Adaptive method and apparatus for forecasting and controlling neurological disturbances under a multi-level control

• US 20020103512A1
• Filed: 12/12/2000
• Published: 08/01/2002
• Est. Priority Date: 12/12/2000
• Status: Active Grant

• Current Assignee: Trustees Of The University Of Pennsylvania (University of Pennsylvania)
• Original Assignee: Trustees Of The University Of Pennsylvania (University of Pennsylvania)
• Inventors: Litt, Brian, Vachtsevanos, George John, Echauz, Javier Ramon, Esteller, Rosana
• Granted Patent: US 6,594,524 B2
• Time in Patent Office: Days
• US Class Current: 607/9
• CPC Class Codes: A61B 5/076 Permanent implantations telemetry using implanted circuitry A61B5/0031; implanted stimulators for therapy A61N1/3605, A61N1/362, A61N1/372
• CPC Class Codes: A61B 5/369 Electroencephalography [EEG] devices for psychotechnics A61B5/16
• CPC Class Codes: A61B 5/37 Intracranial electroencephalography [IC-EEG], e.g. electrocorticography [ECoG]
• CPC Class Codes: A61B 5/372 Analysis of electroencephalograms
• CPC Class Codes: A61B 5/375 using biofeedback
• CPC Class Codes: A61B 5/4094 Diagnosing or monitoring seizure diseases, e.g. epilepsy
• CPC Class Codes: A61B 5/486 Bio-feedback A61B5/375 takes precedence
• CPC Class Codes: A61B 5/7203 for noise prevention, reduction or removal
• CPC Class Codes: A61B 5/726 using Wavelet transforms
• CPC Class Codes: A61B 5/7264 Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems neural networks per se G06N3/00; expert systems per se G06N5/00
• CPC Class Codes: A61B 5/7267 involving training the classification device
• CPC Class Codes: A61B 5/7275 Determining trends in physiological measurement data; Predicting development of a medical condition based on physiological measurements, e.g. determining a risk factor
• CPC Class Codes: A61N 1/36082 Cognitive or psychiatric applications, e.g. dementia or Alzheimer's disease
• CPC Class Codes: G06F 2218/08 Feature extraction
• CPC Class Codes: G16H 20/70 relating to mental therapies, e.g. psychological therapy or autogenous training
• CPC Class Codes: G16H 40/63 for local operation
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems
• CPC Class Codes: G16H 50/30 for calculating health indices; for individual health risk assessment

Abstract

A method and apparatus for forecasting and controlling neurological abnormalities in humans such as seizures or other brain disturbances. The system is based on a multi-level control strategy. Using as inputs one or more types of physiological measures such as brain electrical, chemical or magnetic activity, heart rate, pupil dilation, eye movement, temperature, chemical concentration of certain substances, a feature set is selected off-line from a pre-programmed feature library contained in a high level controller within a supervisory control architecture. This high level controller stores the feature library within a notebook or external PC. The supervisory control also contains a knowledge base that is continuously updated at discrete steps with the feedback information coming from an implantable device where the selected feature set (feature vector) is implemented. This high level controller also establishes the initial system settings (off-line) and subsequent settings (on-line) or tunings through an outer control loop by an intelligent procedure that incorporates knowledge as it arises. The subsequent adaptive settings for the system are determined in conjunction with a low-level controller that resides within the implantable device. The device has the capabilities of forecasting brain disturbances, controlling the disturbances, or both. Forecasting is achieved by indicating the probability of an oncoming seizure within one or more time frames, which is accomplished through an inner-loop control law and a feedback necessary to prevent or control the neurological event by either electrical, chemical, cognitive, sensory, and/or magnetic stimulation.

Claims

1. A method for predicting and controlling the electrographic and clinical onset of a seizure and other neurological events in an individual, comprising the acts of:

generating data that is acquired from a plurality of input signals obtained from at least one sensor located in or on the individual;

fusing the data to combine information from the at least one sensor that is connected to at least one transducer;

selecting and extracting a plurality of features from the fused data;

determining from the extracted features if a seizure or other neurological event is likely to occur within a plurality of specified time frames, and the probability of having a seizure for each specified time frame;

providing an alarm to the individual to inform him of an imminent seizure or neurological event when the probability of seizure is higher than an adaptive threshold; and

applying a control rule to initiate an intervention measure that is commensurate with the probability of the electrographical onset of a seizure for each specified time frame.

2. The method for predicting and controlling the electrographic onset of a seizure of claim 1 further comprising the act of normalizing the selected features before determining if a seizure is likely to occur within the specified time frame.

3. The method for predicting and controlling the electrographic onset of a seizure of claim 1 further comprising preprocessing of the input signals to reduce noise, to enhance the quality, to compensate for undesireable signal variations and to emphasize distinguishability between a pre-seizure class and a non-pre-seizure class.

4. The method for predicting and controlling the electrographic onset of a seizure of claim 1 wherein the intervention measure is an electrical stimulus of a minimally required duration and intensity that is delivered at a time that is based on the probability of seizure for a specified time frame.

5. The method for predicting and controlling the electrographic onset of a seizure of claim 1 wherein the intervention measure is a drug infusion that is activated to deliver a minimally required amount of a drug into the individual at a time that is based on the probability of seizure for a specified time frame.

6. The method for predicting and controlling the electro graphic onset of a seizure of claim 1 wherein the intervention measure is a magnetic stimulus generated by the wearing of a magnetic helmet at a time that is based on the probability of seizure for a specified time frame.

