System and method of multi-sensor fusion of physiological measurements

• US 5,626,140 A
• Filed: 11/01/1995
• Issued: 05/06/1997
• Est. Priority Date: 11/01/1995
• Status: Expired due to Fees

• Current Assignee: Spacelabs Medical Incorporated (OSI Systems Incorporated)
• Original Assignee: Spacelabs Medical Incorporated (OSI Systems Incorporated)
• Inventors: Ebrahim, Mehboob H., Feldman, Jeffrey M.
• Primary Examiner(s): Sykes, Angela D.
• Assistant Examiner(s): HUANG, STEPHEN D
• Application Number:
  US08/551,522
• Time in Patent Office: 552 Days
• Field of Search: 128/671, 128/670, 128/668, 128/696, 128/706-710, 607/17-20, 607/119-129
• US Class Current: 600/484
• CPC Class Codes: G06F 18/256 of results relating to different input data, e.g. multimodal recognition

Abstract

A system and method for fusing independent measures of the physiological parameter uses a Kalman filter for each possible combination of sensor measurements. The Kalman filter utilize probability density functions of a nominal error contamination model and a prediction error model as well as past estimates of the physiological parameter to produce the Kalman filter outputs. A confidence calculator uses Bayesian statistical analysis to determine a confidence level for each of the Kalman filter outputs, and selects a fused estimate for the physiological parameter based on the confidence level. The fused estimate and the confidence level are used to dynamically update the nominal error contamination model and prediction error model to create an adaptive measurement system. The confidence calculator also takes into account the probability of artifactual error contamination in any or all of the sensor measurements. The system assumes a worst case analysis of the artifactual error contamination, thus producing a robust model able to adapt to any probability density function of the artifactual error and a priori probability of artifact.

Claims

1. A system for estimation of heart rate subject, the system comprising:

a plurality of sensors coupled to the subject, each providing a physiological signal capable of producing a heart rate measurement;

a filter circuit receiving said heart rate measurements and producing a fused heart rate estimate for each possible combination of said heart rate measurements from said plurality of sensors in which each of said heart rate measurements in said combinations are considered to be acceptable or unacceptable; and

a confidence calculator coupled to said filter circuit to receive said fused heart rate estimates and to determine a confidence level value for each of said fused heart rate estimates indicative of an accuracy of each of said fused heart rate estimates, said confidence calculator selecting said one of said fused heart rate estimates as the estimated heart rate based on said confidence level.

2. The system of claim 1 wherein said filter circuit is a Kalman filter for each of said possible combinations, said Kalman filter circuit using past heart rate estimates and a statistical measure of contamination of said heart rate measurements by nominal error to produce said fused heart rate estimates.

3. The system of claim 2 wherein said statistical measure of nominal error contamination includes a Gaussian probability density function of said nominal error contamination, said Kalman filter using said Gaussian probability density function to produce said fused heart rate estimates.

4. The system of claim 1 wherein said filter circuit includes a statistical filter and includes a statistical measure of heart rate variability, said filter circuit using said statistical measure of heart rate variability to produce said fused heart rate estimates.

5. The system of claim 4 wherein statistical measure of heart rate variability is a Gaussian probability density function of heart rate variability, said filter circuit using said Gaussian probability density function of heart rate variability to produce said fused heart rate estimates.

6. The system of claim 1 wherein said heart rate measurements are susceptible to contamination by artifactual error, said contaminated heart rate measurements producing an unacceptable heart rate measurement, said confidence calculator performing a statistical analysis of said fused heart rate estimates to determine a statistical probability of contamination of each of said fused heart rate estimates by said artifactual error.

7. The system of claim 6 wherein said filter circuit is a statistical filter and includes a statistical measure of nominal error contamination affecting said heart rate measurements, said confidence calculator combining said statistical measure of nominal error contamination and said statistical probability of contamination of each of said fused heart rate estimates by said artifactual error to determine said selected fused heart rate estimate.

