Use of brain electrophysiological quantitative data to classify and subtype an individual into diagnostic categories by discriminant and cluster analysis

Claim 2

A quantitative process of using brain electro-physiological data (BE) to classify an individual in one of more than two diagnostic categories comprising the machine-implemented steps of: providing BE data for a selected individual; processing said BE data by applying thereto selected spectral analysis and statistical procedures to extract a set of desired features of said BE data; deriving a respective discriminant score for each of a set of more than two diagnostic categories by selectively weighting and combining a number of selected ones of said features of said BE data; combining selected discriminant scores to derive respective probabilities that the individual belongs to selected diagnostic categories; applying selected guardbands or rule-out levels to said probabilities to enhance the reliability of classifying said individual into a diagnostic category on the basis of said probabilities; and classifying the individual into one of more than two diagnostic categories on the basis of said probabilities and said guardbands.

Claim 3

A quantitative process as in claim 1 in which said processing step comprises extracting said features in the form of age-corrected Z-scores which are in the same dimensional units.

Claim 4

A quantitative process as in claim 2 in which said processing step comprises ensuring a Gaussian distribution of the BE data from which said features are extracted.

Claim 5

A quantitative process as in claim 3 in which said processing step comprises extracting features which include absolute and relative spectral power of said BE data in selected frequency bands.

Claim 6

A quantitative BE process as in claim 4 in which said processing step comprises extracting features which include symmetry and coherence measures relating respective brain regions.

Claim 7

A quantitative process as in claim 5 in which said processing step comprises extracting said features for selected monopolar and bipolar derivations of said BE data.

Claim 8

A quantitative BE process as in claim 6 in which said deriving step comprises deriving said discriminant score for a selected diagnostic category by summing the products of said selected features with selected coefficients which typically differ as between both features and diagnostic categories.

Claim 9

A quantitative BE process as in claim 7 in which said combining step comprises dividing a discriminant score for a selected diagnostic category by a selected combination of the same discriminant score and the discriminant score for each remaining diagnostic category which is a part of the same set of at least two diagnostic categories to thereby derive a measure of the probability that the individual belongs to the selected diagnostic category rather than to another category in the same set of at least two categories.

Claim 10

A quantitative process as in claim 8 in which said combining step comprises deriving a discriminant score DS for a diagnostic category (i) in accordance with the expression DS=(1/a)(A 1 F 1 +A 2 F 2 + . . . +A n F n ), where (1/a) is a function coefficient and A 1 -A n are coefficients, F 1 -F n are selected features and n is an integer index.

Claim 11

A quantitative process as in claim 9 in which said step of applying selected guardbands comprises eliminating as unreliable those probabilities which permit more than a selected chance that an individual who belongs in a first category in said set of categories will be classified in another category in said set and said rule-out levels comprise threshold for eliminating as unreliable those probabilities which are below the threshold.

Claim 12

A quantitative process as in claim 10 in which said step of classifying the individual comprises displaying the classification.

Claim 13

A quantitative process as in claim 9 in which the step of deriving disciminant scores comprises deriving said discriminant scores in accordance with the features and coefficients of the following table for the diagnostic conditions stated therein: ______________________________________ NORMAL vs PRIMARY DEPRESSION ______________________________________ VARIABLE NORMAL PRIMARY ______________________________________ F.sub.1 Left Fronto/Temporal -0.03746 0.75010 A.sub.1 Bipolar Absolute Power F.sub.2 Parietal Delta Bipolar -0.02351 -0.50883 A.sub.2 Coherence F.sub.3 Anterior Bipolar Coherence -0.05034 1.05778 A.sub.3 F.sub.4 Central Bipolar Asymmetry -0.02733 0.55487 A.sub.4 F.sub.5 Anterior Bipolar Asymmetry -0 02111 0.81269 A.sub.5 F.sub.6 Anterior Beta -0.10229 0.39695 A.sub.6 Monopolar Relative Power Function Constant -0.70165 -2.89319 FC ______________________________________ The guardbands for this case are: GUARDBANDS a a' a" ______________________________________ Normal 65 75 85 Primary 60 72 90 ______________________________________

Claim 14

A quantitative process as in claim 9 in which the step of deriving disciminant scores comprises deriving said discriminant scores in accordance with the features and coefficients of the following table for the diagnostic conditions stated therein: ______________________________________ NORMAL vs ABNORMAL for AGE GREATER THAN 50 YEARS ______________________________________ VARIABLE NORMAL ABNORMAL ______________________________________ All Regions Delta 0.00000 0.00000 Bipolar Relative Power All Regions Theta -0.02961 0.36594 Bipolar Relative Power All Regions Alpha 0.05106 -0.25480 Bipolar Relative Power All Regions Beta 0.03480 0.31069 Bipolar Relative Power Anterior All Frequencies -0.02934 0.59480 Bipolar Coherence Posterior All Frequencies -0.00501 0.32559 Bipolar Coherence Anterior All Frequencies -0.01403 0.44775 Bipolar Asymmetry Posterior All Frequencies -0.02759 0.36308 Bipolar Asymmetry Anterior Combined Frequencies -0.05146 0.61554 Monopolar Absolute Power Posterior Combined Frequencies 0.00000 0.00000 Monopolar Absolute Power Function Constant -0.69861 -2.23015 ______________________________________ GUARDBANDS a a' a" ______________________________________ Normal 65 75 80 Abnormal 80 85 90 ______________________________________