7. The method for predicting and controlling the electrographic onset of a seizure of claim 1 wherein the intervention measure is a procedure that includes the solving of highly cognitive problems.

8. The method for predicting and controlling the electrographic onset of a seizure of claim 1 wherein the intervention measure is a sensory stimulation including at least one of music therapy, images, flavors, odors and tactile sensations.

9. The method for predicting and controlling the electrographic onset of a seizure of claim 1 wherein the intervention measure is delivered in at least one of a region of onset and a distribution region surrounding the region of offset.

10. The method for predicting and controlling the electrographic onset of a seizure of claim 1 wherein the intervention measure is delivered in subcortical regions including at least one of the thalamus, basal ganglia, and other deep nuclei.

11. The method for predicting and controlling the electro graphic onset of a seizure of claim 1 wherein if the electrograhic onset occurs, applying treatment to either at least one of a general region of onset and deep brain structures to modulate the behavior of the seizure focus.

12. The method for predicting and controlling the electrographic onset of a seizure of claim 1 wherein the intervention measure application includes at least one of:

rhythmic electrical pacing that changes in frequency, intensity and distribution as the probability of a seizure onset reaches and exceeds a threshold;

chaos control pacing;

random electrical stimulation to interfere with developing coherence in activity in a region of, and surrounding, an epileptic focus;

depolarization or hyperpolarization stimuli to silence or suppress activity in actively discharging regions, or regions at risk for seizure spread.

13. The method for predicting and controlling the electro graphic onset of a seizure of claim 12 wherein the intervention measure is delivered to a plurality of electrodes to provide a surround inhibition to prevent a progression of a seizure precursor.

14. The method for predicting and controlling the electrographic onset of a seizure of claim 12 wherein the intervention measure is delivered sequentially in a wave that covers a cortical or subcortical region of tissue so as to progressively inhibit normal or pathological neuronal function in the covered region.

15. The method for predicting and controlling the electrographic onset of a seizure of claim 1 wherein the intervention measure application is an infusion of a therapeutic chemical agent into a brain region where seizures are generated, or to which they may spread.

16. The method for predicting and controlling the electrographic onset of a seizure of claim 15 wherein the chemical agent is delivered in greater quantity, concentration or spatial distribution as the probability of seizure increases.

17. The method for predicting and controlling the electrographic onset of a seizure of claim 15 wherein the intervention measure is applied to at least one of an epilectic focus, an area surrounding the epilectic focus, a region involved in an early spread, and a central or deep brain region to modulate seizure propagation.

18. The method for predicting and controlling the electrographic onset of a seizure of claim 15 wherein the therapeutic chemical agent is activated by oxidative stress and increases in concentration and distribution as the probability of seizure increases.

19. The method for predicting and controlling the electrographic onset of a seizure of claim 1 wherein the intervention measure is delivered to central nerves or blood vessels in a graduated manner as the probability of seizure increases.

20. The method for predicting and controlling the electrographic onset of a seizure of claim 1 wherein the intervention measure is a plurality of artificial neuronal signals delivered to disrupt eletrochemical traffic on at least one neuronal network that includes or communicates with an ictal onset zone.

21. The method for predicting and controlling the electrographic onset of a seizure of claim 1 wherein the alarm is any one of a visual signal, an audio signal and a tactile sensation.

22. The method for predicting and controlling the electrographic onset of a seizure of claim 1 wherein the plurality of features are selected for each individual.

23. The method for predicting and controlling the electrographic onset of a seizure of claim 1 wherein the same plurality of features are selected for each individual.

24. The method for predicting and controlling the electrographic onset of a seizure of claim 1 wherein parameters of the selected features are tuned for each individual.

25. The method for predicting and controlling the electrographic onset of a seizure of claim 24 wherein one of the parameters that is used for each selected feature is a running window length that is used in feature extraction.

26. The method for predicting and controlling the electrographic onset of a seizure of claim 1 wherein a plurality of features are extracted at an analog level.

27. The method for predicting and controlling the electro graphic onset of a seizure of claim 1 wherein a plurality of features are extracted at a digital level.

28. The method for predicting and controlling the electrographic onset of a seizure of claim 1 wherein the plurality of features are extracted over a pre-established window length.

29. The method for predicting and controlling the electrographic onset of a seizure of claim 28 further comprising shifting of the window over the plurality of input signals to allow at least a partial overlap with a previous window, reusing the extracted features in the overlap portion and repeating the extraction of the plurality of features on a new input portion within the window.

30. The method for predicting and controlling the electrographic onset of a seizure of claim 1 wherein the act of fusing the data comprises the act of combining the plurality of signals from at least one sensor using an intelligent tool including a neural network or a fuzzy logic algorithm.

31. The method for predicting and controlling the electrographic onset of a seizure of claim 3 wherein the act of preprocessing of the input signals comprises subtraction of input signals from spatially adjacent sensors that measure the same type of activity.

32. The method for predicting and controlling the electrographic onset of a seizure of claim 1 wherein the plurality of features is selected from a feature library including a plurality of historical and instantaneous features.

33. The method for predicting and controlling the electrographic onset of a seizure of claim 32 wherein the plurality of instantaneous features are generated directly from preprocessed and fused input signals through a running observation window.

34. The method for predicting and controlling the electrographic onset of a seizure of claim 32 wherein the historical features are based on a historical evolution of features over time.

