Individual/Peak Gamma Frequency: What Do We Know?

In recent years, the concept of individualized measures of electroencephalographic (EEG) activity has emerged. Gamma-band activity plays an important role in many sensory and cognitive processes. Thus, peak frequency in the gamma range has received considerable attention. However, peak or individual gamma frequency (IGF) is rarely used as a primary measure of interest; consequently, little is known about its nature and functional significance. With this review, we attempt to comprehensively overview available information on the functional properties of peak gamma frequency, addressing its relationship with certain processes and/or modulation by various factors. Here, we show that IGFs seem to be related to various endogenous and exogenous factors. Broad functional aspects that are related to IGF might point to the differences in underlying mechanisms. Therefore, research utilizing different types of stimulation for IGF estimation and covering several functional aspects in the same population is required. Moreover, IGFs span a wide range of frequencies (30–100 Hz). This could be partly due to the variability of methods used to extract the measures of IGF. In order to overcome this issue, further studies aiming at the optimization of IGF extraction would be greatly beneficial.


Introduction
In recent years, the concept of individualized measures of electroencephalographic (EEG) activity has emerged. For example, peak alpha frequencies or peak theta frequencies are frequently assessed in various conditions and are used as markers for certain processes to track the effect of behavioral manipulations or neurostimulation [1][2][3]. Similarly, interest in peak frequency in the gamma band is increasing [4][5][6][7][8][9].
Gamma oscillations can be measured during the resting-state condition [10], in response to stimulation of basically any sensory modality [11][12][13] and while performing cognitive tasks [14]. However, during the resting state, no prominent peak in the gamma range is observed and it is technically difficult to extract the information [15]. Thus, methods for the estimation/extraction of peak frequencies are being developed and tested [8,16,17]. Nevertheless, it is still a question whether there is a relationship between peak gamma frequency as evoked in response to a certain stimulus or task and the individual gamma frequency (IGF) indexing individual-specific resonant properties of the brain. Indeed, Zaehle et al. [17] demonstrated a correlation between peak gamma frequencies estimated using an auditory steady-state response approach and those extracted from evoked responses to brief auditory stimuli, suggesting that both types of responses might reflect overlapping processes. However, little is known about the nature and functional significance of individual or peak gamma frequency. With this review, we attempt to comprehensively overview available information on the functional properties of peak gamma frequency addressing its relationship with various processes and/or modulation by various factors.
To note, the terminology used to address the phenomenon is highly diverse. While some studies use the terms "peak gamma frequency" [18] or "gamma peak frequency" [19], in others, the term "individual gamma frequency" [7] can be found. All of these terms generally denote a specific individual frequency in the gamma band (mostly  that is characterized by the highest amplitude, power or phase-locking values. Thus, in the present review, the terms "peak gamma frequency" and "individual gamma frequency" are used interchangeably.

Literature Search
To make the inclusion as wide as possible, two searches were performed using the following terms: peak-gamma-frequency OR gamma-peak-frequency OR individual-gammafrequency OR peak-gamma-band OR individual-gamma-band. First, a search in PubMed was performed, resulting in a total of 32 entries. Then an additional search was carried out in Google Scholar, which yielded 964 entries in total. Duplicates, abstracts and review papers were excluded. Titles and abstracts of 847 original articles in English were screened. Specifically, we included only the papers in which (a) individual/peak gamma frequency was estimated and (b) the relationship of individual/peak gamma frequency with certain factors was tested. If the information in the abstract was insufficient, the whole article was viewed. This resulted in 94 papers being mentioned in this review. As the literature search was not conducted on a systematic basis, the evidence reviewed may be not exhaustive.

Methodological Aspects
A systematic review of the search results was not possible due to the fact that IGF is rarely identified as a primary measure of interest and is most frequently reported as a side measure. Furthermore, studies that were selected after the search were characterized by a high heterogeneity of approaches that were used to estimate peak gamma frequencies. In human research, a few studies estimated peak gamma frequency from resting-state EEG activity [20,21]. Others extracted IGF from gamma activity recorded with EEG or magnetoencephalography (MEG) during visual [22,23], auditory [24] or somatosensory [25,26] stimulation or while performing motor [27,28] or cognitive [29,30] tasks. Finally, some authors specifically tested the maximal response range using periodic stimulation [9,31]. A number of animal studies also investigated peak gamma frequency employing electrocorticography (ECoG) during sensory stimulation [32,33] as well as motor [34,35] or cognitive [36,37] activity. In addition, IGF was also extracted from in vitro recordings after inducing gamma activity in brain slices using various substances [38,39]. Depending on the type of stimulation used and its parameters, gamma peaks were estimated from the recordings of different brain areas and in both low gamma (~30-60 Hz) and high gamma (~60-100 Hz) ranges. Despite the high variety of methods and approaches used, for the ease of comprehension, the selected studies were grouped by the factors that demonstrated an association with IGF.

