Rosipal R., Trejo L.J., Rostakova Z, Cimrova B.

Frequency spectra and spatial sources of EEG oscillations are highly specific to individuals and may be manifested differently as functions of pharmaceutical treatments or other interventions. However, most analyses of treatment effects on EEG oscillations use an approach based on standard frequency band powers measured at single electrodes. We have developed and applied a method that accurately and efficiently models individual EEG oscillations and tracks their activation over time or treatment conditions. <p>

To obtain robust and repeatable individual measures of EEG suitable for analytical and statistical testing, we developed and refined a novel approach using irregular-resampling auto-spectral analysis (IRASA) to separate fractal and oscillatory components in the EEG power spectrum and three-way parallel factor analysis (PARAFAC) to isolate elemental oscillatory EEG components or ''atoms'' and track their activations; that is, time-scores over time or conditions [1,2]. We automated the whole process by extracting EEG atoms using a set of PARAFAC model parameters and identifying consistent results by clustering the obtained frequency and spatial loadings of the atoms. We apply standard univariate statistical tests and analysis of (co)variance models to determine statistical significance of changes in atom activations across treatment and experimental conditions. <p>

We simulated effects of dose-related EEG changes by including a range of amplitude effects and signal-to-noise ratios, serving to define the sensitivity and specificity of the approach in comparison to the standard EEG testing based on wider spectral band ranges and separate spatial locations. In one application, using EEG data obtained during the longitudinal motor neurorehabilitation treatment of stroke patients, i) we successfully identify dominant oscillatory patterns associated with the induced motor-related EEG changes, ii) tracked their changes during the treatment period, and iii) demonstrated consistent long-term effects. In another application set, we modeled and tracked the changes in atomic EEG activations as functions of drugs under development for treatment of CNS disorders. In a prior Phase 1B clinical trial of a small molecule being developed for treatment of major depressive disorder, we used PARAFAC atoms to track changes in oscillatory sources in the alpha band and demonstrated significant differences between placebo and actively dosed healthy volunteers. Currently we are using PARAFAC to track EEG changes over time after dosing with other molecules being developed to treat central nervous system disorders. We plan a wide range of future applications to drugs being tested for CNS disorders, which will allow us to verify and validate the atomic EEG decomposition method in comparison to classic EEG band-power analyses.<p>

References <br>
1. Wen H, Liu Z. Separating fractal and oscillatory components in the power spectrum of neurophysiological signal. Brain Topogr. 2016; 19:13-26. <br>
2. Bro R. PARAFAC: Tutorial and applications. Chemom. Intell. Lab. Syst. 1997; 38:149-171.