PropertyValue
is nif:broaderContext of
nif:broaderContext
is schema:hasPart of
schema:isPartOf
nif:isString
  • Study 1-Intrinsic Schumann Resonance in Human Brain Activity: The subjects in this study were 184 individuals, measured singly, who had participated in various experiments within the laboratory between 2009 and 2013. Some subjects were measured more than once so that the total numbers of records were 237. As part of the laboratory protocol, eyes-closed measurements were collected at the beginning of each experiment before testing. While exact indices of age of the participants were not available, the majority of the individuals included in this study were university students between 19 and 25 years of age. Some of the individuals had been referred to the third author’s private practice for psychometric and electroencephalographic assessment (N = 45). The proportion of men (N = 109) and women (N = 128) were similar. Most measurements were completed within a commercial acoustic (echoic) chamber under dim light conditions, otherwise measurements were obtained in quiet rooms on the University campus from participants who volunteered for participation in 4th-year thesis projects. Brain electrical activity was monitored using a Mitsar 201 amplifier equipped with a 19-channel Electro-Cap International. Measurements from 19 sites (Fp1, Fp2, F7, F3, Fz, F4, F8, T3, C3, Cz, C4, T4, T5, P3, Pz, P4, T6, O1, O2) consistent with the International Standard of Electrode Placement were obtained. Impedance for all measurements was maintained below 5 kOhms. The data recorded from the amplifier was delivered to a Dell laptop equipped with WinEEG v.2.8 which produced a digital copy of the recorded voltages. Sixteen-second epochs of eyes-closed data were extracted from each participant and exported into MATLAB software for further filtering and processing. While most data collected with the amplifier were obtained using a sampling rate of 250 Hz, some measurements were collected with a sampling rate of 500 Hz. To insure homogeneity across subjects, data that were collected using a sampling rate of 500 Hz was re-sampled to 250 Hz using the resample.m function within the MATLAB platform. The data for each subject was then filtered between 1.5 and 40 Hz using the eegfiltfft.m function within the freely available EEGLab toolbox [19]. The function uses an inverse FFT algorithm to band-pass filter raw measurements within a specified frequency range. We have found qualitatively and quantitatively that this filtering algorithm produced identical results independent of whether the segment length was 16 seconds or 120 seconds in duration; the correlation between the raw voltage recordings was 0.996. Once filtered, the data were submitted to spectral analysis, using the spectopo.m function, which computed spectral density within discrete frequencies for the Fp1 O2 T3 and T4 sensors using Welch’s periodogram method employing a window size of 2048 (8.2 seconds) to maximize spectral resolution and a Hamming window with 50% overlap between windows. These data was then imported into SPSS for further analysis and for the computation of mean absolute potential difference along the rostral-caudal (Fp1-O2) axis as well as between the left-right temporal (T3 and T4) sensors. Absolute differences between rostral-caudal and left and right temporal sensors were also obtained by subtracting the absolute raw voltages (independent from spectral density) from the extracted EEG record for each 4-millisecond point interval over 16 seconds. Random Sample of 10 Brain Measurements: To appreciate the difference between a sub-sample of the total population and the total population, 16 s of QEEG data (Fp1 02, T3 T4) for 10 subjects were randomly selected using a random number generator. Z-scores for the μV2∙Hz-1 values for the rostral-caudal, left-right measures were obtained first in order to minimize individual differences in absolute baseline values. Spectral analyses from SPSS 16 (a different algorithm from the one applied in the previous section) were then applied to each of the two measures for each of the 10 subjects. The frontal spectral density scores were then subtracted from the occipital spectral density scores for each subject. The z-scores for each of these new means were calculated so that there would be a standardized score for these differences to compare directly across individuals. The average values for the two orthogonal measures were calculated. In order to examine the degree of individual differences of the spectral analyses each subject’s results were assessed visually for the peak spectral density within the 10 Hz range. Because the Δf (increment of frequency) within this frequency range for 250 Hz samples is fractional (less than an integer, i.e., about .01 Hz), the band width of the peak spectral density could be inferred. This was discerned by direct inspection of the quantitative values where the decline in z-scores on either side of the peak was conspicuous and greater than 2 standard deviations for the interval. Discerning the Schumann Resonance Signature in the Brain: After re-sampling and re-filtering between 1.5 and 40 Hz for the 184 individuals by eegfiltfft.m, the anterior, middle, and caudal root mean square signals were derived from integrating frontal (Fp1,Fp2,F7,F3,Fz,F4,F8), middle (T3,C3,Cz,C4,T4) and caudal (T5,P3,Pz,P4,T6,O1,O2) sensors. These derived signals were spectral analyzed by spectropo.m with a window size equal to 