System and method for performing probabilistic classification and decision support using multidimensional medical image databases

• US 20040193036A1
• Filed: 11/06/2003
• Published: 09/30/2004
• Est. Priority Date: 03/12/2003
• Status: Active Grant

• Current Assignee: Siemens Medical Solutions USA Incorporated (Siemens AG)
• Original Assignee: Siemens Medical Solutions USA Incorporated (Siemens AG)
• Inventors: Gupta, Alok, Ramesh, Visvanathan, Zhou, Xiang Sean, Comaniciu, Dorin, Paine, Diane, Duggirala, Bhavani
• Granted Patent: US 7,458,936 B2
• Time in Patent Office: Days
• US Class Current: 600/407
• CPC Class Codes: A61B 8/0883 for diagnosis of the heart
• CPC Class Codes: G06F 18/214 Generating training patterns; Bootstrap methods, e.g. bagging or boosting
• CPC Class Codes: G06F 18/24147 Distances to closest patterns, e.g. nearest neighbour classification
• CPC Class Codes: G06V 10/764 using classification, e.g. of video objects
• CPC Class Codes: G06V 10/774 Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
• CPC Class Codes: G06V 2201/03 Recognition of patterns in medical or anatomical images
• CPC Class Codes: G16H 30/40 for processing medical images, e.g. editing
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems
• CPC Class Codes: Y10S 128/92 Computer assisted medical diagnostics
• CPC Class Codes: Y10S 128/924 using artificial intelligence

Abstract

A system and method for providing decision support to a physician during a medical examination is disclosed. Data is received from a sensor representing a particular medical measurement. The received data includes image data. The received data and context data is analyzed with respect to one or more sets of training models. Probability values for the particular medical measurement and other measurements to be taken are derived based on the analysis and based on identified classes. The received image data is compared with training images. Distance values are determined between the received image data and the training images, and the training images are associated with the identified classes. Absolute value feature sensitivity scores are derived for the particular medical measurement and other measurements to be taken based on the analysis. The probability values, distance values and absolute value feature sensitivity scores are outputted to the user.

Claims

1. A method for assigning probability classification values to a set of identified classes based on a set of measurements taken during a medical procedure of a patient in order to provide decision support for rendering a medical diagnosis, the method comprising:

receiving data from a sensor representing one or more medical measurements;

analyzing the received data by applying decision rules and training models derived from knowledgebase data and prior physician input;

calculating probability values for the identified classes based on the analysis;

determining sensitivity values for the one or more medical measurements based on the analysis; and

outputting the probability values for each identified class and the sensitivity measurements for the one or more measurements.

2. The method of claim 1 wherein the identified classes include classifications for one or more diseases.

3. The method of claim 2 wherein the one or more disease is Dilated Cardiomyopathy (DCM).

4. The method of claim 1 wherein the identified classes include a non-diseased classification.

5. The method of claim 4 wherein the non-diseased classification is non-Dilated Cardiomyopathy (non-DCM).

6. The method of claim 1 wherein the medical sensor is an ultrasound transducer.

7. The method of claim 1 wherein the medical procedure is an echocardiogram examination.

8. The method of claim 1 wherein the step of analyzing the received data by applying decision rules and training models derived from knowledgebase data and prior physician input further includes using induction algorithms to learn probabilistic models.

9. The method of claim 8 wherein the identified classes are linearly separable classes.

10. The method of claim 9 wherein the induction algorithm uses non-parametric discriminant analysis.

11. The method of claim 10 wherein generative modeling is used in reduced discriminative space to obtain likelihood maps for each class.

12. The method of claim 8 wherein the identified classes have nonlinear class boundaries.

13. The method of claim 12 wherein the induction algorithm uses kernel discriminant analysis.

14. The method of claim 10 wherein generative modeling is used in reduced discriminative space to obtain likelihood maps for each class.

