtering.

4. A method as in claim 2, in which the step of estimating the predetermined function in turn comprises the following steps:

measuring values of a frequency-domain transfer function between the indicator input signal and the indicator output signal for each period of the indicator input signal, the measured frequency-domain transfer function forming input signals for both local blood output value and trend blood output value estimation;

selecting state parameters of a predetermined transfer function model relating the indicator output signal to the indicator input signal;

determining optimal local state parameters as a predetermined optimization function of the transfer function model and the measured frequency-domain transfer function values, and estimating the local blood output value as a predetermined output function of at least one of the optimal local state parameters; and

recursively estimating optimal trend state parameters by Kalman filtering the measured frequency-domain transfer function values, and estimating the trend blood output value as the predetermined output function of at least one of the optimal local state parameters.

5. A method as in claim 4, further including the step of initializing the Kalman filtering step using the optimal local state parameters.

6. A method as in claim 4, further including the following steps:

A) generating the input signal profile as a pseudo-random binary sequence (PRBS);

B) determining an autocorrelation value Cxx of the input signal and converting the autocorrelation value Cxx to the frequency domain;

C) determining a cross-correlation value Cxy between the input signal and the output signal and converting the cross-correlation value Cxy to the frequency domain; and

D) computing the measured frequency-domain transfer function values as a predetermined function of the quotient between the frequency-converted cross-correlation and autocorrelation values.

7. A method as in claim 4, further including the step of removing any low-frequency noise trend from the indicator output signals before estimating either the local blood output value or the trend blood output value.

8. A method as in claim 1, in which the step of estimating the predetermined function comprises measuring a frequency-domain transfer function between the indicator input signal and the indicator output signal for each period of the indicator input signal, the measured frequency-domain transfer function forming input signals for both local blood output value and trend blood output value estimation.

9. A method as in claim 1, further including the following steps:

grouping sensed indicator output signals into a sequence of runs, each run comprising the indicator output signals sensed during a predetermined run period;

determining a signal edit threshold and a signal edit condition as predetermined functions of the indicator output signals;

comparing the indicator output signals with the signal edit threshold; and

setting to a predetermined default value any indicator output signals that meet the signal edit condition before the step of estimating the blood output values.

10. A method as in claim 9, wherein the step of determining the signal edit condition comprises comparing absolute values of the indicator output signals with the signal edit threshold, the signal edit condition being that the absolute values of the indicator output signals exceed the signal edit threshold, thus indicating the presence of a substantially impulsive noise source in the flow region.

11. A method as in claim 9, wherein the step of determining the signal edit condition comprises determining a maximum output value range for each run, and the step of setting to the predetermined default value comprises setting to the predetermined default value all output signals in any run whose maximum output value range exceeds the signal edit threshold, thus indicating the presence of a non-impulsive noise source in the flow region.

12. A method for estimating blood output through a flow region of a patient'"'"'s body, comprising the following steps:

A) injecting a thermal indicator as an indicator input signal having a pseudo-random binary sequence (PRBS) input signal profile and a period into an upstream position in the flow region;

B) sensing the presence of the indicator at a downstream position in the flow region to determine an indicator output signal;

C) removing any low-frequency noise trend from the indicator output signals;

D) selecting state parameters of a predetermined transfer function model relating the indicator output signal to the indicator input signal;

E) determining an autocorrelation value Cxx of the input signal and converting the autocorrelation value Cxx to the frequency domain;

F) determining a cross-correlation value Cxy between the indicator input signal and the indicator output signal and converting the cross-correlation value Cxy to the frequency domain;

G) computing measured transfer function values as a predetermined function of the quotient between the frequency-converted cross-correlation and autocorrelation values;

H) determining optimal local state parameters as a predetermined optimization function of the transfer function model and the measured transfer function values, and estimating a local blood output value as a predetermined output function of at least one of the optimal local state parameters; and

I) recursively estimating optimal trend state parameters by Kalman filtering the measured transfer function values, and estimating a trend blood output value as the predetermined output function of at least one of the optimal local state parameters.

13. A system for estimating blood output through a flow region of a patient'"'"'s body, comprising:

A) injection means for injecting an indicator as an indicator input signal having an input signal profile and a period into an upstream position in the flow region;

B) indicator sensing means for sensing the presence of the indicator at a downstream position in the flow region to determine an indicator output signal;

C) estimation means including a local estimator for estimating over a local estimation time a local blood output value as a predetermined function of the input signal profile and the indicator output signal, and a trend estimator for estimating a trend blood output value over a trend estimation time, in which the trend estimation time is longer than the local estimation time, thereby providing estimated blood output values corresponding to both relatively fast and slow changes in blood output.