8. The system of claim 6 wherein said confidence calculator calculates a minimum confidence value for one of said measurement combinations for which all of the physiological signals are assumed to be contaminated by said artifactual error, said confidence calculator further calculating a confidence value for remaining measurement combinations using a worst-case probability of contamination by said artifactual error while also having a minimum probability of said confidence level exceeding said minimum confidence level, said selected fused heart rate estimate having the greatest confidence value.

9. The system of claim 1, further including a statistical model for heart rate variability, said confidence calculator using said statistical model to determine said confidence level and modifying said statistical model following the selection of said selected fused heart rate estimate.

10. The system of claim 1, further including a statistical model for sensor error to characterize susceptibility of said sensors to nominal error contamination, said confidence calculator using said statistical model to determine said confidence level and modifying said statistical model following said selection of said selected fused heart rate estimate.

11. A system for fusion of physiological sensor measurements in a subject having a plurality of sensor coupled thereto, each of the sensors producing a sensor measurement related to a physiological parameter, the system comprising:

a statistical model affecting the fusion of the sensor measurements;

a statistical filter circuit receiving the sensor measurements and said statistical model and producing a fused parameter estimate for each possible combination of the sensor measurements from the plurality of sensors in which each of the sensor measurements in said combinations are considered to be acceptable or unacceptable; and

a confidence calculator coupled to said filter circuit to receive said fused parameter estimates and to determine a confidence level for each of said fused parameter estimates indicative of an accuracy of each of said fused parameter estimates, said confidence calculator selecting one of said fused parameter estimate as a final parameter estimates based on said confidence level.

12. The system of claim 11 wherein said filter circuit also receives a previous estimate of the final parameter estimate selected by said confidence calculator, said filter circuit producing said fused parameter estimates based on the sensor measurements, said statistical model and said previous estimate.

13. The system of claim 11 wherein said statistical model is a parameter variability statistical model characterizing changes in the parameter over time and said filter circuit produces said parameter estimates using said parameter variability statistical model.

14. The system of claim 13, further including a sensor measurement error statistical model characterizing susceptibility of the sensor measurements to nominal error contamination, said filter circuit producing said fused parameter estimates using said parameter variability statistical model and said sensor error statistical model.

15. The system of claim 11 wherein said statistical model is a sensor measurement error statistical model characterizing susceptibility of the sensor measurements to nominal error contamination, said filter circuit producing said parameter estimates using said sensor error statistical model.

16. The system of claim 15, further including a parameter variability statistical model characterizing changes in the parameter over time, said filter circuit producing said parameter estimates using said sensor error statistical model and said parameter variability statistical model.

17. The system of claim 11 wherein said confidence calculator alters said statistical model following the selection of said parameter estimate, whereby the system is adaptive by using said final parameter estimate to alter said statistical model.

18. The system of claim 11 wherein said filter circuit is a Kalman filter for each of said possible combinations, each of said Kalman filters using a previous parameter estimate and said statistical model to produce said fused parameter estimates.

19. The system of claim 18 wherein said statistical model is a sensor measurement error statistical model having a Gaussian probability density function characterizing susceptibility of the sensor measurements to nominal error contamination, said Kalman filter using said Gaussian probability density function to produce said fused parameter estimates.

20. The system of claim 18 wherein said statistical model is a parameter variability model having a Gaussian probability density function characterizing changes in the parameter over time, said Kalman filter using said Gaussian probability density function to produce said fused parameter estimates.

21. The system of claim 11 for use with physiological sensor measurements that are susceptible to contamination by artifactual error, said contaminated physiological sensor measurements being considered unacceptable, said confidence calculator analyzing said fused parameter estimates to determine a statistical probability of error of each of said fused parameter estimates caused by said artifactual error.

22. The system of claim 21 wherein said statistical model is a sensor error statistical model characterizing the susceptibility of the sensor measurements to nominal error contamination, said confidence calculator combining said sensor error statistical model and said statistical probability of contamination of each of said fused parameter estimates by said artifactual error to determine said final parameter estimate.