Endogenous Determinants of Peak Gamma Frequency
Peak gamma frequency appears to be a relatively stable characteristic of brain functioning and tends to show high consistency within a subject. Tan et al. [40] demonstrated the stability of visually induced peak gamma frequencies across repeated measurements within a subject. High test-retest consistency of auditory-evoked IGF was also reported [41]. Moreover, van Pelt et al. [42] suggested the genetic determination of the gamma-band peak frequency to be around 91%. In their study, visually induced peak gamma frequencies were highly correlated in monozygotic twins, whereas no correlation was found in dizygotic twins or unrelated subjects. Yet, certain endogenous factors were shown to contribute to the individual stability and variability of IGFs. Among those, anatomical features of the brain, neurochemical balance in normal and diseased conditions and the state of the subject can be highlighted.

Anatomical Features
The size of the neural networks is known to determine the frequency of the network operation [43]. Certain anatomical factors, including the surface area or volume of the visual cortex [44][45][46] and cuneus [47], were shown to be positively related to visually evoked/induced peak gamma frequencies (Table 1). In line with this finding, lesions of the primary visual cortex in monkeys resulted in decreased IGFs in the extrastriate cortex [48]. In contrast, a negative correlation of gamma peak frequency with the occipital cortex surface area and pericalcarine volume was demonstrated [23]. Although peak frequencies within the gamma range seem to be unrelated to the cortical thickness of the primary visual cortex [22,23,46,49], a positive relationship was shown for the pericalcarine [50,51], the cuneus [47] or the whole occipital cortex [23]. In addition, higher IGF in the presence of higher white matter density within the corpus callosum [52] and gray matter thickness in the occipital cortex [23] were reported. For the somatosensory gamma activity, the thickness of the primary somatosensory cortex was inversely related to peak gamma frequency [26]. On the contrary, some studies failed to detect correlations between IGF and various cortical measures (e.g., [22,46,49]); however, it should be noted that these experiments were performed in relatively small samples (12-34 subjects) and could have been underpowered to detect relevant associations. Abbreviations: EEG-electroencephalogram; fMRI-functional magnetic resonance imaging; IGF-individual gamma frequency; LFP-local field potential; MEG-magnetoencephalogram; MRS-magnetic resonance spectroscopy; ND-never depressed; RD-remitted depression.

Neurochemical Balance
Another endogenous aspect that was shown to be related to peak gamma frequency is the balance of neurotransmitters. The generation of gamma oscillations is based on the balance of excitation and inhibition [53,54]. Rapid adjustments in the inhibition result in changes in the oscillation interval and in the frequency of oscillation [55,56]. Consequently, most studies addressed the association between the concentration of GABA and peak gamma frequency (Table 2). However, the results are not conclusive as both positive relationship between IGF and GABA concentration levels [29,[57][58][59] and no relationship [22,24,51,60] were shown. Again, it is possible that small sample sizes (14-50 subjects) could have hindered the ability to reveal potential relationships. To note, a positive association between GABA A receptor density in the primary visual cortex and visually induced gamma frequency [61] and an inverse relationship between IGF and glutamate concentration in the lateral occipital cortex [30] were also observed. However, other studies did not find a significant correlation between IGF and glutamate [24,60].  Abbreviations: ACPA-arachidonylcyclopropylamide; DHPG-(RS)-3,5-dihydroxyphenylglycine; DLPFCdorsolateral prefrontal cortex; EEG-electroencephalogram; fMRI-functional magnetic resonance imaging; IGF-individual gamma frequency; LFP-local field potential; MEG-magnetoencephalogram; MRI-magnetic resonance imaging; MRS-magnetic resonance spectroscopy; PET-positron emission tomography.

