Signal complexity indicators of health status in clinical EEG

Figure 1. Brain- and non-brain-acting medication use in study sample. (A) Medication use in patients diagnosed with epilepsy (N = 66); (B) Medication use in patients without a diagnosis of epilepsy (N = 97). Indicators of comorbidities (use of multiple brain-acting medications, non-brain-acting medications, or both) shown in green. BA: brain-acting; nonBA: non-brain-acting. Figure panels created in MATLAB v.9.1 and merged using Adobe Illustrator CS6; colors were also edited in Illustrator for clarity.

Figure 2. Brain signal complexity differentiates brain disorders. (A) Correlation coefficients and (B) bootstrap ratios of the first latent variable relating clinical data to MSE curves. (C) Average (± SEM) MSE curves, with subjects split into two groups according to their LV-scores. MSE curves were first averaged across electrodes within subjects, then averaged across subjects within each group. In (A), variables whose coefficients are significantly different from 0 are indicated in color for ease of interpretation. Figure panels created in MATLAB v.9.1 and merged using Adobe Illustrator CS6.

Figure 3. Brain signal complexity differs for older unhealthy males. Correlation coefficients (A) and bootstrap ratios (B) of the second latent variable relating clinical data to MSE curves. (C) Average (± SEM) MSE curves, with subjects split into two groups according to their LV-scores. MSE curves were first averaged across electrodes within subjects, then averaged across subjects within each group. In (A), variables whose coefficients are significantly different from 0 are indicated in color for ease of interpretation. Figure panels created in MATLAB v.9.1 and merged using Adobe Illustrator CS6.

Figure 4. Spectral power density differentiates epilepsy from other brain disorders. (A) First latent variable (p < 0.001; 50.1% cross-block covariance) (B) second latent variable (p < 0.001; 29.9% cross-block covariance), and (C) third latent variable (p = 0.068; 10.1% cross-block covariance) of a PLS analysis relating clinical data to SPD. Left panels: correlation coefficients; middle panels: bootstrap ratios; right panels: average (± SEM) SPD, with subjects split into two groups according to their LV-scores. SPD functions were first averaged across electrodes within subjects, then averaged across subjects within each group. Figure panels created in MATLAB v.9.1 and merged using Adobe Illustrator CS6.