23. The system of claim 21 wherein said confidence calculator calculates a minimum confidence value for one of said measurement combinations for which all of the physiological sensor measurements are considered unacceptable, said confidence calculator further selecting calculating a confidence value for remaining measurement combinations using a worst-case probability of contamination by said artifactual error while also having a minimum probability of said confidence level exceeding said minimum confidence level, said confidence calculator selecting as said final parameter estimate the fused parameter estimate with a maximum confidence value.

24. A method for fusion of sensor measurements in a subject having a plurality of sensor coupled thereto, each of the sensors producing a sensor measurement related to a physiological parameter, the method comprising the steps of:

generating a statistical model affecting the fusion of the sensor measurements;

producing a fused parameter estimate based on the sensor measurements and said statistical model for each possible combination of the sensor measurements from the plurality of sensors in which each of the sensor measurements in said combinations are considered to be acceptable or unacceptable;

determining a confidence level value for each of said fused parameter estimates indicative of an accuracy of each of said parameter estimates; and

selecting a final parameter estimate based on said confidence level.

25. The method of claim 24 wherein said step of producing said fused parameter estimates bases said fused parameter estimates on the sensor measurements, said statistical model and a previous final parameter estimate.

26. The method of claim 24 wherein said statistical model is a parameter variability statistical model characterizing changes in the parameter over time, and said step of producing said fused parameter estimates uses said parameter variability statistical model.

27. The method of claim 26, further including the step of generating a sensor measurement error statistical model characterizing susceptibility of the sensor measurements to nominal error contamination, and said step of producing said fused parameter estimates uses said parameter variability statistical model and said sensor error statistical model.

28. The method of claim 24 wherein said statistical model is a sensor measurement error statistical model characterizing susceptibility of the sensor measurements to nominal error contamination, and said step of producing said fused parameter estimates uses said sensor measurement error statistical model.

29. The method of claim 28, further including the step of generating a parameter variability statistical model characterizing changes in the parameter over time and said step of producing said fused parameter estimates uses said sensor measurement error statistical model and said parameter variability statistical model.

30. The method of claim 24, further including the step of altering said statistical model following said step of selecting said final parameter estimate, whereby the method is adaptive by using said selected final parameter estimate to alter said statistical model.

31. The method of claim 24 wherein said step of producing said fused parameter estimates is performed by a Kalman filter for each of said possible combinations and uses a previous final parameter estimate and said statistical model in said Kalman filters to produce said fused parameter estimates.

32. The method of claim 31 wherein said statistical model is a sensor error statistical model having a Gaussian probability density function characterizing susceptibility of the sensor measurements to nominal error contamination, and said step of producing said fused parameter estimates in said Kalman filters uses said Gaussian probability density function to produce said fused parameter estimates.

33. The method of claim 31 wherein said statistical model is a parameter variability model having a Gaussian probability density function characterizing changes in the parameter over time, and said step of producing said fused parameter estimates in said Kalman filters uses said Gaussian probability density function to produce said fused parameter estimates.

34. The method of claim 24 for use with sensor measurements that are susceptible to contamination by artifact, said contaminated sensor measurements being considered unacceptable, said step of calculating said confidence level analyzing said fused parameter estimates to determine a statistical probability of contamination of each of said fused parameter estimates by said artifact.

35. The method of claim 34 wherein said statistical model is a sensor measurement error statistical model characterizing the susceptibility of the sensors to nominal error contamination, the method further including the step of combining said sensor measurement error statistical model and said statistical probability of contamination of each of said fused parameter estimates by said artifact to determine said fused parameter estimate, said step of calculating said confidence value using said combined sensor measurement error statistical model and said statistical probability of contamination by said artifact.

36. The method of claim 34 wherein said step of calculating said confidence level calculates a minimum confidence value for one of said measurement combinations for which all physiological sensor measurements are considered unacceptable, the method further including the steps of calculating a confidence value for each of the remaining ones of said measurement combinations using a worst case a priori probability of contamination by said artifact while also having a minimum probability of said confidence level exceeding said minimum confidence level, said step of selecting said final parameter estimate selecting one of said fused parameter estimates with a maximum confidence value.