Method of classifying raw EEG signals

A computer implemented method of classifying raw electroencephalogram (EEG) signals by extracted features for the detection of high ability and low ability with regard to fluid intelligence in a patient, comprising the steps of:

• placing a set of electrodes on the patient's scalp, wherein the placement includes at least the frontal-left and frontal-right and the parietal-left and the parietal-right brain regions;
• providing a computer system, the computer system including a processor and a memory device having instructions stored thereon, wherein the instructions in response to executions by the processor cause the processor to:
• receive a set of raw EEG signals directly from a set of EEG channels associated with the brain regions of the patient, wherein the raw EEG signals are associated with a brain state activated by an event, the set of raw EEG signals being represented as x1, x2, x3, . . . , xn, where n represents a total number of the EEG channels;
• calculate a mean EEG signal, x, wherein the step of calculating the mean EEG signal comprises calculating x as
• x
• _
• =
• 1
• n
• ⁢
• ∑
• 1
• n
• ⁢
• x
• i
• ;
• calculate a standard deviation of the set of EEG signals, std, as
• std
• =
• 1
• n
• ⁢
• ∑
• 1
• n
• ⁢
• (
• x
• i
• -
• x
• _
• )
• 2
• ;
• calculate a set of zero-centered EEG signals, wherein an ith one of the zero-centered EEG signals, yi, is calculated as yi=xi−x for i=1, 2, 3, . . . , n;
• calculate a unit variance for each of the zero-centered EEG signals, wherein an ith one of the unit variances, ϕi, is calculated as
• ϕ
• i
• =
• y
• i
• std
• and constitutes a transformed signal;
• calculate a nuclear matrix, N, as N=ATA, wherein A is a matrix representing the transformed signals and being formed as A=[ϕ1 ϕ2 . . . ϕn], further wherein the matrix represents either high ability or low ability;
• apply singular value decomposition to the nuclear matrix to generate a set of singular values, such that N=UDVT, where U and V are unitary matrices and D is a diagonal matrix formed from a set of singular values, wherein an ith one of the singular values is represented as σi for i=1, 2, 3, . . . , n, σ1 and σ2 being defined as dominant singular values such that a nuclear feature vector, F, is defined as F=[σ1, σ2]T;
• use a class means based minimum distance classifier (CMMDC) to calculate a first distance, d1, as d1=|F−Cμ1|, wherein Cμ1 represents a mean nuclear feature vector of a first class of a sample dataset established through prior EEG readings of a sample set of test subjects measuring fluid intelligence;
• use a class means based minimum distance classifier (CMMDC) to calculate a second distance, d2, as d2=|F−Cμ2|, wherein Cμ2 represents a mean nuclear feature vector of a second class of the sample dataset;
• classify the set of raw EEG signals of the patient as being in the second class if d1>d2, and otherwise classifying the set of EEG signals of the patient as being in the first class, wherein the first class represents low ability and the second class represents high ability;
• input the set of raw EEG signals directly from the set of EEG channels associated with the brain regions of the patient into a support vector machine (SVM) for generating a classification of the fluid intelligence of the brain regions being assessed and comparing the classification results with the CMMDC result to assess the accuracy of the CMMDC classification;
• and display the CMMDC classification result.