35. The method for predicting and controlling the electro graphic onset of a seizure of claim 32 wherein the historical and instantaneous features are limited to a focus region in the brain of an individual.

36. The method for predicting and controlling the electrographic onset of a seizure of claim 32 wherein the historical and instantaneous features are derived as a spatial feature from a combination of a plurality of regions in the brain of an individual.

37. The method for predicting and controlling the electrographic onset of a seizure of claim 32 wherein the feature library includes a collection of custom routines to compute the features.

38. The method for predicting and controlling the electrographic onset of a seizure of claim 32 wherein the plurality of features are extracted from different domains.

39. The method for predicting and controlling the electrographic onset of a seizure of claim 38 wherein at least one feature is a ratio of a short term value and a long term value of that feature

40. The method for predicting and controlling the electrographic onset of a seizure of claim 38 wherein the different domains include at least two of time, frequency, wavelet, fractal geometry, stochastic processes, statistics, and information theory domains.

41. The method for predicting and controlling the electrographic onset of a seizure of claim 40 wherein the time domain features include at least one of an average power, a power derivative, a fourth-power indicator, an accumulated energy, an average non-linear energy, a thresholded non-linear energy, a duration of thresholded non-linear energy, and a ratio of short term and long term power feature.

42. The method for predicting and controlling the electrographic onset of a seizure of claim 41 wherein the fractal geometry features include at least one of a fractal dimension of analog signal, a curve length, a fractal dimension of digital signals, a ratio of short term and long term curve length, an a ratio of short term and long term fractal dimensions of digital signals.

43. The method for predicting and controlling the electrographic onset of a seizure of claim 41 wherein the frequency domain features include at least one of a power spectrum, a power on frequency bands, a coherence between intracranial channels, a mean crossings and a zero crossings feature.

44. The method for predicting and controlling the electrographic onset of a seizure of claim 41 wherein the wavelet domain features include at least one of a spike detector, a density of spikes over time, and an absolute value of a wavelet coefficient.

45. The method for predicting and controlling the electro graphic onset of a seizure of claim 41 wherein the statistics and stochastic process domains include at least one of a mean frequency index, a cross-correlation between different intracranial channels, and autoregressive coefficients.

46. The method for predicting and controlling the electrographic onset of a seizure of claim 41 wherein the information theory features include at least one of an entropy feature and an average mutual information feature.

47. The method for predicting and controlling the electrographic onset of a seizure of claim 34 wherein at least one historical feature is generated as a feature of other features by a second or higher level of feature extraction.

48. The method for predicting and controlling the electrographic onset of a seizure of claim 25 wherein a determination of the running window length and a starting time for feature extraction over an input signal for every feature includes the acts of:

determining a window range based on stationarity criteria and a minimum length to compute a feature under analysis;

determining a feature value for each of a plurality of different window sizes;

calculating a feature effectiveness measure based on class distinguishability for the plurality of different window sizes used for every feature;

determining the window length that corresponds to a best class distinguishability as indicated by a maximum value or minimum value of the feature effectiveness measure; and

aligning the plurality of windows with the window having the maximum length such that the right edge of all windows coincide.

49. The method for predicting and controlling the electrographic onset of a seizure of claim 48 wherein the maximum or minimum values of the feature effectiveness measure that provides the best class distinguishability depends on the feature effectiveness measure in use.

50. The method for predicting and controlling the electrographic onset of a seizure of claim 48 wherein the feature effectiveness measure determines the window length that maximizes the distinguishability between a preictal/ictal class and a baseline class.

51. The method for predicting and controlling the electrographic onset of a seizure of claim 50 wherein the act of selecting and extracting a plurality of features comprises the acts of:

extracting a set of candidate features from the feature library;

ranking the extracted features by the feature effectiveness measure; and

determining a smallest subset of features that satisfies a performance criterion.

52. The method for predicting and controlling the electrographic onset of a seizure of claim 51 further comprising the acts of:

performing an initial pre-selection from the feature library to discard a plurality of features with inferior class separability; and

evaluating individual feature performance using at least one criterion for every feature that is not discarded during the initial pre-selection.

53. The method for predicting and controlling the electrographic onset of a seizure of claim 51 wherein the act or ranking the extracted features by the feature effectiveness measure uses an overlap measure criterion, a modified add-on algorithm and heuristics to select a final feature set.

54. The method for predicting and controlling the electrographic onset of a seizure of claim 51 further comprising the acts of constructing and evaluating two-dimensional feature spaces to validate qualitatively that the final feature set is complementary and has low correlation among the final features.

55. The method for predicting and controlling the electrographic onset of a seizure of claim 53 wherein the overlap measure criterion is based on functions proportional to the estimated conditional probability distributions of the features under analysis for both a pre-seizure class and a non-pre-seizure class.

56. The method for predicting and controlling the electrographic onset of a seizure of claim 30 wherein the neural network or fuzzy logic algorithm include at least one of a probabilistic neural network, a k-nearest neighbor neural network, a wavelet network, and a combination probabilistic/k-nearest neighbor neural network.

57. The method for predicting and controlling the electrographic onset of a seizure of claim 3 wherein the act of preprocessing the input signals comprises classification of an individual'"'"'s awareness state within at least one of the categories of awake, asleep, and drowsy using algorithms based on frequency and time information.

