Method and system for modeling cardiovascular disease using a probability regession model

• US 7,462,153 B2
• Filed: 07/23/2004
• Issued: 12/09/2008
• Est. Priority Date: 07/23/2004
• Status: Active Grant

• Current Assignee: AusculSciences, Inc.
• Original Assignee: SonoMedica, Inc.
• Inventors: Mohler, Sailor H., Bostian, Aj A., Bostian, Sidney E. Jr.
• Primary Examiner(s): Layno; Carl H.
• Assistant Examiner(s): Gedeon; Brian T
• Application Number:
  US10/897,121
• Publication Number:
  US 20060020220A1
• Time in Patent Office: 1,600 Days
• Field of Search: 600/508, 600/513, 600/514, 600/528
• US Class Current: 600/508
• CPC Class Codes: G16H 50/50 for simulation or modelling of medical disorders
• CPC Class Codes: G16Z 99/00 Subject matter not provided for in other main groups of this subclass

Abstract

A method and system for modeling cardiovascular disease using a probability regression model is provided. A parameter estimate of a probability regression model for cardiovascular disease can be generated using predictors derived from cardiovascular sound signals and disease status information. A probability of cardiovascular disease can be generated using a probability regression model that includes a predictor derived from cardiovascular sound signals.

Claims

1. A processor-readable medium comprising code representing instructions to cause a processor to:

generate at least one parameter estimate of a probability regression model for cardiovascular disease at least partially based on (1) a plurality of predictors generated from cardiovascular sound signals of a plurality of subjects and (2) disease status information for cardiovascular disease of the plurality of subjects;

determine a probability of cardiovascular disease for a new subject with the probability regression model, the probability of cardiovascular disease being generated at least partially based on the at least one parameter estimate and another plurality of predictors generated from cardiovascular sound signals of the new subject; and

update the at least one parameter estimate at least partially based on disease status information associated with the new subject.

2. The processor-readable medium of claim 1, further comprising code representing instructions to cause a processor to:

determine a probability of cardiovascular disease for a second new subject with the probability regression model, the probability of cardiovascular disease for the second new subject being generated at least partially based on the updated at least one parameter estimate.

3. A processor-readable medium comprising code representing instructions to cause a processor to:

generate a probability of cardiovascular disease using a probability regression model;

$\mathrm{Pr}<br><br>\left(Y❘<br><br>X,b\right)=\frac{1}{\sqrt{2<br><br>\mathrm{\pi <br><br>}}}<br><br>{\int <br><br>}_{s=-\mathrm{\infty <br><br>}}^{X^{\mathrm{\prime <br><br>}}<br><br>b}<br><br>\exp \left(-\frac{s^2}{2}\right)<br><br>d<br><br>s;<br><br>$

wherein X is at least one predictor generated from cardiovascular sound signals;

wherein b is at least one parameter estimate previously generated from the probability regression model based on cardiovascular sound signals of a plurality of subjects; and

wherein Pr(Y|X, b) is the probability of the existence of cardiovascular disease Y given the at least one predictor X and the at least one previously generated parameter estimate b.

4. The processor-readable medium of claim 3, wherein the at least one previously generated parameter estimate b is previously generated from the probability regression model also based on clinical data of the plurality of subjects.

5. A processor-readable medium comprising code representing instructions to cause a processor to:

generate at least one parameter estimate of a probability regression model for cardiovascular disease at least partially based on (1) a predictor generated from cardiovascular sound signals of at least one subject and (2) disease status information for cardiovascular disease;

wherein the probability regression model is one of a logit model, a multinomial probit model, a multinomial logit model, an ordered probit model, an ordered logit model, a Weibull model, a Cox proportional hazards model, an exponential model, a log-logistic model, a lognormal model, and a Kaplan-Meier model.

6. The processor-readable medium of claim 5, further comprising code representing instructions to cause a processor to:

update the at least one parameter estimate at least partially based on information associated with a new subject.

7. The processor-readable medium of claim 5, wherein the cardiovascular disease is coronary artery disease.

8. The processor-readable medium of claim 5, wherein the cardiovascular disease is one of carotid artery disease, congestive heart failure, hypertension, and bundle branch block.

9. The processor-readable medium of claim 5, wherein the probability regression model is a model of a current cardiovascular disease event.

10. The processor-readable medium of claim 5, wherein the probability regression model is a model of a future cardiovascular disease event.

