Neurophysiology, Vol. 44, No. 1, April, 2012 Age-Related Changes in the Energy and Spectral Composition of EEG O. Vysata,1 J. Kukal,2 A. Prochazka,2 L. Pazdera,3 and M. Valis4 Received November 9, 2011. Age-related changes in the EEG energy and spectral composition were examined in 17,722 healthy subjects (truck drivers), 20 to 70 years old. Linear correlations between age vs global EEG energy and spectral powers (SPs) of EEG frequency ranges were estimated by linear regression analysis. Significant dependences of the global EEG energy and SPs of all EEG rhythms on age (age-related decreases) were found (most significant for the alpha range). An age-related decline of the EEG energy may be, in part, explained by age-related generalized brain atrophy and increased thickness of the scalp. While similarity measures show an agerelated decrease of the correlation between different sources of EEG, activity phase cancellation should also be taken into account. Smaller decreases in the absolute SPs of the beta and gamma ranges together with a general drop in the overall EEG power result in a trend toward relatively higher normalized representation of these frequency components with aging. The highest significance of decline in the alpha range can be related to a slowing oscillation frequency of alpha generators toward the theta band. Keywords: electroencephalography, power spectral density, energy, aging, normal values. Aging is a multifactorial process caused by accumulated damage to a variety of cellular components in both the nervous system and all other systems of the organism. Increasing age in mammals correlates with increased levels of mitochondrial DNA mutations and deteriorating of the functions of the respiratory chain. Respiratory chain-deficient cells are prone to apoptosis, and increased cell loss is likely to be an important consequence of age-associated mitochondrial dysfunction [1]. In the human brain, this results in shrinkage of large neurons and increases in the relative numbers of small neurons and gliocytes [2]. It is logical to believe that these shifts should inevitably be reflected in the EEG global power and frequency composition. The generalized age-related brain atrophy means that there are some decreases in the number of neurons producing electrical potentials and in the number of neural connections. A reduction of EEG slow-wave activity (power density in the range 0.75 to 4.5 Hz) and increase in the amount of spindle activity were found in older people during sleep [3]. According to another clinical study, the aging process does not differentially influence the EEG power density during sleep in men and women. There are data that the decrease in the EEG power with age was not restricted to the slow-wave activity but also included theta and gamma activity. It was also reported that increasing age was associated with a higher power of the beta range [4]. Thus, there are some contradictions with respect to age-related modifications of the pattern of EEG activity. In our study, we examined age-related changes in the EEG global energy and frequency composition in a rather representative sampling of healthy subjects. 1 METHODS INTRODUCTION 2 3 4 Institute of Chemical Technology, Prague, Czech Republic. Czech Technical University, Prague, Czech Republic. Neurocenter Caregroup Ltd., Rychnov and Kneznou, Czech Republic. Faculty Hospital of the Hradec Kralove Charles University in Prague, Czech Republic. Correspondence should be addressed to O. Vysata (e-mail: vysatao@gmail.com). Subjects. In our retrospective population study, EEG data were obtained during examinations of 31,009 truck drivers. All subjects underwent neurological and neuropsychological examinations. Some of the subjects 63 0090-2977/12/4401-0063 © 2012 Springer Science+Business Media, Inc. O. Vysata et al. 64 (n = 879) with potential damages to the brain in the anamnesis were excluded from the studied group. The main exclusion criteria were alcoholism, drug abuse, abnormal results of neurological examination, and abnormal results of neuropsychological examination. Another part (12,390 EEG recordings) was also excluded due to high noise