Neuropsychiatric Disorders
It is well known that both neuroanatomical changes and neurotransmitter imbalances are linked with various pathological conditions, e.g., schizophrenia [69], depression [70] and neurodegenerative disorders [71]. Hence, peak gamma frequencies have been shown to be altered in neuropsychiatric disorders (Table 3). A few studies reported that patients with schizophrenia display lower IGFs compared to controls [9,29]. In addition, lower peak gamma frequencies in Alzheimer's disease [21] and dyslexia [5] patients compared to controls were reported. On the contrary, an increase in peak gamma frequency was observed in people with autistic traits and autism spectrum disorder [18,72]. Finally, no IGF difference was found between healthy controls and patients with photosensitive epilepsy [73], visual snow syndrome [74] or schizoaffective bipolar disorder [75]. A relationship between phenomenological symptoms and cognitive abilities was also demonstrated: Subjects with schizophrenia reporting higher self-disorder scores had lower parietal peak gamma frequencies in response to proprioceptive stimulation [76], similar to multiple sclerosis patients with cognitive impairment who showed lower IGFs compared to patients with retained cognitive functions and controls [77].

Brain States
Peak frequencies in the gamma range were shown to depend on the subjects' general state (Table 4). For example, peak gamma frequencies detected in response to auditory stimulation were shown to decrease with higher doses of isoflurane [78] and propofol [79] sedation. Similarly, lower IGF was shown in anesthetized monkeys compared to their awake state [80]. In contrast, Saxena et al. [81] found no differences in visually induced IGF between awake and propofol-induced sedation states. Lozano-Montes et al. [34] demonstrated that peak gamma frequency was higher in rats during quiet wakefulness than during self-grooming behaviors. Moreover, peak gamma frequencies were shown to depend on the state of the female organism: During the luteal phase, peak gamma frequencies were significantly higher compared to the peak frequencies in the follicular phase during the menstrual cycle [82].
Finally, as neuromodulational techniques are increasingly used to change the state of the brain, the effect of neuromodulation on peak gamma frequencies was tested. However, the knowledge is still limited: Transcranial alternating current stimulation (tACS) was reported to increase IGF [4,83,84], whereas transcranial direct current stimulation (tDCS) did not modulate IGF [85][86][87]. In addition, non-invasive vagal nerve stimulation was shown to decrease IGF [88]. The abovementioned suggests the potential of gamma peak frequencies to be modulated.

Age
Lastly, age is regarded as one of the factors modulating peak gamma frequency (Table 5). Significant reductions in gamma peak frequencies were observed in the medial frontal cortex of aged rats, but not in the hippocampus [37]. In humans, the slowing of IGF in the primary visual cortex [22,23,47,50,[89][90][91][92][93][94] and primary motor cortex [59] with age was observed. Similarly, Purcell et al. [31] demonstrated an age-related decrease in peak gamma frequencies obtained from the auditory envelope following responses within the 30-50 Hz range. While many studies report a negative relationship between IGF and age, Duygun et al. [20] showed that peak gamma frequency tends to increase from childhood to young adulthood; however, it decreases with age in the adult population. In contrast, Poulsen et al. [95,96] reported an increase in IGF with age, whereas a few studies [24,46,51] showed no significant relationship between gamma peak frequencies and age, potentially due to testing subjects of relatively narrow age ranges (see Table 5). However, Proskovec et al. [26] also failed to detect a significant link between age and IGF, despite including subjects of a wide range of ages. Abbreviations: AD-Alzheimer's disease; ASD-autism spectrum disorder; EEG-electroencephalogram; HC-healthy children; HE-healthy elderly; HY-healthy adults; IGF-individual gamma frequency; MEGmagnetoencephalogram; MRI-magnetic resonance imaging; MRS-magnetic resonance spectroscopy; NDnever depressed; RD-remitted depression; rsEEG-resting-state electroencephalogram.

Peak Gamma Activity in Response to Sensory Stimulation
Gamma activity is generated in response to a stimulus of any modality [11][12][13]. Accumulating evidence suggests that various sensory processes have an effect on IGF.