58. The method for predicting and controlling the electrographic onset of a seizure of claim 1 further comprising the act of fusing the selected features to include establishing an individual-tuned variable normalization level that uses an individual'"'"'s state of awareness to normalize an accumulated energy or other feature and decide if a seizure is approaching when a normalized threshold value is exceeded.

60. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 59 further comprising program instructions that initiate a preproccessing of the input signals to reduce noise and to enhance the quality, and to emphasize distinguisability between a pre-seizure class and a non-pre-seizure class.

61. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 59 further comprising program instructions that initiate an electrical stimulus of a minimally required duration and intensity that is delivered at a time that is based on the probability of seizure for a specified time frame.

62. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 59 further comprising program instructions that initiate activation of a drug infusion to deliver a minimally required amount of a drug into the individual at a time that is based on the probability of a seizure for a specified time frame.

63. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 59 further comprising program instructions that initiate generation of a magnetic stimulus through the wearing of a magnetic helmet at a time that is based on the probability of seizure for a specified time frame.

64. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 59 further comprising program instructions that provide an indication that a cognitive problem should be solved as an intervention measure.

65. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 59 further comprising program instructions that provide an indication that a sensory stimulation should be applied as an intervention measure.

66. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 59 further comprising program instructions that initiate activation of any one of a visual alarm, an audio alarm, and a tactile sensation.

67. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 59 further comprising program instructions that select a plurality of features for each individual.

68. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 59 further comprising program instructions that select the same plurality of features for each individual.

69. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 59 further comprising program instructions that tune the parameters of the selected features for each individual.

70. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 59 further comprising program instructions that determine a running window length which is used in feature extraction for each selected feature.

71. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 59 further comprising program instructions that extract a plurality of features at an analog level.

72. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 59 further comprising program instructions that extract a plurality of features at a digital level.

73. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 70 further comprising program instructions that extract a plurality of features over a pre-established window length.

74. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 73 further comprising program instructions that shift the window over the plurality of input signals to allow at least a partial overlap with a previous window and repeat the extraction of the plurality of features.

75. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 59 further comprising program instructions that combine the plurality of signals from at least one sensor using an intelligent tool that includes a neural network or a fuzzy logic algorithm.

76. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 60 wherein the program instruction for preprocessing of the input signals further comprises program instructions that subtract the signals from spatially adjacent sensors that measure the same type of activity.

77. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 59 further comprising program instructions that select a plurality of features from a feature library that includes a plurality of historical and instantaneous features.

78. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 77 further comprising program instructions that generate a plurality of instantaneous features directly from pre-processed and fused input signals through a running observation window.

79. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 77 further comprising program instructions that generate historical features based on a historical evolution of features over time.

80. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 77 further comprising program instructions that limit the historical and instantaneous features to a focus region in the brain of an individual.

81. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 77 further comprising program instructions that derive historical and instantaneous features as a spatial feature from a combination of a plurality of regions in the brain of an individual.

82. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 77 further comprising program instructions collected as custom routines within the feature library to compute the features.

83. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 77 further comprising program instructions that extract a plurality of features from different domains.

84. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 83 further comprising program instructions that determine at least one feature as a ratio of a short term value and a long term value of that feature.

85. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 83 wherein the different domains include at least two of time, frequency, wavelet, fractal geometry, stochastic processes, statistics, and information theory domains.

86. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 85 further comprising program instructions that determine at least one of an average power, a power derivative, a fourth-power indicator, an accumulated energy, and average non-linear energy, a thresholded non-linear energy, a duration of thresholded non-linear energy, and a ratio of short term and long term power as time domain features.

87. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 86 further comprising program instructions that determine at least one of a fractal dimension of analog signals, a curve length, a fractal dimension of digital signals, a ratio of a short term and a long term fractal dimension of digital signals, and a ratio of short term and long term curve length as fractal geometry features.

88. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 86 further comprising program instructions that determine at least one of a power spectrum, a power on frequency bands, a coherence between intracranial channels, a mean crossings and a zero crossings feature.

89. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 86 further comprising program instructions that determine at least one of a spike detector, a density of spikes over time, and an absolute value of a wavelet coefficient as wavelet domain features.

90. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 86 further comprising program instructions that determine at least one of a mean frequency index, a cross-correlation between different intracranial channels, and autoregressive coefficients as features in the statistics and stochastic process domains.

91. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 86 further comprising program instructions that determine at least one of an entropy feature and an average mutual information feature as information theory features.

92. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 70 wherein the program instructions for determining the running window length further comprise:

program instructions that determine a window range based on stationarity criteria and a minimum length to compute a feature under analysis;

program instructions that determine a feature value for each of a plurality of different window sizes;

program instructions that calculate a feature effectiveness measure for each feature for the plurality of different window sizes;

program instructions that determine the optimal window length for each feature from the plurality of windows examined that corresponds to a value of the feature effectiveness measure wherein the distinguisability between a preictal class and a non-preictal class is maximized; and

program instructions that align the plurality of optimal windows determined for each feature with the feature window having the maximum length.

93. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 92 further comprising program instructions that initiate re-execution of the program instructions that determine a feature value and the program instructions that calculate a feature effectiveness measure for each selected feature.

94. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 92 further comprising program instructions that maximize the distinguishability between a preictal/ictal class and a baseline class as the feature effectiveness measure.

95. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 94 wherein the program instructions that select and extract a plurality of features comprise:

program instructions that extract a set of candidate features from the feature library;

program instructions that rank the extracted features by the feature effectiveness measure; and

program instructions that determine a smallest subset of features that satisfies a performance criterion.

96. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 95 further comprising:

program instructions that perform an initial pre-selection from the feature library to discard a plurality of features with inferior class separability; and

program instructions that evaluate individual feature performance using at least one criterion for every feature that is not discarded during the initial pre-selection.

97. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 95 wherein the program instructions that rank the extracted features by the feature effectiveness measure use an overlap measure criterion, a modified add-on algorithm and heuristics to select a final feature set.

98. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 97 further comprising program instructions that construct and evaluate two-dimensional feature spaces to validate qualitatively that the final feature set is complementary and has low correlation among the final features.

99. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 97 further comprising program instructions that base the overlap measure criterion on estimated conditional probability distributions of each particular feature under analysis for both a pre-seizure class and non-pre-seizure class.

100. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 75 further comprising program instructions that determine at least one of a probabilistic neural network, a k-nearest neighbor neural network, a wavelet network, and a combination probabilistic/k-nearest neighbor neural network.

101. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 60 wherein the program instructions for preprocessing of the input signals further comprises program instructions that classify an individual'"'"'s awareness state within at least one of the categories of awake, asleep, and drowsy.

102. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 101 wherein the program instructions that classify an individual'"'"'s awareness state within the categories of awake, asleep and drowsy are based on frequency and time information.

103. The computer program product for predicting and controlling the electrographic onset of a seizure of claim 59 further comprising program instructions that fuse the selected features to include establishing an individual-tuned variable normalization level that uses an individual'"'"'s state of awareness to normalize an accumulated energy or other feature and decide if a seizure is approaching when a normalized threshold value is exceeded.

173. The method for predicting and controlling the electrographic onset of a seizure of claim 1 further comprising the act of implanting a plurality of electrodes in each focus region of the individual.

180. The method for predicting and controlling the electrographic onset of a seizure of claim 1 further comprising the act of implanting a plurality of electrodes in a unique focus region and in at least one other region of the brain of the individual.

59. A computer readable medium containing a computer program product for predicting and controlling the electrographic and clinical onset of a seizure and other neurological events in an individual, the computer program product comprising:

program instructions that generate data acquired from a plurality of input signals obtained from at least one sensor located in or on the individual;

program instructions that fuse the data to combine information from the at least one sensor that is connected to at least one transducer;

program instructions that select and extract a plurality of features from the fused data;

program instructions that determine from the extracted features if a seizure or other neurological event is likely to occur within a plurality of specified time frames, and the probability of having a seizure for each specified time frame;

program instructions that generate an alarm to the individual to inform him of an imminent seizure or neurological event when the probability of seizure is higher than an adaptive threshold; and

program instructions that apply a control rule to initiate an intervention measure that is commensurate with the probability of the electrographical onset of a seizure.

104. A system for predicting and controlling the electrographic and clinical onset of a seizure and other neurological disturbances in an individual, comprising:

a data generation component to acquire physiological signals from the individual;

an intelligent data processing unit to preprocess the physiological signals, to extract and select a plurality of features, and to provide an estimation of the probability of seizure for at least one time frame; and

a low level controller connected to the intelligent data processing unit to automatically activate a therapeutic intervention measure to control the onset of a seizure in the individual in response to the probability of seizure exceeding a threshold.

105. The system for predicting and controlling the electrographic onset of a seizure of claim 104 further comprising at least one sensor for detecting physiological signals that indicate the state of activity in the brain of the individual.

106. The system for predicting and controlling the electrographic onset of a seizure of claim 105 wherein the sensor is at least one of an implanted intracranial electrode, an epidural electrode, a scalp electrode, a sphenoidal electrode, a foramen ovale electrode, an intravascular electrode, a chemical sensor, a pupil dilation sensing device, an eye movement sensor, a heart rate sensor, and a body temperature sensor.

107. The system for predicting and controlling the electrographic onset of a seizure of claim 104 further comprising a high level controller that communicates with the intelligent data processing unit to retune at least one parameter that is used to extract and select a feature.

108. The system for predicting and controlling the electro graphic onset of a seizure of claim 104 further comprising an external portable module including an external communications unit that enables the transfer of physiological data that is sensed in the individual to the external portable module for analysis and storage.

109. The system for predicting and controlling the electrographic onset of a seizure of claim 108 where in the external portable module further comprises a display device that shows the probability output from the intelligent data processing unit for having a seizure in at least one time frame.

110. The system for predicting and controlling the electrographic onset of a seizure of claim 108 wherein the external portable module further comprises an alarm device which is activated to alert the individual of an oncoming seizure when the probability of having a seizure in at least one time frame exceeds an adaptive threshold.

111. The system for predicting and controlling the electrographic onset of a seizure of claim 108 wherein the external portable module further comprises a battery recharger.

112. The system for predicting and controlling the electrographic onset of a seizure of claim 108 wherein the external portable module further comprises at least one of a microprocessor, a digital signal processor, a field programmable gate array, and an application specific integrated circuit.

113. The system for predicting and controlling the electro graphic onset of a seizure of claim 108 wherein the external communications unit communicates with the intelligent data processing unit by any one of telemetry, magnetic induction, direct electrical connection, optical communication and ultrasonic communication.