2048 FFT points to maximize spectral resolution employing the same spectral analysis parameters mentioned above. Other researchers, e.g., Abeyuriya et al [20], have recently employed the same approach to discern non-linear harmonics of sleep spindles in human electroencephalographic recordings. Because of a connection between activity within the parahippocampus and naturally-occurring geomagnetic activity established in a previous publication [10] and its prominent role as an integrator of experience before long-term memory processes within the hippocampus, inferences of left and right parahippocampal current source density (μA·mm-2) were computed in sLORETA [21] within the classical frequency bands used in more conventional electroencephalographic studies [i.e. delta (1.5–4 Hz); theta (4–7 Hz); low-alpha (7–10 Hz); high-alpha (10–13 Hz); beta-1 (13–20 Hz); beta-2 (20–25 Hz); beta-3 (25–30 Hz); gamma (>30 Hz)]. Corradini and Persinger [22] have found that source localization performed with 19 sensors within a clinical population was able to localize the source of infarctions and closed-head injuries and was validated against accepted psychometric inferences of the functioning of these regions. We employed the exact same methodology for computing the current source density for the parahippocampal regions as in a previously published paper [10]. Each of the 237 16-second recordings were entered into sLORETA. First, the 19-channel series was translated into the frequency-domain using the EEG to Cross-Spectrum function and computed cross-spectral densities in the defined frequency bands [i.e. delta (1.5–4 Hz); theta (4–7 Hz); low-alpha (7–10 Hz); high-alpha (10–13 Hz); beta-1 (13–20 Hz); beta-2 (20–25 Hz); beta-3 (25–30 Hz); gamma (>30 Hz)]. Next the Cross-Spectrum to sLORETA function was used to compute current source density; the transformation matrix used here was defined by the 19-channel array produced within the software. Finally inferences of left (MNI-co-ordinates: X = -28, Y = -40, Z = -12) and right (X = 28, Y = -40, Z = -12) parahippocampal current source density for each of the frequency bands (for each participant) were extracted by using the sLORETA to ROI function using a customized ROI seed file. The same epochs of electroencephalographic activity used in the sLORETA analyses were also imported into MapWin software [23] for the computation of microstates and their related parameters (i.e. duration of microstate, occurrence of microstate class, etc.). The strategies involved in producing microstates involves clustering of topographic maps based upon voltage. Because of the ‘stochastic’ nature of EEG signals, only those maps that exceed a certain threshold of ‘signal to noise’ are included [24]. The criterion for this threshold is defined by global field potential (GFP); only those maps with elevated GFP peaks that are high are considered for clustering. This equation is expressed as a spatial standard deviation and can be modeled mathemtaically as [24]: GFP=(∑(v(t)−V(t)2/n)1/2(1) where v is the voltage at a given channel, t is the time point of interest, and V is the mean voltage across all channels. These maps are then entered into a k-means clustering algorithm within the MapWin software itself and classes of microstate maps are produced irrespective of polarity. Effectively the resultant ‘classes’, or clusters, are the mean centers for each channel interpolated onto a 5x5 matrix with approximate locations of channels identified. The montage in this case was: Fp1Fp2F7F3FzF4F8T3C3CzC4T4T5P3PzP4T6O1O2 The 16-second epochs for each of the 237 available records were segmented into 4x4 second segments for later averaging. Then, microstate computations were computed separately for each participant utilizing the following methods. First, each of the 4 segments were imported into MapWin where a digital filter between 2–20 Hz was applied, in accordance for standard procedures described by Koenig et al [25]. Next, microstate clusters were then produced using the Compute Microstates function. Here we selected to produce only 4-clusters (maximum = 8) with a convergence criterion set to 25 iterations. All maps produced here were polarity insensitive and were only generated at GFP peaks. After this process was repeated for all 237 cases, the resultant average microstate maps (N = 237) were entered into the Combine Microstates function which clustered all of the scalp maps of each individual to produce 4 mean microstate-classes; the classes were almost identical to those reported by Koenig et al. [25] and explained approximately 72% of the variance in scalp map classification. Next, 4 microstate statistics (occurrence, coverage and duration) for each microstate class as well as model percent variance (which effectively described how well the average model fit the individual subjects) were computed within the software. To accomplish this all 4-second segments (of each participant) were re-imported into Mapwin software. The average 4-class model computed above was then applied to each of these segments (N = 237x4) separately and statistics were computed using the Microstate Statistics function which exported each statistic for each epoch into a Microsoft Excel file. Averages of each of the statistics for each individual were then computed. Finally, the raw voltages, spectral densities, inferences of parahippocamapal activity, and microstate parameters were exported into a singular SPSS dataset for further analyses.
rdf:type