Visual Processing
So far, research on peak frequency is mostly focused on visually evoked and induced gamma responses. It seems that depending on the type of a visual stimulus, IGF can be affected differently (Table 6). Higher IGFs are obtained with increasing stimulus contrast [23,33,36,73,[97][98][99][100][101], velocity [23,91,92,[102][103][104] and in the presence of orthogonal masks [99]. An increase in IGF was also observed for preferred stimulus orientation [103] and grated stimulus discontinuity [104] in monkeys. In contrast, peak gamma frequencies are inversely related to the size of a visual stimulus [19,32,98,105] and grating [32]. Yet, some studies showed that the size of a visual stimulus has no effect on peak gamma frequency [44,59]. Several studies provided contradictory results regarding visual stimulus eccentricity: Lima et al. [33] and van Pelt and Fries [19] reported an inverse relationship between eccentricity and IGF, while Gregory et al. [44] observed an increase in IGF in the presence of higher eccentricity. A biphasic pattern of gamma peak frequency modulation in the early visual cortex related to the repetition of the same visual stimulus was demonstrated in both humans [106] and monkeys [107]: IGF remained unchanged in humans and decreased in monkeys during the initial repetitions of a stimulus, but further repetitions resulted in a steady increase in IGF, possibly reflecting adaptive neuronal processing of recurring stimuli. In general, some studies suggest that individual system properties as well as the moment of assessment might affect the results of IGF estimation. For example, Kahlbrock et al. [89] reported higher IGF in individuals with higher critical flicker frequency, whereas Brunet et al. [19] observed a significantly decreased peak gamma frequency immediately after saccades in monkeys.  Abbreviations: ASD-autism spectrum disorder; EEG-electroencephalogram; fMRI-functional magnetic resonance imaging; IGF-individual gamma frequency; LFP-local field potential; MEG-magnetoencephalogram; MRI-magnetic resonance imaging.

Auditory Processing
Research in the auditory domain has demonstrated a relationship between IGF and auditory perception (Table 7). Multiple studies showed that higher IGFs are related to better performance in the gap detection task [31,41,83], which is commonly used as a measure of temporal auditory acuity. Moreover, peak gamma frequency was significantly related to the maximum perceptible modulation frequency [31]. In a study assessing gamma-band activity in musically trained children, IGF was shown to be higher over frontal areas, compared to children without musical training [109]. Abbreviations: EEG-electroencephalogram; IGF-individual gamma frequency; tACS-transcranial alternating current stimulation.

Somatosensory Processing
Recent studies demonstrated the association of IGF with certain aspects of somatosensory processing (Table 8). Spooner et al. [110] reported a significant increase in peak gamma frequency for the second somatosensory stimulus compared to the first one of the pair, pointing to IGF involvement in sensory gating. Interestingly, the elevation was even higher in HIV-infected adults [110]. Similarly, Cheng et al. [111] observed higher IGF for subjects showing more pronounced sensory gating responses in the early P35m component of somatosensory-evoked magnetic fields. On the contrary, in paroxysmal kinesigenic dyskinesia disorder, lower peak gamma frequencies of somatosensory-evoked potentials were discovered compared to healthy subjects [25]. Abbreviations: IGF-individual gamma frequency; MEG-magnetoencephalogram; PKD-paroxysmal kinesigenic dyskinesia.

Peak Gamma Frequency in Motion and Cognitive Processes
The activity within the gamma range is known to be implicated in many processes, including motor commands [112] and cognitive processing [14].

Motor Activity
High-frequency gamma oscillations are involved in movement execution [27]. Some evidence suggests that each individual has a profile of gamma peaks for specific motor outputs that is consistent over time [113]. However, peak gamma frequencies in the primary motor cortex differ depending on the effector: For example, the peak gamma frequency during foot dorsiflexion is significantly lower compared to the peak frequency during elbow flexion [27]. Muthukumaraswamy [28] demonstrated that IGF in the primary motor cortex is higher during the first movement of a sequence in comparison to the peak frequency during the following movements. Zheng [35] showed that peak gamma frequency increased in rats as their speed of running increased. In subjects with paroxysmal kinesigenic dyskinesia, Liu et al. [25] demonstrated a substantially reduced peak frequency compared with healthy controls. Importantly, Heinrichs-Graham et al. [114] showed that the peak frequency of the movement-related gamma synchronization can be modulated by visual distractors: IGF was significantly higher during incongruent conditions than the congruent ones. More details are provided in Table 9. Abbreviations: IGF-individual gamma frequency; LFP-local field potential; MEG-magnetoencephalogram; MRI-magnetic resonance imaging; PKD-paroxysmal kinesigenic dyskinesia.