15. The method of claim 13 wherein a Support Vector Model (SVM) is used to perform the kernel discriminant analysis.

16. The method of claim 1 wherein a user uses the probability values to assist in making a medical diagnosis.

17. The method of claim 1 wherein the probability values are provided in real time.

28. The method of claim 1 wherein a user uses the distance values to assist in making a medical diagnosis.

29. The method of claim 1 wherein the distance values are provided in real time.

35. The method of claim 1 wherein the absolute value feature sensitivity scores are provided in real time.

18. A method for comparing an image corresponding to a case in question to a set of training images based on similar content in the images during a medical procedure for a patient in order to provide a medical diagnosis, the method comprising:

receiving image data from a sensor representing a particular medical measurement;

comparing the received image data with training images;

deriving distance values between the received image data and the training images; and

outputting a set of images that have shortest distance measurements and corresponding distance values.

19. The method of claim 18 wherein the identified classes include classifications for one or more diseases.

20. The method of claim 19 wherein the one or more disease is Dilated Cardiomyopathy (DCM).

21. The method of claim 18 wherein the identified classes include a non-diseased classification.

22. The method of claim 18 further comprising the steps of:

receiving an indication from a physician of a selection of a subset of the set of images having the shortest distance measurements;

deriving a new set of distance measurements between the received image data and the training images based on the selected subset of images; and

outputting a second set of images that have shortest distance measurements and corresponding distance values.

23. The method of claim 18 wherein the medical sensor is an ultrasound transducer.

24. The method of claim 18 wherein the medical procedure is an echocardiogram examination.

25. The method of claim 18 wherein the distance measurement is defined in Cartesian space.

26. The method of claim 18 wherein the distance measurement is defined in a reduced discriminative space.

27. The method of claim 18 wherein the training models are selected by a user.

30. A method for assigning feature sensitivity values to a set of measurements to be taken during a medical procedure of a patient in order to provide a medical diagnosis, the method comprising:

receiving data from, a sensor representing a particular medical measurement;

analyzing the received data and context data with respect to one or more sets of training models;

deriving absolute value feature sensitivity scores for the particular medical measurement and other measurements to be taken based on the analysis; and

outputting the absolute value feature sensitivity scores.

31. The method of claim 30 wherein the absolute value feature sensitivity score is determined by:

S_i=log(K)log(τ

IG_i+1)/log(τ

+1)

where S_i is the output value;

log(K) is the maximum absolute value;

τ

is a parameter that controls the degree of enhancement of the small values; and

IG_i is the input value.

32. The method of claim 30 wherein the medical sensor is an ultrasound transducer.

33. The method of claim 30 wherein the medical procedure is an echocardiogram examination.

34. The method of claim 30 wherein a user uses the absolute value feature sensitivity scores to assist in making a medical diagnosis.

36. A method for providing decision support to a physician during a medical examination, the method comprising the steps of:

receiving data from a sensor representing a particular medical measurement, said received data including image data;

analyzing the received data and context data with respect to one or more sets of training models;

deriving probability values for a set of identified classes based on the analysis;

comparing the received image data with training images;

deriving distance values between the received image data and the training images;

deriving absolute value feature sensitivity scores for the particular medical measurement and other measurements to be taken based on the analysis; and

outputting the probability values, a set of images that have shortest distance measurements and corresponding distance values and absolute value feature sensitivity scores for the particular measurement and other measurements.

37. The method of claim 36 wherein the identified classes include classifications for one or more diseases.

38. The method of claim 37 wherein the one or more disease is Dilated Cardiomyopathy (DCM).

39. The method of claim 36 wherein the identified classes include a non-diseased classification.

40. The method of claim 36 further comprising the steps of:

receiving an indication from a physician of a selection of a subset of the set of images having the shortest distance measurements;

deriving a new set of distance measurements between the received image data and the training images based on the selected subset of images; and

outputting a second set of images that have shortest distance measurements and corresponding distance values.

41. The method of claim 36 wherein the medical sensor is an ultrasound transducer.

42. The method of claim 36 wherein the medical procedure is an echocardiogram examination.

43. The method of claim 36 wherein the step of analyzing the received data and context data with respect to one or more sets of training models further includes using induction algorithms to learn probabilistic models.

