We achieve stable and reproducible FFTs in three steps:

  1. Permit the user to select a minimum of 600 msec selections of artifact free digital EEG samples and then splice together all of the user’s selections as a series of continuous 256 digital values at 128 samples/second (2 second epoch length) of “artifact free” EEG samples;

  2. Remove splice artifact (maximum = 2 per any 256 sample) by padding zeros as EEG data points at t = 0 and then applying a 5th order Butterworth band pass filer 1 Hz to 40 Hz to baseline and minimize the “discontinuities” of splicing the EEG samples over the entire selection of “artifact free” EEG; and

  3. Use sliding averages of overlapping FFT windows (75% overlap as Kaiser and Sterman, J. Neurotherapy, 2001) so as to minimize the boundaries of windows = 0.

We experimented with different degrees of overlapping windows and different filters and we computed the cross-validated Z scores for each of these experiments. We essentially replicated Kaiser and Sterman’s method to reduce window effects and we achieved quite reliable results no matter what the splice segment lengths with overlapping of the 2 second epochs of EEG for both the normative EEG samples and patient EEG samples. Tests using the worst case scenario of segment splicing of the troughs and peaks of sine waves revealed side band ringing in the FFT which was maximally about 5% of the signal. This worst case scenario is never obtained in practice and the effects of splicing are < 5% by limiting the minimal segment length to 600 msec and by filtering.

The users of NeuroGuide™ can compare the effects of windowing in the FFT by comparing the ASCII FFT spectral values for overlapping vs. non-overlapping FFTs.

Click on the menu Analysis>LORETA Export>Overlapping Windows and save. Repeat by clicking Analysis>LORETA Export>Successive in order to save non- overlapping FFTs of the same data. NeuroGuide™ gives an option for users to export with or without splice correction. However, even in this case it is best to select artifact free EEG and to analyze EEG digital data using the equivalent or approximately the same system that acquired the original EEG digital values.