Cognitive Processes
Kucewicz et al. [115] proposed that there are two distinct gamma oscillatory activity types involved in cognitive processing that can be categorized into narrowband and broadband gamma activities, with the narrowband gamma activity being centered on a specific gamma frequency peak. Indeed, most studies that estimated IGF during cognitive task performance found it to be within the lower gamma range (30-60 Hz) (Table 10). Research in monkeys demonstrated that spatial attention increases peak frequency [36] which also differs depending on where attention is directed: Higher peak frequency was observed when attention was directed to the background, compared to when the foreground was attended [33]. Bosman et al. [116] showed an increase in IGF in the primary visual cortex of monkeys when attention was directed to behaviorally relevant stimuli, compared to irrelevant stimuli. However, human studies showed no evidence of peak gamma frequency modulation in the primary visual cortex due to spatial attention [117] or monitoring of moving stimuli [118]. The alternation rate for bistable images in perceptual rivalry tasks was negatively related to IGF in the primary visual cortex [119]. Chen et al. [29] showed a significant positive correlation between peak gamma frequency and working memory performance. In line with this, some studies reported a positive relationship between IGFs and performance in memory tasks during different stages of sedation with either propofol [79] or isoflurane [78]. Furthermore, patients diagnosed with multiple sclerosis whose gamma peak frequencies were slower tended to perform worse in the neuropsychological tests of verbal memory, attention and executive functions [77]. Supporting results were reported by Insel et al. [37] in rats where IGF negatively correlated with the median decision times. Yet, when the control of age was added, the relationship was non-significant. Finally, the relationship of peak gamma frequencies with language processing was observed by Rufener and Zaehle [5] who found a positive correlation between IGF and phonological awareness in dyslexic patients. Nevertheless, it should be mentioned that a few studies did not observe any relationship between estimated IGFs and performance in complex planning tasks [6] or inhibition interference [7].

Summary
In the present work, we overviewed the available findings concerning the relationships of IGF with various factors. Even though there is a relatively large number of studies that assessed IGF, only a small fraction considered IGF as a primary measure of interest. The studies included in this review showed that IGFs span a wide range of frequencies (~30-100 Hz). In addition to the inter-individual differences, the wide range of estimated IGFs could be partly due to the variability of methods used to extract these measures. For convenience, the included studies are summarized in Table 11, wherein the number of studies for each IGF estimation modality, task and factor related to IGF is provided. The majority of studies used visual stimulation for IGF estimation, far fewer utilized auditory stimulation, while only a few studies employed somatosensory stimulation or motor and cognitive tasks. However, it is evident that factors related to IGF vary considerably among different stimulation types used for IGF extraction. For example, IGF estimated from the responses to visual stimulation seems to be mainly related to endogenous factors, such as anatomical and neurochemical measures, as well as visual processing, while IGF from the auditory modality was also shown to be linked with cognitive processes. The abovementioned suggests that IGF estimated from different modalities may reflect different functional aspects, being reflective of the activity in the underlying sources. For example, auditory steady-state stimulation results in the activation of both auditory cortices and the frontal region [6,8] potentially being related to gamma implicated in cognitive processing [120]. Conversely, visually evoked or induced gamma activity is generated in the occipital regions [47,50,51] where visual information is processed. Research employing different sensory modalities to estimate IGF and covering several different functional aspects in the same population is required to establish more robust IGF relationships with various processes. In addition, future studies should address the need for the optimization of IGF extraction methods, which could help to estimate the IGF range and correlates more precisely. In general, the present review suggests that IGF seems to reflect certain properties of brain functioning; thus, future studies could potentially provide the basis for the use of IGF as an electrophysiological biomarker. Table 11. A summary of selected human and animal studies. The number of studies was counted for each IGF estimation modality, specific tasks within the modalities and the factors related to IGF. In addition, the IGF range within each modality is presented. Factors related to IGF include only those studies that found a statistically significant relationship.  Abbreviation: IGF-individual gamma frequency. * Wyss et al. [24] reported the IGF range extracted from auditory stimulation to be between 30 and 160 Hz; however, in the other studies, the highest IGF values were up to around 63 Hz.

Conflicts of Interest:
The authors declare no conflict of interest.