Cochlear implant MAP optimization with use of a genetic algorithm

• US 6,879,860 B2
• Filed: 03/11/2003
• Issued: 04/12/2005
• Est. Priority Date: 03/11/2003
• Status: Active Grant

• Current Assignee: Cochlear Limited
• Original Assignee: Cochlear Limited
• Inventors: Honert, Christopher van den, Wakefield, Gregory Howard
• Primary Examiner(s): Schaetzle, Kennedy
• Application Number:
  US10/385,880
• Publication Number:
  US 20040181266A1
• Time in Patent Office: 763 Days
• Field of Search: 607 55- 57, 607/59, 600/559, 381/23.1, 381/312, 381/314, 128/920, 128/924, 128/925, 735/85
• US Class Current: 607/57
• CPC Class Codes: A61N 1/0541 Cochlear electrodes
• CPC Class Codes: A61N 1/36039 fitting procedures

Abstract

A method for fitting a cochlear implant to a patient. A genetic algorithm operates to generate successive generations of multiple groups of values for a parameter subset, and patient feedback during execution of the genetic algorithm determines the multiple groups of subset values in successive generations. In each generation, half of the groups of values for the parameter subset are selected and used to determine the groups of values for the next generation. Values for the parameters not included in the subset are selected by any traditional method that does not use a genetic algorithm. In one illustrative embodiment of the invention, there are only three parameters in the subset that are selected by a genetic algorithm—parameters relating to rate, number of channels and filtering characteristics.

Claims

1. A method of at least partially fitting a cochlear implant to a patient comprising the following acts:

(a) executing a genetic algorithm to select values for a subset of all the parameters for which values must be selected to fit the implant, the genetic algorithm operating to generate successive generations of multiple groups of values for said parameter subset, and

(b) using patient feedback during execution of the genetic algorithm to determine the multiple groups of parameter subset values in successive generations.

2. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 1 further including the following act:

(c) selecting values for the parameters not included in said subset by a method that does not use a genetic algorithm.

3. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 2 wherein said patient feedback involves selecting, without ranking, preferred groups of parameter subset values in one generation from which the groups of parameter subset values in the next generation are determined.

4. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 3 wherein the values of said subset of parameters are represented by a binary string of 6-12 bits.

5. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 3 wherein there are 6-12 groups of parameter subset values in each generation.

6. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 3 wherein the values of said subset of parameters are represented by a string of bits, with different sub-strings of bits in said string representing values of respective parameters.

7. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 6 wherein the parameters in said subset are related to rate, number of channels and filtering characteristics.

8. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 3 wherein the values of said subset of parameters are represented by a string of bits, each different valued string identifying a predetermined group of parameter values.

9. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 8 wherein the parameters in said subset are related to rate, number of channels and filtering characteristics.

10. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 3 wherein the values of said subset of parameters are represented by a string of bits, some of the bits in said string being divided into separate sub-strings each of which represents the value of a respective parameter, and the others of the bits in said string identifying a predetermined group of parameter values.

11. A method of at least partially fitting a cochlear implant to a patient comprising the steps of:

(a) executing a genetic algorithm to select values for a subset of all the parameters for which values must be selected to fit the implant, the genetic algorithm operating to generate successive generations of multiple groups of values for said parameter subset, and

(b) based on the subjective listening judgment of the patient during execution of the genetic algorithm, selecting substantially less than all of the groups of values for said parameter subset in each generation and determining from them a larger number of groups of values for the next generation.

12. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 11 wherein N groups of values for said parameter subset are determined for each generation, and N/2 of said groups are selected in each generation for the determination of the N groups in the next generation.

13. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 12 wherein each of the N/2 of said groups selected in each generation is paired with each of two others of said groups, with each pair being used for the determination of two groups in the next generation.

14. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 13 wherein the patient selects N/2 groups in each generation and the selected groups are then used to determine the groups in the next generation independent of any rankings among the N/2 groups.

15. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 11 wherein the selected groups in each generation are used to determine the groups in the next generation independent of any rankings among the selected groups.

16. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 11 wherein there are 6-12 groups of parameter subset values in each generation.

17. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 11 wherein the values of said subset of parameters are represented by a string of bits, the string of bits being divided into separate sub-strings each of which represents the value of a respective parameter.

18. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 17 wherein the string of bits for representing each group of values for the next generation is determined by combining parts of two different strings of bits that represent groups of values for the current generation.

19. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 18 wherein the parts of strings of bits that are combined are cut from strings at positions that separate parameter values.

20. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 17 wherein the parts of strings of bits that are combined are cut from strings at positions that do not vary from generation to generation.

21. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 17 wherein the parts of strings of bits that are combined are cut from strings at positions that vary from generation to generation.

22. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 11 wherein the genetic algorithm is executed a fixed number of times to generate a fixed number of successive generations of multiple groups of values for said parameter subset, and a final group of values for the parameter subset is selected from the groups in the final generation.

23. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 11 wherein a record is made of any group of values for said parameter subset that the patient identifies as satisfactory in any generation, and a final group of values for the parameter subset is selected from the groups in the final generation and those previously identified by the patient.

24. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 11 wherein the parameters in each subset are related to rate, number of channels and filtering characteristics.

25. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 11 wherein successive generations of multiple groups of values for said parameter subset are determined in part by mutations.

26. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 11 wherein the values for said subset of parameters are represented by a string of bits, each string of bits being associated with a predetermined set of parameter values.

27. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 11 wherein the values for said subset of parameters are represented by a string of bits, and successive generations of multiple groups of values for said parameter subset are determined in part by mutations.

28. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 27 wherein in each generation the probability of each bit in a string of bits mutating is fixed for all bits.

29. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 27 wherein, in any generation in which mutations are allowed, the probability of each bit in a string of bits mutating is between 2% and 4%.

30. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 11 wherein the multiple groups of values for said parameter subset that are used in the first generation of the execution of the genetic algorithm are based on predetermined clinician judgment.

31. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 11 wherein the multiple groups of values for said parameter subset that are used in the first generation of the execution of the genetic algorithm is random.

32. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 11 wherein the multiple groups of values for said parameter subset that are used in the first generation of the execution of the genetic algorithm is based at least partially on the results of an earlier execution of the genetic algorithm.

33. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 32 wherein the parameters in said subset are related to rate, number of channels and filtering characteristics.

34. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 11 wherein the values of said subset of parameters are represented by a string of bits, some of the bits in said string being divided into separate sub-strings each of which represents the value of a respective parameter, and the others of the bits in said string identifying a predetermined group of parameter values.

35. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 11 further including the step of:

(c) selecting values for the parameters not included in said subset by a method that does not use a genetic algorithm.

36. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 35 wherein said patient selects preferred groups of parameter subset values in one generation from which the groups of parameter subset values in the next generation are determined.

37. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 36 wherein said patient selects preferred groups of parameter subset values by classifying them as selected or unselected without ranking them.

38. A method of at least partially fitting a cochlear implant to a patient comprising the steps of:

(a) executing a genetic algorithm to select values for at least a subset of all the parameters for which values must be selected to fit the implant, the genetic algorithm operating to generate successive generations of multiple groups of values for said parameter subset, and

(b) using patient feedback during execution of the genetic algorithm to select a specified number of groups of values to survive in each generation from which multiple groups of values for said parameter subset are determined for the next generation.

39. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 38 wherein N groups of values for said parameter subset are determined for each generation, and N/2 of said groups survive in each generation from which the N groups in the next generation are determined.

40. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 39 wherein each of the N/2 of said groups that survive in each generation is paired with each of two others of said groups, with each pair being used for the determination of two groups in the next generation.

41. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 38 wherein the patient selects N/2 groups in each generation and the selected groups are then used to determine the groups in the next generation independent of any rankings among the N/2 groups.

42. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 38 wherein the surviving groups in each generation are used to determine the groups in the next generation independent of any rankings among the surviving groups.

43. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 38 wherein there are 6-12 groups of parameter subset values in each generation.

44. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 38 wherein the values of said subset of parameters are represented by a string of bits, the string of bits being divided into separate sub-strings each of which represents the value of a respective parameter.

45. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 44 wherein the string of bits for representing each group of values for the next generation is determined by combining parts of two different strings of bits that represent surviving groups of values for the current generation.

46. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 45 wherein the parts of strings of bits that are combined are cut from strings at positions that separate parameter values.

47. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 45 wherein the parts of strings of bits that are combined are cut from strings at positions that do not vary from generation to generation.

48. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 45 wherein the parts of strings of bits that are combined are cut from strings at positions that vary from generation to generation.

49. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 38 wherein the genetic algorithm is executed a fixed number of times to generate a fixed number of successive generations of multiple groups of values for said parameter subset, and a final group of values for the parameter subset is selected from the groups in the final generation.

50. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 38 wherein a record is made of any group of values for said parameter subset that the patient identifies as satisfactory in any generation, and a final group of values for the parameter subset is selected from the groups in the final generation and those previously identified by the patient.

51. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 38 wherein the parameters in each subset are related to rate, number of channels and filtering characteristics.

52. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 38 wherein successive generations of multiple groups of values for said parameter subset are determined in part by mutations.

53. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 38 wherein the values for said subset of parameters are represented by a string of bits, each string of bits being associated with a predetermined set of parameter values.

54. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 38 wherein the values for said subset of parameters are represented by a string of bits, and successive generations of multiple groups of values for said parameter subset are determined in part by mutations.

55. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 54 wherein in each generation the probability of each bit in a string of bits mutating is the same for all bits.

56. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 54 wherein, in any generation in which mutations are allowed, the probability of each bit in a string of bits mutating is between 2% and 4%.

57. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 38 wherein the multiple groups of values for said parameter subset that are used in the first generation of the execution of the genetic algorithm is based on predetermined clinician judgment.

58. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 38 wherein the multiple groups of values for said parameter subset that are used in the first generation of the execution of the genetic algorithm is random.

59. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 38 wherein the values of said subset of parameters are represented by a string of bits, with different sub-strings of bits in said string representing values of respective parameters.

60. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 59 wherein the parameters in said subset are related to rate, number of channels and filtering characteristics.

61. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 38 wherein the values of said subset of parameters are represented by a string of bits, some of the bits in said string being divided into separate sub-strings each of which represents the value of a respective parameter, and the others of the bits in said string identifying a predetermined group of parameter values.

62. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 38 further including the step of:

(c) selecting values for the parameters not included in said subset by a method that does not use a genetic algorithm.

63. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 62 wherein said patient selects preferred groups of parameter subset values in one generation from which the groups of parameter subset values in the next generation are determined.

64. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 63 wherein said patient selects preferred groups of parameter subset values by classifying them as selected and unselected without ranking them.

65. A method of at least partially fitting a cochlear implant to a patient comprising the steps of forming a set of digit sequences, each sequence being divided into groups which represent the values of respective parameters for the implant;

testing the implant on the patient using the values represented in said sequences and selecting the sequences that provide superior performance;

combining parts of selected sequences to form new sequences; and

iteratively repeating the testing, selecting and combining steps to derive a sequence for use in finally fitting the implant.

66. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 65 wherein sequences are selected based on patient feedback without ranking sequences.

67. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 65 wherein said sequences comprise 6-12 binary digits.

68. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 67 wherein there are 6-12 sequences in each iteration.

69. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 65 wherein there are 6-12 sequences in each iteration.

70. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 65 wherein the parameters whose values are represented in said sequences are related to rate, number of channels and filtering characteristics.

71. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 65 wherein the values of respective parameters for the implant in a sequence are represented by a string of bits, the string of bits being divided into separate sub-strings each of which represents the value of a respective parameter.

72. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 71 wherein the string of bits for representing a value in a sequence is determined by combining parts of two different strings of bits each of which represents one or more parameter values.

73. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 72 wherein the parts of strings of bits that are combined are cut from strings at positions that separate parameter values.

74. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 73 wherein the parts of strings of bits that are combined are cut from strings at positions that do not vary from generation to generation.

75. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 73 wherein the parts of strings of bits that are combined are cut from strings at positions that vary from generation to generation.

76. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 65 wherein a record is made of any sequence that is identified as being satisfactory, and a final sequence is selected from the last set of sequences that is used in testing and those previously identified as being satisfactory.

77. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 65 wherein the parameters represented in each sequence are related to rate, number of channels and filtering characteristics.

78. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 65 wherein new sequences are determined in part by randomly changing digits.

79. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 65 wherein the values of respective parameters for the implant in a sequence are represented by a string of bits, each string of bits being associated with a predetermined set of parameter values.

80. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 65 wherein the values of respective parameters for the implant in a sequence are represented by a string of bits, and successive sequences are determined in part by random changes.

81. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 80 wherein the probability of each bit in a string of bits being randomly changed is fixed for all bits.

82. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 81 wherein the probability of each bit in a string of bits being randomly changed is between 2% and 4%.

83. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 65 wherein the sequences that are first used are based on predetermined clinician judgment.

84. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 65 wherein the sequences that are first used are random.

85. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 65 wherein the values of respective parameters for the implant are represented in a sequence by a string of bits, with different sub-strings of bits in said string representing values of respective parameters.

86. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 85 wherein said parameters are related to rate, number of channels and filtering characteristics.

87. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 65 wherein the values of respective parameters for the implant are represented in a sequence by a string of bits, some of the bits in said string being divided into separate sub-strings each of which represents the value of a respective parameter, and the others of the bits in said string identifying a predetermined group of parameter values.

88. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 65 wherein said patient selects preferred groups of sequences from which parts are combined to form new sequences.

89. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 88 wherein said patient selects preferred groups of sequences by classifying them as selected or unselected without ranking them.

90. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 65 wherein sequences are selected based at least in part on patient feedback.

91. A method of at least partially fitting a cochlear implant to a patient comprising:

(a) selecting values for at least a subset of all the parameters for which values must be selected to fit the implant by generating successive generations of multiple groups of values for said parameter subset, and

(b) for each generation using patient feedback to select a specified number of groups of values from which multiple groups of values for said parameter subset are determined for the next generation.

92. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 91 wherein groups of values in each generation are used to determine the groups of values in the next generation independent of any rankings among the groups in said each generation.

93. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 91 wherein there are 6-12 groups of parameter subset values in each generation.

94. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 91 wherein the values of said subset of parameters are represented by a string of bits, and the string of bits for representing each group of values for the next generation is determined by combining parts of two different strings of bits that represent groups of values for the current generation.

95. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 91 wherein multiple groups of values for said parameter subset are selected for a fixed number of successive generations, and a final group of values for the parameter subset is selected from the groups in the final generation.

96. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 91 wherein a record is made of any group of values for said parameter subset that the patient identifies as satisfactory in any generation, and a final group of values for the parameter subset is selected from the groups in the final generation and those previously identified by the patient.

97. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 91 wherein the parameters in each subset are related to stimulation rate, number of electrodes used, number of maxima per stimulus frame, steepness of output compression, and combination of input audio filtering options.

98. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 91 wherein successive generations of multiple groups of values for said parameter subset are determined in part by mutations.

99. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 91 wherein the values for said subset of parameters are represented by a string of bits, and successive generations of multiple groups of values for said parameter subset are determined in part by mutations.

100. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 99 wherein in each generation the probability of each bit in a string of bits mutating is the same for all bits.

101. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 99 wherein, in any generation in which mutations are allowed, the probability of each bit in a string of bits mutating is between 2% and 4%.

102. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 91 wherein the multiple groups of values for said parameter subset that are used in the first generation are based on predetermined clinician judgment.

103. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 91 wherein the multiple groups of values for said parameter subset that are used in the first generation are random.

104. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 91 wherein the multiple groups of values for said parameter subset that are used in the first generation are based on the results of a previous use of the method on the same patient.

105. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 91 wherein said patient selects preferred groups of parameter subset values in one generation from which the groups of parameter subset values in the next generation are determined.

106. A method of at least partially fitting a cochlear implant to a patient in accordance with claim 105 wherein said patient selects preferred groups of parameter subset values by classifying them as selected and unselected without ranking them.