Claim 15

A quantitative process as in claim 9 in which the step of deriving disciminant scores comprises deriving said discriminant scores in accordance with the features and coefficients of the following table for the diagnostic conditions stated therein: ______________________________________ NORMAL vs ABNORMAL for AGE LESS THAN OR EQUAL TO 50 YEARS ______________________________________ VARIABLE NORMAL ABNORMAL ______________________________________ All Regions Delta 0.00000 0.00000 Bipolar Relative Power All Regions Theta 0.00000 0.00000 Bipolar Relative Power All Regions Alpha 0.04777 -0.48710 Bipolar Relative Power All Regions Beta 0.02004 0.46486 Bipolar Relative Power Anterior All Frequencies -0.02881 0.75448 Bipolar Coherence Posterior All Frequencies -0.00612 0.30889 Bipolar Coherence Anterior All Frequencies -0.01651 0.48136 Bipolar Asymmetry Posterior All Frequencies -0.02686 0.38016 Bipolar Asymmetry Anterior Combined Frequencies -0.05588 0.53385 Monopolar Absolute Power Posterior Combined Frequencies 0.00000 0.00000 Monopolar Absolute Power Function Constant -0.69794 -2.08874 ______________________________________ GUARDBANDS a a' a" ______________________________________ Normal 65 75 85 Abnormal 65 75 80 ______________________________________

Claim 16

A quantitative process as in claim 9 in which the step of deriving disciminant scores comprises deriving said discriminant scores in accordance with the features and coefficients of the following table for the diagnostic conditions stated therein: __________________________________________________________________________ NORMAL vs PRIMARY DEPRESSION vs DEMENTIA __________________________________________________________________________ VARIABLE NORMAL PRIMARY DEMENTIA __________________________________________________________________________ All Regions Theta -0.02079 0.17234 1.30229 Bipolar Relative Power Left Fronto/Temporal -0.04421 1.02014 0.04645 Bipolar Absolute Power Parietal Delta -0.00752 -0.74484 -0.27939 Bipolar Coherence Fronto/Temporal Delta -0.00899 -0.53506 0.52687 Bipolar Coherence Anterior Bipolar Coherence -0.04753 0.77702 0.94400 Central Bipolar Asymmetry -0.03740 0.57220 0.62493 Anterior Bipolar Asymmetry -0.01438 0.64457 0.43654 Cz Theta -0.00180 -0.79164 1.09048 Monopolar Relative Power P3 Alpha -0.24440 2.82880 1.73974 Monopolar Relative Power O1 Alpha 0.27980 -4.06245 -1.92960 Monopolar Relative Power Anterior Beta -0.08958 0.35577 0.59821 Monopolar Relative Power Function Constant -1.11186 -4.39809 -4.63901 __________________________________________________________________________ GUARDBANDS a a' a" __________________________________________________________________________ Normal 42 80 90 Primary 40 70 75 Dementia 45 75 95 __________________________________________________________________________

Claim 17

A quantitative process as in claim 9 in which the step of deriving disciminant scores comprises deriving said discriminant scores in accordance with the features and coefficients of the following table for the diagnostic conditions stated therein: ______________________________________ UNIPOLAR vs BIPOLAR DEPRESSION ______________________________________ VARIABLE UNIPOLAR BIPOLAR ______________________________________ Right Temporal Beta -0.91564 1.26876 Bipolar Relative Power Left Lateral Alpha -0.15510 1.53144 Monopolar Relative Power O1 Delta Monopolar Asymmetry -0.13401 0.13963 Right Temporal Alpha -0.72519 1.57794 Bipolar Relative Power Left Temporal Alpha -0.06270 -1.59622 Bipolar Relative Power Beta C3 Monopolar Relative Power 0.04106 -1.97474 Beta Cz Monopolar Relative Power 0.26946 1.57156 Fronto-temporal Bipolar 0.01198 -0.75104 Absolute Power Asymmetry Function Constant -0.89218 -2.09228 ______________________________________ GUARDBANDS a a' a" ______________________________________ Unipolar 50 77 92 Bipolar 55 82 87 ______________________________________

Claim 18

A quantitative process as in claim 9 in which the step of deriving disciminant scores comprises deriving said discriminant scores in accordance with the features and coefficients of the following table for the diagnostic conditions stated therein: ______________________________________ PRIMARY DEPRESSION vs ALCOHOL ABUSE ______________________________________ VARIABLE PRIMARY ALCOHOLC ______________________________________ Left Central Beta -0.07380 0.58635 Bipolar Relative Power Left Fronto/Temporal Beta -0.08163 0.58543 Bipolar Relative Power Right Fronto/Temporal 0.76931 -0.20846 Bipolar Absolute Power Fronto/Temporal Delta -0.80124 -0.05935 Bipolar Coherence Fronto/Temporal Beta -0.26573 0.60565 Bipolar Coherence Posterior Beta 0.45822 -0.59687 Bipolar Asymmetry Fz Theta Monopolar Relative Power -0.07747 -0.96582 Constant -1.79117 -2.61208 ______________________________________ RULE-OUT LEVEL: ______________________________________ Alcoholic 65 (if P.sub.i ≧ 65%, rule out alcoholism) ______________________________________