44. The method of claim 43 wherein the identified classes are linearly separable classes.

45. The method of claim 44 wherein the induction algorithm uses non-parametric discriminant analysis.

46. The method of claim 45 wherein generative modeling is used in reduced discriminative space to obtain likelihood maps for each class.

47. The method of claim 43 wherein the identified classes have nonlinear class boundaries.

49. The method of claim 46 wherein generative modeling is used in reduced discriminative space to obtain likelihood maps for each class.

50. The method of claim 49 wherein a Support Vector Model (SVM) is used to perform the kernel discriminant analysis.

51. The method of claim 36 wherein a user uses the probability values to assist in making a medical diagnosis.

52. The method of claim 36 wherein the probability values are provided in real time.

53. The method of claim 36 wherein the distance measurement is defined in Cartesian space.

54. The method of claim 36 wherein the distance measurement is defined in a reduced discriminative space.

55. The method of claim 36 wherein the training models are selected by a user.

56. The method of claim 36 wherein the absolute value feature sensitivity score is determined by:

S_i=log(K)log(τ

IG_i+1)/log(τ

+1)

where S_i is the output value;

log(K) is the maximum absolute value;

τ

is a parameter that controls the degree of enhancement of the small values; and

IG_i is the input value.

48. The method of claim 48 wherein the induction algorithm uses kernel discriminant analysis.

57. A Computer Aided Diagnosis (CAD) system comprises:

means for receiving data from a sensor representing a particular medical measurement, said received data including image data;

means for analyzing the received data and context data with respect to one or more sets of training models;

means for deriving probability values for a set of identified classes based on the analysis;

means for comparing the received image data with training images;

means for deriving distance values between the received image data and the training images, the training images being associated with identified classes;

means for deriving absolute value feature sensitivity scores for the particular medical measurement and other measurements to be taken based on the analysis; and

means for outputting the probability values, a set of images that have shortest distance measurements and corresponding distance values and absolute value feature sensitivity scores for the particular measurement and other measurements.

58. The system of claim 57 wherein the identified classes include classifications for one or more diseases.

59. The system of claim 58 wherein the one or more disease is Dilated Cardiomyopathy (DCM).

60. The system of claim 57 wherein the identified classes include a non-diseased classification.

61. The system of claim 57 further comprising the steps of:

receiving an indication from a physician of a selection of a subset of the set of images having the shortest distance measurements;

deriving a new set of distance measurements between the received image data and the training images based on the selected subset of images; and

outputting a second set of images that have shortest distance measurements and corresponding distance values.

62. The system of claim 57 wherein the medical sensor is an ultrasound transducer.

63. The system of claim 57 wherein the medical procedure is an echocardiogram examination.

64. The system of claim 57 wherein the step of analyzing the received data and context data with respect to one or more sets of training models further includes using induction algorithms to learn probabilistic models.

65. The system of claim 64 wherein the identified classes are linearly separable classes.

66. The system of claim 65 wherein the induction algorithm uses non-parametric discriminant analysis.

67. The system of claim 66 wherein generative modeling is used in reduced discriminative space to obtain likelihood maps for each class.

68. The system of claim 64 wherein the identified classes have nonlinear class boundaries.

69. The system of claim 68 wherein the induction algorithm uses kernel discriminant analysis.

70. The system of claim 67 wherein generative modeling is used in reduced discriminative space to obtain likelihood maps for each class.

71. The system of claim 69 wherein a Support Vector Model (SVM) is used to perform the kernel discriminant analysis.

72. The system of claim 57 wherein a user uses the probability values to assist in making a medical diagnosis.

73. The system of claim 57 wherein the probability values are provided in real time.

74. The system of claim 57 wherein the distance measurement is defined in Cartesian space.

75. The system of claim 57 wherein the distance measurement is defined in a reduced discriminative space.

76. The system of claim 57 wherein the training models are selected by a user.

77. The system of claim 57 wherein the absolute value feature sensitivity score is determined by:

S_i=log(K)log(τ

IG_i+1)/log(τ

+1)

where S_i is the output value;

log(K) is the maximum absolute value;

τ

is a parameter that controls the degree of enhancement of the small values; and

IG_i is the input value.