14. A system as in claim 13, in which the estimation means includes a recursive estimator for estimating the trend output value.

15. A system as in claim 14, in which the recursive estimator is a Kalman filter.

16. A system as in claim 14, in which:

transfer function measuring means is provided for measuring values of a frequency-domain transfer function between the indicator input signal and the indicator output signal for each period of the indicator input signal, the measured frequency-domain transfer function forming input signals for both the local estimator and the trend estimator;

the local estimator is provided for determining optimal local state parameters as a predetermined optimization function of the transfer function model and the measured transfer function values, and estimating the local blood output value as a predetermined output function of at least one of the optimal local state parameters; and

the trend estimator is a Kalman filter recursively estimating optimal trend state parameters by Kalman filtering the measured transfer function values, and estimating the trend blood output value as the predetermined output function of at least one of the optimal local state parameters.

17. A system as in claim 16, in which the local estimator is connected to the trend estimator, the optimal local state parameters forming initialization state parameters for the Kalman filter.

18. A system as in claim 16, in which:

A) the input signal profile is a pseudo-random binary sequence (PRBS);

B) the transfer function measuring means is further provided;

1) for determining an autocorrelation value Cxx of the input signal and converting the autocorrelation value Cxx to the frequency domain;

2) for determining a cross-correlation value Cxy between the input signal and the output signal and converting the cross-correlation value Cxy to the frequency domain; and

3) for computing the measured transfer function values as a predetermined function of the quotient between the frequency-converted cross-correlation and autocorrelation values.

19. A system as in claim 13, further including transfer function measuring means for measuring a frequency-domain transfer function between the indicator input signal and the indicator output signal for each period of the indicator input signal, the measured frequency-domain transfer function forming input signals for both the local estimator and the trend estimator.

20. A system as in claim 13, further including pre-filtering means for removing any low-frequency noise trend from the indicator output signals before local or trend estimation.

21. A system as in claim 13, further including:

means for grouping sensed indicator output signals into a sequence of runs, each run comprising the indicator output signals sensed during a predetermined run period;

means for determining a signal edit threshold and a signal edit condition as predetermined functions of the indicator output signals;

means for comparing the indicator output signals with the signal edit threshold; and

means for setting to a predetermined default value any indicator output signals that meet the signal edit condition before the step of estimating the blood output values.

22. A system as in claim 21, wherein the means for determining the signal edit condition comprises means for comparing absolute values of the indicator output signals with the signal edit threshold, the signal edit condition being that the absolute values of the indicator output signals exceed the signal edit threshold, thus indicating the presence of a substantially impulsive noise source in the flow region.

23. A system as in claim 21, wherein the means for determining the signal edit condition comprises means for determining a maximum output value range for each run, and the means for setting to the predetermined default value sets to the predetermined default value all output signals in any run whose maximum output value range exceeds the signal edit threshold, thus indicating the presence of a non-impulsive noise source in the flow region.

24. A system for estimating blood output through a flow region of a patient'"'"'s body, comprising:

A) indicator injection means for injecting a thermal indicator having a pseudo-random binary sequence (PRBS) input signal profile into an upstream position in the flow region as an indicator input signal;

B) thermistor means for sensing the presence of the indicator at a downstream position in the flow region to determine an indicator output signal;

C) pre-filter means for removing any low-frequency noise trend from the indicator output signals;

D) transfer function measuring means;

1) for computing an autocorrelation value Cxx of the input signal and converting the autocorrelation value Cxx to the frequency domain;

2) for computing a cross-correlation value Cxy between the indicator input signal and the indicator output signal and converting the cross-correlation value Cxy to the frequency domain;

3) for computing measured transfer function values as a predetermined function of the quotient between the frequency-converted cross-correlation and autocorrelation values;

E) local estimation means for determining optimal local state parameters as a predetermined optimization function of the transfer function model and the measured transfer function values, and estimating a local blood output value as a predetermined output function of at least one of the optimal local state parameters; and

F) trend estimation means for recursively estimating optimal trend state parameters by Kalman filtering the measured transfer function values, and estimating a trend blood output value as the predetermined output function of at least one of the optimal local state parameters.