2017: A Model for QEEG, and sLORETA Correlates to Predicting and Enhancing Human Performance: A Multivariate Approach (Plenary Session)

The derivation of QEEG predictive functions may play a critical role in using QEEG measures that predict potential functional weaknesses or individuals “at risk” for specific academic or occupational challenges. Such measures can also hypothetically be used in a multivariate algorithm approach in neurofeedback to achieve human performance enhancement. EEG was obtained from 493 individuals ranging in age from 4 to 75 years diagnosed with a variety of disorders. All EEG data used a 19 site monopolar montage and referenced to Cz in acquisition and utilizing linked ears reference to derive digitized information in each of the following broad bands: Delta (1.5-3.5 Hz), theta4-7.5 Hz), alpha (8-12.5 Hz), beta (13-25 Hz) and beta2 (25.5-35 Hz) frequency bands with derived measures of absolute power, relative power, power asymmetry, (inter and intrahemispheric), coherence (inter and intrahemispheric), and mean frequency for each broad band was attained and then converted to z-scores relative to a database of age-matched normal. Univariate as well as complex multivariate variables collapsed across selected regions and combination of frequencies were derived for a total of 13,712 variables. Additionally, the sLORETA voxels including weighted function voxels for subcortical structures were derived at very narrow band frequencies (.39 Hz bands) ranging from 1.5 Hz to 35.5 Hz of the EEG (87 variables). Z-scores of all voxels for each ROI standard in sLORETA as well as a number of weighted voxels estimating subcortical locations were derived for each very narrow band frequency for a total 6,896 variables. Data reduction methods for this total 599,952 variable matrix were utilized by deriving the mean Z-score score of all voxels within each ROI and then averaging these mean Z-scores across the narrow band frequencies that define each of a number of broad band frequencies for each subject. Step-wise regression analyses of the resultant reduced variable sets were used to define specific weighted polynomial multivariate equations accounting for over 90% of variance that predict standard scores from neuropsychological tests and their subtests for many cognitive and behavior measures. Analyses revealed distinctive predictive equations for human performance spanning a wide range of human performance. It is proposed that these algorithms represent electrophysiological base networks (as opposed to fMRI based networks) at “resting state” that correspond to gradients of psychological performance. Pilot data from equations demonstrated predictive ability to test neuropsychological performance were tested in independent patient samples to test validity and reliability. This study demonstrates that QEEG can be used to screen for brain functional impairment with prediction of specific neuropsychological deficits that may require further assessment and intervention. A discussion will be provided regarding the use of these same algorithms for neurofeedback training optimized human performance.