114. The system for predicting and controlling the electrographic onset of a seizure of claim 108 wherein the external portable module further comprises a communications port that enables the external portable module to be connected to a serial or a parallel port of a computer system, and that enables the transmission of stored data from the external portable module through an Internet connection to another computer system where the transmitted data can be downloaded and stored.

115. The system for predicting and controlling the electrographic onset of a seizure of claim 104 wherein the intelligent data processing unit is contained in an implantable device.

116. The system for predicting and controlling the electrographic onset of a seizure of claim 115 wherein the implantable device is implanted in the brain of the individual.

117. The system for predicting and controlling the electrographic onset of a seizure of claim 104 wherein the intelligent data processing unit is programmed into any one of a microprocessor, a digital signal processor, a field programmable gate array, and an application specific integrated circuit (ASIC) embedded on a microchip.

118. The system for predicting and controlling the electrographic onset of a seizure of claim 104 wherein the intelligent data processing unit comprises a preprocessor to amplify and filter the physiological signals.

119. The system for predicting and controlling the electrographic onset of a seizure of claim 104 wherein the intelligent data processing unit comprises a first feature extraction module to extract analog features from the preprocessed physiological signals.

120. The system for predicting and controlling the electrographic onset of a seizure of claim 119 wherein the intelligent data processing unit further comprises a second feature extraction module to extract digital features from the preprocessed physiological signals.

121. The system for predicting and controlling the electrographic onset of a seizure of claim 120 wherein the intelligent data processing unit further comprises a feature vector generator module that combines a plurality of extracted features based on a running window technique.

122. The system for predicting and controlling the electrographic onset of a seizure of claim 104 wherein the intelligent data processing unit comprises an on-board memory to record the physiological signals over a period of time based on a capacity of the memory.

123. The system for predicting and controlling the electrographic onset of a seizure of claim 121 wherein the intelligent data processing unit further comprises an intelligent prediction analysis and classification module operating on a central processor that analyzes the feature vector to provide an estimation of the probability of having a seizure for one or more time frames.

124. The system for predicting and controlling the electrographic onset of a seizure of claim 104 further comprising a neural network to perform the analysis of the feature vector.

125. The system for predicting and controlling the electrographic onset of a seizure of claim 124 wherein the neural network is at least one of a probabilistic neural network, a k-nearest neighbor neural network, and a wavelet neural network.

126. The system for predicting and controlling the electro graphic onset of a seizure of claim 123 further comprising an internal communications unit to enable the transfer of physiological data that is sensed in the individual by a central processor in the intelligent data processing unit to an external portable module that displays the probability of seizure for at least one time frame.

127. The system for predicting and controlling the electrographic onset of a seizure of claim 104 further comprising an internal electrical stimulation unit activated by the low level controller to electrically stimulate focal points to prevent synchronized nerve im pulses as the therapeutic intervention measure.

128. The system for predicting and controlling the electrographic onset of a seizure of claim 104 further comprising a drug delivery system activated by the low level controller to provide chemical stimulation as by releasing small quantities of a drug as the therapeutic intervention measure.

129. The system for predicting and controlling the electrographic onset of a seizure of claim 104 further comprising a special helmet activated by the low level controller to provide magnetic stimulation as the therapeutic intervention measure.

130. The system for predicting and controlling the electrographic onset of a seizure of claim 104 wherein the low level controller activates a stimulation unit to instruct the individual to initiate a sensory/perceptive stimulus as the therapeutic intervention measure.

131. The system for predicting and controlling the electro graphic onset of a seizure of claim 104 wherein the low level controller activates a stimulation unit to instruct the individual to initiate a cognitive stimulus as the therapeutic intervention measure.

132. The system for predicting and controlling the electrographic onset of a seizure of claim 130 wherein the sensory/perceptive stimulus is any of a visual, an auditory, a tactile, a smell and a taste stimulus.

133. The system for predicting and controlling the electrographic onset of a seizure of claim 131 wherein the cognitive stimulus is any of a reading, a mathematical computation, and a logic reasoning problem stimulus.

135. The adaptive multi-level hierarchical control system of claim 134 wherein the knowledge base comprises a priori information for the individual.

136. The adaptive multi-level hierarchical control system of claim 135 wherein the a priori information for the individual comprises seizure frequency over time, seizure duration, type of seizure, and aura frequency collected before an implantation of the intelligent data processing device.

137. The adaptive multi-level hierarchical control system of claim 134 wherein the low level controller determines and adjusts the parameter settings of the actuators in the stimulation device continuously.

138. The adaptive multi-level hierarchical control system of claim 134 wherein the high level supervisory controller can operate in an automatic mode or in a semi-automatic mode.

139. The adaptive multi-level hierarchical control system of claim 138 further comprising a master program that monitors a set of controlled variables and updates the applied feedback control laws when operating in the automatic mode.

140. The adaptive multi-level hierarchical control system of claim 138 wherein a physician or specialist inputs parameters directly into the intelligent data processing device through a master program user interface when operating in a semi-automatic mode.

141. The adaptive multi-level hierarchical control system of claim 134 wherein the high level supervisory controller is a computer external to the intelligent data processing device for providing a coordination layer control.

142. The adaptive multi-level hierarchical control system of claim 141 wherein the coordination layer of control returns system parameters including parameters related to fusion of sensory data, feature extraction, feature normalization, neural network retraining, fuzzy logic adjustments, and fault diagnoses of actuators, sensors and implantable device.