Claim 19

A quantitative process as in claim 9 in which the step of deriving disciminant scores comprises deriving said discriminant scores in accordance with the features and coefficients of the following table for the diagnostic conditions stated therein: ______________________________________ BIPOLAR DEPRESSION vs ALCOHOL ABUSE ______________________________________ VARIABLE BIPOLAR ALCOHOLC ______________________________________ Right Parietal/Occipital Beta 0.58452 0.08115 Bipolar Relative Power Parietal/Occipital Delta -0.72523 -0.20808 Bipolar Coherence Fronto/Temporal Delta -0.95221 0.20967 Bipolar Coherence Parietal/Occipital Beta -0.01620 0.45588 Bipolar Coherence Posterior Bipolar Asymmetry 0.65121 -0.80888 Posterior Alpha 0.83039 -0.51708 Monopolar Relative Power Fz Theta Monopolar Absolute Power 1.02508 -1.29232 Constant -2.16285 -1.88824 ______________________________________ RULE-OUT LEVEL: ______________________________________ Alcoholic 51 ______________________________________

Claim 20

A quantitative process as in claim 9 in which the step of deriving disciminant scores comprises deriving said discriminant scores in accordance with the features and coefficients of the following table for the diagnostic conditions stated therein: ______________________________________ ELDERLY DEMENTIA vs ALCOHOL ABUSE ______________________________________ VARIABLE DEMENTIA ALCOHOLC ______________________________________ Left Central Theta -0.03469 0.91110 Bipolar Relative Power Right Central Theta 1.06487 -1.18625 Bipolar Relative Power Left Central Beta 0.03015 0.89450 Bipolar Relative Power Parietal/Occipital Beta -0.08565 0.89726 Bipolar Coherence C3-C4 Theta -0.40997 0.09382 Monopolar Asymmetry Constant -1.82594 -1.98809 ______________________________________ RULE-OUT LEVEL: ______________________________________ Alcoholic 51 ______________________________________

Claim 21

A quantitative process as in claim 9 in which the step of deriving disciminant scores comprises deriving said discriminant scores in accordance with the features and coefficients of the following table for the diagnostic conditions stated therein: ______________________________________ UNIPOLAR DEPRESSION vs ALCOHOL ABUSE ______________________________________ VARIABLE UNIPOLAR ALCOHOLC ______________________________________ Left Central Beta -0.12390 0.86921 Bipolar Relative Power Left Fronto/Temporal Beta -0.31110 0.86605 Bipolar Relative Power Fronto/Temporal Delta -0.59866 -0.14523 Bipolar Coherence Left Central 0.88585 -0.14734 Bipolar Absolute Power Parietal/Occipital Beta 0.00458 0.90921 Bipolar Coherence Posterior Bipolar Asymmetry 0.50132 -0.57529 Constant -1.71888 -2.47368 ______________________________________ RULE-OUT LEVEL: ______________________________________ Alcoholic 51 ______________________________________

Claim 22

A quantitative EEG method comprising: deriving quantitative EEG data for an individual and subjecting said EEG data to preprocessing including artifact rejection to reduce said EEG data from artifact-free segments, subjecting said artifact-free segments to spectral processing to extract desired features thereof, applying transforms to said features as needed to ensure Gaussianity and Z-transforming said features on the basis of normative age regression data for a population of individuals assumed to be normal to thereby derive Z-scores for selected features and any desired combinations of features; and deriving discriminant scores for selected diagnostic categories from Z-scores derived from an individual's EEG data by combining for each respective discriminant function selected ones of the individual's Z-scores weighted by selected coefficients specific to the respective diagnostic category, combining selected discriminant scores to evaluate the probability that the individual belongs to a particular one of two or more selected diagnostic categories and applying guardbands to the evaluated probabilities to ascertain the level of confidence that the probability correctly classifies the individual into a diagnostic category.

Claim 23

A quantitative EEG method comprising: deriving quantitative EEG data for an individual and subjecting said EEG data to preprocessing including artifact rejection to reduce said EEG data from artifact-free segments, subjecting said artifact-free segments to spectral processing to extract desired features thereof, applying transforms to said features as needed to ensure Gaussianity and Z-transforming said features on the basis of normative age regression data for a population of individuals assumed to be normal to thereby derive Z-scores for selected features and any desired combinations of features; and deriving discriminant scores for selected diagnostic categories from Z-scores derived from an individual's EEG data by combining for each respective discriminant function selected ones of the individual's Z-scores weighted by selected coefficients specific to the respective diagnostic category, combining selected discriminant scores to evaluate the probability that the individual belongs to a particular one of two or more selected diagnostic categories and applying guardbands to the evaluated probabilities to ascertain the level of confidence that the probability correctly classifies the individual into a diagnostic category.