levels that could not be filtered or corrected. The remaining 17,722 subjects included 17,540 men and 182 women, from 20 to 70 years old, with a mean age of 43.2 ± 11.2 (M ± s.d.) years. EEG. All recordings were performed under similar standard conditions. The subjects were in a comfortable position, on a bed, with their eyes closed. Electrodes were positioned according to the 10-20 system of electrode placement; the recording was conducted on a 21-channel digital EEG setup (TruScan 32, Alien Technik Ltd., Czech Republic) with a 22-bit AD conversion and a sampling frequency of 128 sec –1. Filter settings were 0.5 and 60 Hz. The linked ear contacts were used as references [5]. EEG Pre-Processing. Stored digitized data were zero-phase digitally filtered using a bandpass FIR filter (100 coefficients, Hamming window) of 0.560 Hz and a bandstop filter of 49-51 Hz [6]. The analysis started by automatic artifact recognition and removal. Such recognition was based on the regularized LD classifier and removed 91% of the artifacts. An off-line expert then manually removed artifacts from the pre-processed data. The EEG signal was not “prewhitened” because “pink” (or complex) noise with the amplitude dependence of the frequency A = 1/f · α is a characteristic feature of cerebral electrical activity [7]. The parameter α is highly invariant across the subjects [8]. Global EEG Energy. Global EEG energy was counted as the mean absolute value of the summarized amplitude per sample (µV/sample). Power Spectral Densities. Power spectral densities (PSDs) of the EEG frequency components were estimated using the Welch method. These parameters for EEG samples recorded from 19 leads were estimated for the delta (0.5-3.5 Hz), theta (4-7.5 Hz), alpha (8-12.5 Hz), beta (13-29.5 Hz), and gamma (30-60 Hz) ranges as µV 2/Hz. Relative PSDs were calculated as ratios of the absolute PSD within the corresponding range vs the integral energy in all ranges. Statistics. We investigated the presence of significant linear correlations between the values of age and energy (PSD) by linear regression analysis [9]. The F statistic and its P value were calculated. Estimates of the error variance at the 99% confidence interval are shown in the graphs. RESULTS Despite the rather broad natural interindividual variability, there was a clear age-related trend toward a decline in the global EEG energy (Fig. 1). For this index, the y-intercept, slope, and P value were 6.65 µV/sample, –3.2 · 10 –2, and <10 –20, respectively. The same can be said as to the absolute values of the PSDs of all EEG frequency ranges (Table 1). The yearly decline for the global energy was found to be 0.47 %. Despite the fact that the global energy and PSD values were computed using dissimilar methods, there was a good correspondence between them. The global energy estimated as the mean absolute value of amplitude per sample (y-intercept = 6.65 µV/sample) corresponds well to the value derived from the summed PSDs of all frequency ranges (y-intercept = = 6.76 µV/sample). The highest significance of changes was found for the alpha rhythm (Table 1, Fig. 2). The yearly declines of the regression graphs for absolute values of PSDs of the delta, theta, alpha, beta, and gamma ranges were 1.16, 1.02, 0.75, 0.27, and 0.18%, respectively. Values of the relative energy (PSD) demonstrated significant trends toward decreases in the delta, theta (Fig. 3), and alpha ranges, while mild but significant increases in the beta and gamma ranges were observed (see negative and positive signs of the slopes in Table 2, Fig. 4). TABLE 1. Significance of the Trends of Age-Related Changes in the Absolute PSDs of the EEG Frequency Ranges EEG rhythm Y-intercept Slope P value Delta 473.2 –5.5 1.1 · 10–8 Theta 132.1 –1.4 9.1 · 10–13 Alpha 121.8 –0.9 <10–20 –3 Beta 3.3 –8.8 · 10 5.2 · 10–6 –3 Gamma 0.86 –1.6 · 10 4.4 · 10–4 TABLE 2. Significance of the Trends of Age-Related Changes in the Relative PSDs of the EEG Frequency Ranges EEG rhythm Y-intercept Slope P value Delta 0.40 –1.2 · 10–3 <10–20 Theta 0.17 –4.7 · 10–4 <10–20 –3 Alpha 0.32 –1.5 · 10 1.9 · 10–3 –3 Beta 0.08 1.5 · 10 <10–20 –4 Gamma 0.03 5.5 · 10 <10–20