143. The adaptive multi-level hierarchical control system of claim 134 further comprising an external computer for providing a research layer control to evaluate any new algorithms for control of seizures or brain disturbances, for prediction and detection of the unequivocal electrographic onset of seizure, for control strategies, or for other types of parameter adjustments.

144. The adaptive multi-level hierarchical control system of claim 143 wherein the research layer computer analyzes physiological mechanisms to explain seizure and other brain disturbances.

145. The adaptive multi-level hierarchical control system of claim 143 wherein the research layer collects information from a plurality of individuals to form a research and development database.

146. The adaptive multi-level hierarchical control system of claim 134 wherein the multi-level hierarchical control is provided by a feedback control law updated by the low level controller and a knowledge base control law updated by the high level supervisory controller.

147. The adaptive multi-level hierarchical control system of claim 146 wherein the adaptive hierarchical control is provided by the updated knowledge base control law.

148. The adaptive multi-level hierarchical control system of claim 134 wherein the processor for the high level supervisory controller operates a logic module that executes optimization algorithms and determines self-evaluation metrics to establish the supervisory controller'"'"'s performance over time, to determine required adjustments in the intelligent data processing device'"'"'s set points, and to generate an updated feedback control law that is downloaded into the intelligent data processing device.

149. The adaptive multi-level hierarchical control system of claim 134 wherein the knowledge database is updated at discrete steps by downloading new information from the intelligent data processing device.

150. The adaptive multi-level hierarchical control system of claim 134 further comprising an external portable module including an external communications unit that enables the transfer of physiological data that is sensed in the individual to the external portable module for analysis and storage.

151. The adaptive multi-level hierarchical control system of claim 150 wherein the external portable module further comprises a display device that shows the probability from the intelligent data processing unit for having a seizure in at least one time frame.

152. The adaptive multi-level hierarchical control system of claim 150 wherein the external portable module further comprises an alarm device which is activated to alert the individual of an oncoming seizure when the probability of having a seizure in at least one time frame exceeds an adaptive threshold.

153. The adaptive multi-level hierarchical control system of claim 134 wherein the intelligent data processing device is implanted into the individual.

154. The adaptive multi-level hierarchical control system of claim 153 wherein the intelligent data processing device includes a learning capability based on artificial intelligence tools and an analysis of previously stored information that enables an adaptation of the intelligent data processing device to the individual in which it is implanted and a specific state of the individual at any time.

155. The adaptive multi-level hierarchical control system of claim 134 further comprising at least one sensor for detecting physiological signals that indicate the state of activity in the brain of an individual.

157. The method for predicting and controlling the onset of a seizure of claim 156 wherein the implanted device is feature/parameter-tuned with features that are selected for each patient based on the features that can capture the unequivocal electrographic onset of seizure in advance.

158. The method for predicting and controlling the onset of a seizure of claim 156 wherein the implanted device is parameter-tuned with the same features used for each individual receiving an implanted device in which the parameters are tuned on an individual basis.

159. The method for predicting and controlling the onset of a seizure of claim 156 wherein the act of installing the at least one sensor includes determining the focus region for correct installation.

160. The method for predicting and controlling the onset of a seizure of claim 156 wherein the act of initializing the parameter settings includes the acts of:

recording sensor data into the intermediate storage device continuously from a pair of input channels;

downloading the recorded sensor data from the intermediate storage device into an external processing device;

preprocessing and fusing the downloaded sensor data by the external processing device;

extracting and selecting features in the external processing device;

selecting a best feature set by the external processing device to establish a feature vector; and

transferring and setting the selected feature algorithms from the external processing device into the implantable device.

161. The method for predicting and controlling the onset of a seizure of claim 156 wherein the acts of analyzing and classifying the selected features includes the acts of:

performing real-time processing of the input signals from the at least one sensor by subtracting a focal channel input signal from an adjacent channel, and filtering the difference signal;

extracting each selected feature at an analog level or a digital level based on the characteristics of the selected feature;

combining the extracted features using a running-window technique to generate a feature vector;

normalizing the feature vector by a processor in the implanted device;

performing analysis of the feature vector for each time frame using a fuzzy system or a neural network to provide an estimation of the probability of having a seizure for at least one time frame.

162. The method for predicting and controlling the onset of a seizure of claim 161 further comprising the acts of:

displaying a probability output of having a seizure for at least one time frame on the display of the external portable module; and

activating an alarm to alert the individual of an oncoming seizure when the probability output exceeds an adaptive threshold.

163. The method for predicting and controlling the onset of a seizure of claim 161 further comprising the acts of:

scheduling the download of recorded sensor data from a buffer in the implanted device into the intermediate storage device by a processor in the external portable module;

transferring data between the external processing device and the external portable module to establish supervisory control actions and to communicate the control actions to the implanted device.

164. The method for predicting and controlling the onset of a seizure of claim 163 further comprising the act of establishing a communications link between a central processor in the implanted device and the processor in the external portable module.

165. The method for predicting and controlling the onset of a seizure of claim 161 further comprising the act of recording physiological input signals in an internal buffer of the implanted device for a period of time that depends on the memory capability of the buffer.

166. The method for predicting and controlling the onset of a seizure of claim 165 further comprising the act of downloading physiological input signals, the feature vector and a plurality of controlled variables from the internal buffer to the intermediate storage device via a communications link.

167. The method for predicting and controlling the onset of a seizure of claim 166 further comprising the act of downloading data from the intermediate storage device to the external processing device.

