Supporting Research and Applications


EEG Acquisition remains the same, using standard and accepted guidelines.


Currently there are over well over 100,000 publications in the National Library of Medicine, the vast majority includes studies using digitized EEG. A large number of these, while titled as “EEG”, are actually techniques of Quantitative EEG studies. These publications represent peer reviewed journals and an array of textbooks.

A large number of these, while titled as “EEG”, are actually techniques of Quantitative EEG studies. These publications represent dozens of peer reviewed journals and an array of textbooks.

Foundations of Measurement:

BrainDx combines historic applications of systems theory with cutting edge advances in imaging and statistical prediction

Validity – the accuracy in which meaningful and relevant measurements can be made both face and construct, convergent.


Reliability – the reproducibility of conditions or findings using a method of measurement that incorporates test- re-test and cross validation.


Methods used in Neurometrics:

Digital EEG collected from 19 regions of the International 10/20 placement system
Impedances all below 5000Ωs
Referenced to linked ears
Bandpass = 0.5 to 50 Hz, Sample rate >= 100Hz
20-30 min. EEG recorded
State = eyes closed resting
selection of 1-2 min of artifact-free data

Computation of Features

Power Spectrum and related features including:
Absolute power
Relative power
Mean frequency
Intra- and Inter-hemispheric symmetry
Intra- and Inter-hemispheric coherence

Computed for Each Bandpass
Delta (1.5-3.5Hz), Theta (3.5-7.5 Hz), Alpha (7.5-12.5 Hz), Beta (12.5-25 Hz) and Beta2(25-35Hz)

Source Localization
Measures from LORETA Regions of interest (ROIs).
for narrow frequency bands,
and precisley time coded events.

Additional non-linear features include measures of complexity and connectivity


Approach to Norming

Child and Adult norming followed the same procedures as described in published articles. Initial norming was done under government funding (BEH, NSF, NIA) and followed strict protocols for inclusion/exclusion criteria and data acquisition (details given in publications). The following points are important to note:

Norms were constructed using split-half approach where regression equations were formed on one half and tested on the other (independent replication) – with the expectation that less than the expected random number of “hits” (z-values <0.05) were obtained on the second half. The two halves were then combined and final equations constructedSubjects were added to the population until adding more subjects (across the age range) did not change the regression equation. Thus, the size of the population required to norm were statistically determined. As expected many more children were required (more rapid change across age) than adults

Additional evaluations were used to demonstrate high test/retest reliablity and stability of the normsOver the years since initial norming additional subjects have been added to the child and adult populations to represent advances in amplifiers, etc. All new subjects were recruited according to the same criteria as the initial projects

Adult Sample

N = 154 Selected based on extensive psychiatric and neuropsychological evaluations Psychiatric/Neurologic examination Evaluations of achievement, dominance (hand,eye, foot) IQ had to be normal

Detailed developmental, medical, psychosocial histories

Exclusion variables: use of drugs, history of head injury or loss of consciousness, previous

EEG or neurological examination, febrile convulsions

Child Sample

N = 310
Normal Medical and Developmental histories

Excluded extreme prenatal or perinatal trauma
High febrile illness
Loss of consciousness (concussions, convulsions)
Extreme behavior problems
Failure at any school level
WRAT scores below 90 on any skill

John, 1987 (Handbook Chapter);John et. al, 1988;

Age Distribution of Norming Subjects, Closed Eyes Condition
sLORETA Norming

LORETA (Low Resolution Electromagnetic Tomographic Analyses) is a source localization inverse problem method for localizing the mathematically most probable source of the voltages recorded from the scalp.

Working together with Roberto Pascual-Marqui at the BRL, voxel norms were computed using the same BRL/NYU normative database

The methodology described in Neurometrics was applied to the sLORETA norming, allowing the z-transformation of each voxel in the model which can be displayed as statistical color-coded images of the mathematically most probable underlying sources of the scalp recorded EEG data

For each voxel, an individual’s values are compared statistically to the expected norms for their age;

Statistical significance for each voxel is encoded in color superimposed upon slices from a Probabilistic MRI Atlas;

Extensive literature exists demonstrating similar findings with conventional neuroimaging and EEG source localization

The age regression equations that were developed help standardize the quantitative EEG measures so that they may be interpreted independent of the subjects age.

The sLORETA images below are plots of the correlation of subject-wise relative power grey matter voxels with age over the range of 16 to 80 years (N = 154). The regression is linear (a straight line fit) with the logarithm of age.

The two volumes shown each comprise 20% of the grey matter volume.

All the voxels in the red volume have a positive correlation with age greater than .48 at the frequency 17.2 Hz. The maximum value of .58 is in the left Insula.

All the voxels in the blue volume have correlation less than -.44 at the frequency of 10.2 Hz.

In general this illustrates an increase in Beta in the temporal lobes and and a decrease in Alpha with increasing age.

The current density estimate of each voxel is divided by the total energy of the EEG, (subject-wise relative power), thus removing the influence of the overall size of the EEG from this measure.


Multivariate Measures

Using Z-scores allows a common metric that allows computation of multivariate “system” features. These Multivariate computations form a super-set of features that are often important in summarizing concepts like degree of abnormality and they make important contributions to the discriminant functions described below.more typos:

Discriminant Functions

BrainDx offers the use of Multivariate discriminant Analyses to statistically evaluate the match of patient qEEG profile with specifically defined clinical profiles to augment diagnostic processes. It is important to note that this methodology is not intended to be used as a substitute for current psychiatric or psychological diagnostic methods but strictly as a supplemental tool to help with the confirmation of a diagnostic consideration. There are strict criteria for the use of these discriminant functions and the BrainDx software will direct the user to be able to use only those functions which meet history and symptom criteria.


Examples of Discriminate Functions for DSM Clinical Groups

Primary Progressive Dementia (Alzheimer’s Type Dementia)

  • Depression as distinguished from Dementia
  • Vascular Dementia

Major Affective Disorders (Depression)

  • Unipolar as distinguished from Bipolar Depression


Learning Disabilities (LD)

  • Normal vs LD
  • Normal vs ADHD
  • Stimulant (e.g., Ritalin) Responder as distinguished from non-Responder


Autism Spectrum Disorder (ASD)

  • ASD as distinguished from ADHD

Co-morbid Alcohol Abuse

Obsessive Compulsive Disorder

Post-Traumatic Stress Disorder vs Post Concussive Syndrome (In Development)


Applying Z-score statistics with source localization, the degree of functional deviation for age can be better visualized and compared to other forms of neuroimaging when desired