11. The processor-readable medium of claim 5, wherein the predictor is indicative of bruits present in the cardiovascular sound signals of the at least one subject.

12. The processor-readable medium of claim 5, wherein the predictor is indicative of a third heart sound present in the cardiovascular sound signals of the at least one subject.

13. The processor-readable medium of claim 5, wherein the predictor is indicative of a fourth heart sound present in the cardiovascular sound signals of the at least one subject.

14. The processor-readable medium of claim 5, wherein the predictor is indicative of peak acoustic power present in a diastolic interval of the cardiovascular sound signals of the at least one subject.

15. The processor-readable medium of claim 5, wherein the predictor is one of a plurality of predictors, each predictor from the plurality of predictors being generated from the cardiovascular sound signals of the at least one subject.

16. The processor-readable medium of claim 15, the plurality of predictors being different ones of:

a predictor indicative of bruits present in the cardiovascular sound signals of the at least one subject;

a predictor indicative of a third heart sound present in the cardiovascular sound signals of the at least one subject;

a predictor indicative of a fourth heart sound present in the cardiovascular sound signals of the at least one subject; and

a predictor indicative of peak acoustic power present in a diastolic interval of the cardiovascular sound signals of the at least one subject.

17. The processor-readable medium of claim 5, wherein the generating of the at least one parameter estimate also uses at least one supplemental predictor that is not generated from the cardiovascular sound signals of the at least one subject.

18. The processor-readable medium of claim 17, wherein the at least one supplemental predictor includes a heart rate of the at least one subject.

19. The processor-readable medium of claim 17, wherein the at least one supplemental predictor includes at least one of:

a sex of the at least one subject;

an age of the at least one subject;

a body mass index of the at least one subject;

a cholesterol level of the at least one subject;

a C-reactive protein level of the at least one subject; and

a socioeconomic classification of the at least one subject.

20. The processor-readable medium of claim 17, wherein the at least one supplemental predictor includes at least one of:

sex, age, diabetic state, cholesterol level, C-reactive protein level, chest morphology data, chest hair data, breast size data, body mass index, and blood viscosity data.

21. The processor-readable medium of claim 5, wherein the disease status information for cardiovascular disease includes a categorical indication of a degree of cardiovascular disease.

22. The processor-readable medium of claim 5, wherein the disease status information for cardiovascular disease includes at least one of:

information from a catheterization analysis;

information from an electrocardiogram analysis;

information from a stress-test analysis;

information from an echocardiographic analysis;

information from an electron-beam tomography analysis; and

information from a magnetic resonance imaging analysis.

23. The processor-readable medium of claim 5, wherein the parameter estimate is generated by a method of maximum likelihood.

24. The processor-readable medium of claim 5, wherein the parameter estimate is generated by one of a generalized method of moments, a simulation method, and an exact method.

25. The processor-readable medium of claim 5, wherein the at least one generated parameter estimate is one of a plurality of generated parameter estimates of a probability regression model for cardiovascular disease.

26. The processor-readable medium of claim 25, wherein the predictor is one of a plurality of predictors generated from cardiovascular sound signals of the at least one subject.

27. The processor-readable medium of claim 25, wherein the predictor is one of a plurality of predictors generated from cardiovascular sound signals of a plurality of different subjects.

28. The processor-readable medium of claim 27, wherein the plurality of different subjects include at least a first group of subjects and a second group of subjects, the first group of subjects having a common first characteristic and the second group of subjects having a common second characteristic, the first characteristic being different from the second characteristic.

29. The processor-readable medium of claim 28, the first characteristic and the second characteristic being different ones of:

sex, age, diabetic status, and smoking status.

30. The processor-readable medium of claim 5, wherein the at least one subject is a plurality of subjects.

31. The processor-readable medium of claim 5, the at least one subject being a plurality of subjects, the at least one parameter estimate being a first parameter estimate for the plurality of subjects, the processor-readable medium further comprising code representing instructions to cause a processor to update the method of modeling cardiovascular disease by:

generating a second parameter estimate of the probability regression model for cardiovascular disease using (1) a new set of predictors generated from cardiovascular sound signals of a new set of subjects and (2) clinical data of the new set of subjects, the clinical data including disease status information for cardiovascular disease.