168. The method for predicting and controlling the onset of a seizure of claim 161 further comprising the act of performing an initial adaptation of the implanted device at periodically discrete times by connecting the external portable module to a high level supervisory control in the external processing device.

169. The method for predicting and controlling the onset of a seizure of claim 156 wherein the act of applying a multi-level control includes the acts of:

activating the closed-loop control system via a high level supervisory control through the external portable module;

generating feedback control signals by the low level controller to prevent seizures by producing an intermittent electrical, chemical or a magnetic stimulation;

estimating prediction and prevention performance by evaluating a plurality of key parameters;

computing an overall performance metric from the prediction and prevention performance;

adjusting the parameters of a stimulation device and determining a type of stimulation to apply and a corresponding start time, intensity, duration and frequency;

updating feedback control and knowledge base laws;

adapting the feedback control laws to internal and external changes over time to prevent seizure with less-invasive intervention measures; and

tuning internal feature parameters and analysis and classification parameters adaptively based on the combined information contained in the feedback control signals and the overall performance measures.

170. The method for predicting and controlling the onset of a seizure of claim 169 further comprising the acts of:

activating an input channel by the individual via the keypad in the external portable module;

automatically adjusting the hierarchical control system in response to the activation of an input channel;

assessing hierarchical control system performance by using information regarding the probability of seizure in conjunction with preictal and ictal recorded data.

171. The method for predicting and controlling the onset of a seizure of claim 170 wherein the hierarchical control system performance evaluation is performed automatically at a regulatory feedback control level and at a high level supervisory controller.

172. The method for predicting and controlling the onset of a seizure of claim 170 wherein the hierarchical control system performance is activated by an authorized person.

174. The method for predicting and controlling the electrographic onset of a seizure of claim 173 wherein the act of fusing the data comprises subtracting the input signals from adjacent electrodes to form a bipolar signal, and combining the bipolar signals from different focus regions at the data level.

175. The method for predicting and controlling the electrographic onset of a seizure of claim 173 wherein the act of fusing the data comprises subtracting the input signals from adjacent electrodes to form a bipolar signal, and combining the bipolar signals from different focus regions at the feature level.

176. The method for predicting and controlling the electrographic onset of a seizure of claim 174 wherein the input signals are combined into a signal data stream either before or after a preprocessing stage.

177. The method for predicting and controlling the electrographic onset of a seizure of claim 176 wherein the input signals are intracranial electroencephalogram data.

178. The method for predicting and controlling the electrographic onset of a seizure of claim 175 wherein the features derived from the input signals and coincident or aligned in time are combined into a single feature using a nonlinear procedure.

179. The method for predicting and controlling the electrographic onset of a seizure of claim 178 wherein the nonlinear procedure comprises selecting the maximum value of the input signals at each sample time.

181. The method for predicting and controlling the electrographic onset of a seizure of claim 180 wherein the at least one other region is a focal adjacent channel.

182. The method for predicting and controlling the electrographic onset of a seizure of claim 180 wherein the act of fusing the data comprises subtracting the input signals from a pair of electrodes placed in different regions to form a bipolar signal, and combining a plurality of bipolar signals at the data level.

183. The method for predicting and controlling the electrographic onset of a seizure of claim 180 wherein the act of fusing the data comprises subtracting the input signals from a pair of electrodes placed in different regions to form a bipolar signal, and combining a plurality of bipolar signals at the feature level.

184. The method for predicting and controlling the electrographic onset of a seizure of claim 182 wherein the input signals are combined into a signal data stream either before or after a preprocessing stage.

185. The method for predicting and controlling the electrographic onset of a seizure of claim 184 wherein the input signals are intracranial electroencephalogram data.

186. The method for predicting and controlling the electrographic onset of a seizure of claim 183 wherein the features derived from the input signals and coincident or aligned in time are combined into a single feature using a nonlinear procedure.

187. The method for predicting and controlling the electrographic onset of a seizure of claim 186 wherein the nonlinear procedure comprises selecting the maximum value of the input signals at each sample time.

134. An adaptive multi-level hierarchical control system for predicting and controlling the electrographic onset of a seizure and other neurological disturbances in an individual, comprising:

a data generation component that acquires physiological signals from the individual;

an intelligent data processing device that processes the physiological signals to extract features which are analyzed and classified and selected to form a feedback vector;

a low level controller including a stimulation device that is activated to apply an intervention measure in response to the feedback vector to control the onset of seizure and to adjust internal parameter settings of the actuators in the stimulation device; and

a high level supervisory controller including a knowledge database and a processor that adapts to feedback vector changes over time and re-tunes the intelligent data processing device parameter settings dynamically.

156. A method for predicting and controlling the electrographic onset of a seizure in an individual using a multi-level hierarchical control system including an implanted device, comprising the acts of:

installing at least one sensor on or in the individual to detect input signals indicative of brain activity;

implanting the device into the brain of the individual;

initializing and tuning a plurality of parameters in the implanted device;

installation of an external portable module that contains an external communications unit, a settings adjustment unit with a display and a keypad and an intermediate storage device;

selecting features to extract from the input signals;

analyzing and classifying the selected features extracted from the input signals in order to predict the probability of having a seizure in a plurality of time frames;

activating a closed-loop control system in the implanted device through the external portable module; and

applying a multi-level control to the implanted device to initiate an intervention measure that is based on the probability of having a seizure in a plurality of time frames.