32. A processor-readable medium comprising code for predicting cardiovascular disease, the code representing instructions to cause a processor to:

generate a probability of cardiovascular disease using a probability regression model, the probability regression model using at least one predictor generated from cardiovascular sound signals;

wherein the probability regression model is one of a logit model, a multinomial probit model, a multinomial logit model, an ordered probit model, an ordered logit model, a Weibull model, a Cox proportional hazards model, an exponential model, a log-logistic model, a lognormal model, and a Kaplan-Mejer model.

33. The processor-readable medium of claim 32, wherein the cardiovascular disease is coronary artery disease.

34. The processor-readable medium of claim 32, wherein the cardiovascular disease is one of carotid artery disease, congestive heart failure, hypertension, and bundle branch block.

35. The processor-readable medium of claim 32, wherein the probability regression model is a model of a current cardiovascular disease event.

36. The processor-readable medium of claim 32, wherein the probability regression model is a model of a future cardiovascular disease event.

37. The processor-readable medium of claim 32, wherein the at least one predictor is indicative of bruits present in the cardiovascular sound signals.

38. The processor-readable medium of claim 32, wherein the at least one Predictor is indicative of a third heart sound present in the cardiovascular sound signals.

39. The processor-readable medium of claim 32, wherein the at least one predictor is indicative of a fourth heart sound present in the cardiovascular sound signals.

40. The processor-readable medium method of claim 32, wherein the at least one predictor is indicative of peak acoustic power present in a diastolic interval of the cardiovascular sound signals.

41. The processor-readable medium of claim 32, wherein the at least one predictor includes a plurality of predictors each generated from the cardiovascular sound signals.

42. The processor-readable medium of claim 41, the plurality of predictors being different ones of:

a predictor indicative of bruits present in the cardiovascular sound signals;

a predictor indicative of a third heart sound present in the cardiovascular sound signals;

a predictor indicative of a fourth heart sound present in the cardiovascular sound signals; and

a predictor indicative of peak acoustic power present in a diastolic interval of the cardiovascular sound signals.

43. The processor-readable medium of claim 32, wherein the code representing instructions to cause a processor to generate the probability of cardiovascular disease also uses at least one supplemental predictor that is not generated from the cardiovascular sound signals.

44. The processor-readable medium of claim 43, wherein the at least one supplemental predictor that is not generated from the cardiovascular sound signals includes a heart rate.

45. The processor-readable medium of claim 43, wherein the at least one supplemental predictor that is not generated from the cardiovascular sound signals includes at least one of a sex, an age, a body mass index, a cholesterol level, a C-reactive protein level, and a socioeconomic indicator.

46. The processor-readable medium of claim 32, wherein the at least one supplemental predictor is also generated from clinical data.

47. The processor-readable medium of claim 46, wherein the clinical data includes at least one of sex, age, diabetic state, cholesterol level, C-reactive protein level, chest morphology data, chest hair data, breast size data, body mass index, and blood viscosity data.

48. The processor-readable medium of claim 32, wherein the probability regression model includes at least one previously generated parameter estimate.

49. The processor-readable medium of claim 48, wherein the cardiovascular sound signals are first cardiovascular sound signals, the parameter estimate being generated using at least another predictor that is generated from second cardiovascular sound signals that are different than the first cardiovascular sound signals.

50. The processor-readable medium of claim 46, wherein the parameter estimate is also generated using clinical data that at least includes disease status information for cardiovascular disease.

51. A system comprising:

a processor configured to use a probability regression model for cardiovascular disease; and

a parameter estimate generator configured to generate a parameter estimate for the probability regression model using (1) a predictor generated from cardiovascular sound signals of at least one subject and (2) disease status information for cardiovascular disease;

wherein the probability regression model is one of a logit model, a multinomial probit model, a multinomial logit model, an ordered probit model, an ordered logit model, a Weibull model, a Cox proportional hazards model, an exponential model, a log-logistic model, a lognormal model, and a Kaplan-Meier model.

52. The system of claim 51, wherein the processor is further configured to determine a probability of cardiovascular disease using the probability regression model and the parameter estimate.

53. A system comprising:

a memory configured to store a probability regression model for cardiovascular disease; and

a probability generator configured to generate a probability of cardiovascular disease using the probability regression model, the probability regression model using at least one predictor generated from cardiovascular sound signals;

wherein the probability regression model is one of a logit model, a multinomial probit model, a multinomial logit model, an ordered probit model, an ordered logit model, a Weibull model, a Cox proportional hazards model, an exponential model, a log-logistic model, a lognormal model, and a Kaplan-Meier model.