Part 1:Independent Dynamics Of Low, Intermediate, And High Frequency Spectral Intracranial EEG Activities During Human Memory Formation
Mar 18, 2022
Contact: Audrey Hu audrey.hu@wecistanche.com
Victoria S. Marksa, Krishnakant V. Saboob, Çağdaş Topçuc,d,e, Michal Lechc,d,e,
Theodore P. Thayibf, Petr Nejedlye,g, Vaclav Kremen,h, Gregory A. Worrell, I,
Michal T. Kucewiczc,e,i,∗
a Graduate School of Biomedical Sciences, Mayo Clinic, USA
bDepartment of Electrical and Computer Engineering, University of Illinois, Urbana-Champaign, IL, USA
c Multimedia Systems Department, Faculty of Electronics, Telecommunications and Informatics, BioTechMed Center, Gdansk University of Technology, Gdansk, Poland d Nencki Institute of Experimental Biology, Polish Academy of Sciences, Warsaw, Poland
department of Neurology, Mayo Clinic, Rochester, MN, USA
department of Computer Engineering, Iowa State University, Ames, Iowa, USA
g The Czech Academy of Sciences, Institute of Scientific Instruments, Brno, Czech Republic
h Robotics, and Cybernetics, Czech Institute of Informatics, Czech Technical University in Prague, Prague, Czech Republic
department of Physiology and Biomedical Engineering, Mayo Clinic, USA
a b s t r a c t
A wide spectrum of brain rhythms is engaged throughout the human cortex in cognitive functions. How the rhythms of various frequency ranges are coordinated across the space of the human cortex and time of memory processing is inconclusive. They can either be coordinated together across the frequency spectrum at the same cortical site and time or induced independently in particular bands. We used a large dataset of human intracranial electroencephalography (EEG) to parse the spatiotemporal dynamics of spectral activities induced during the formation of verbal memories. Encoding of words for subsequent free recall activated low-frequency theta, in- intermediate frequency alpha and beta, and high-frequency gamma power in a mosaic pattern of discrete cortical sites. A majority of the cortical sites recorded activity in only one of these frequencies, except for the visual cortex where spectral power was induced across multiple bands. Each frequency band showed characteristic dynamics of the induced power specific to the cortical area and hemisphere. The power of the low, intermediate, and high-frequency activities propagated in independent sequences across the visual, temporal and prefrontal cortical areas throughout subsequent phases of memory encoding. Our results provide a holistic, simplified model of the spectral activities engaged in the formation of human memory, suggesting an anatomically and temporally distributed mosaic of coordinated brain rhythms.
1. Introduction
Brain rhythms are thought to support memory and cognitive functions by coordinating activities of connected neurons into synchronous interactions (Singer, 1999; Buzsaki, 2006; Fries, 2015; Klimesch, 1996; Fell and Axmacher, 2011). The spectrum of rhythmic activities involved in these interactions includes the theta (3–9 Hz) and the gamma
(30–120 Hz) frequency bands with specific roles proposed for the low and high-frequency activities in-memory processing (Düzel et al., 2010; Sauseng et al., 2010; Fries et al., 2007; Tallon-Baudry and Bertrand, 1999; Nyhus and Curran, 2010; Lisman and Jensen, 2013; Herweg et al., 2020). The intermediate frequency alpha (9–12 Hz) and beta (12–25 Hz) bands were also associated with processes necessary for memory functions (Klimesch et al., 2007; Hanslmayr et al., 2011; Spitzer and Haegens 2017; Schmidt et al., 2019; Michalareas et al.,2016). How these various rhythms are operating together in the function at a defined timescale, engaging only a subset of the spectral activities in a given frequency range. For example, only theta rhythms would be indued at a given site. Both scenarios agree that the brain ac- activities at defined frequency ranges can be conceptualized as “spectral fingerprints” that subserve specific perceptual or cognitive processes at a corresponding temporal scale of neural interactions (Siegel et al., 2012; Hanslmayr and Staudigl, 2014). Resolving the dynamics of these spectral fingerprints accurately in the anatomical space and time is critical to test these hypotheses and provide a holistic model of the neural activities engaged during cognitive tasks.
Intracranial electroencephalographic (EEG) recordings present a unique opportunity to probe with high resolution the large scale of neural activities engaged during cognitive and other brain functions in discrete areas of the human cortex (Lhatoo et al., 2019; Johnson et al., 2020; Jacobs and Kahana 2010; Engel et al., 2005). The EEG signals are sampled from electrode contacts implanted either directly on the surface of the neocortex or into its deeper layers and subcortical structures. Thus, spectral activities generated by local neural populations and recorded from multiple separate contacts can be tracked in discrete cortical areas during the time of memory processing. Previous iEEG studies commonly employed the spectral power induced in high gamma frequency ranges (60–120 Hz) to track cortical processing in memory and cognitive tasks (Jerbi et al., 2009; Lachaux et al., 2012; Crone et al., 2006; Lundqvist et al., 2018). Theta (Sheehan et al., 2018; Lin et al., 2017) and alpha (Staresina et al., 2016) were also investigated previously. Oscillatory activities in the theta and alpha frequencies were shown to propagate as local, independent, traveling waves in discrete cortical areas during memory processing(Zhang et al., 2018), but the relative spatiotemporal dynamics of these low-frequency oscillations and the faster beta and gamma activities during memory processing remain largely unexplored.
Here, we took advantage of a large iEEG dataset of 164 participants encoding lists of words for subsequent free verbal recall. Previous studies with this task revealed a distributed network of brain regions associated with sensory processing and with higher-order declarative memory functions (Burke et al., 2013; Burke et al., 2014a; Kucewicz et al., 2019). The induced high gamma power was observed first in the visual areas of the occipitotemporal cortex and then in more anterior areas of the prefrontal cortex, inspiring a two-stage model of memory encoding with an early sensory and a late association stage(Burke et al., 2014a).
A similar sequence of posterior-to-anterior order of induced power was also observed in the theta frequency band (Burke et al., 2013) and con- firmed in activities beyond the gamma frequency range (Kucewicz et al., 2014). Our recent study found the sequence of the induced high gamma power during memory encoding to be continuous with increasing latencies along a hierarchy of gradually higher-order association areas (Kucewicz et al., 2019), culminating in the anterior prefrontal cortex late into word encoding. The previous iEEG studies of verbal memory processing were either limited to smaller datasets of participant recordings with sparse electrode coverage of the neocortex or focused only on selected frequency bands. Hence, spectral power across various bands would either be averaged spatially across multiple electrode sites of larger cortical regions, thereby losing the resolution and granularity of individual sites, or would not be studied across the theta, alpha, and beta, and gamma brain activities.
Given the wide scope of the frequency spectrum and brain areas analyzed collectively in the previous studies, we hypothesized that specific spectral activities are induced in a granular pattern of discrete cortical
locations. Hence, aggregating activity from several locations would give rise to a broadband tilt in power across the entire spectrum (Kilner et al.,2005), which could be composed of multiple, specific, "spectral fingerprints" (Fellner et al., 2019). The exact localization of different fingerprints would either be common to the same cortical sites of electrode recording or would be distributed in a mosaic of distinct anatomical sites. Likewise, in terms of timing, the low, intermediate, and high-frequency spectral activities would be induced at different times of memory encoding or all at the same time. In addition to the theta and gamma activities, alpha and beta rhythms were expected to be induced at distinct cortical sites and times of memory encoding. In general, we tested for independent spatiotemporal dynamics of the spectral activities induced during the formation of human verbal memory traces. Our goal was to provide a holistic model of brain rhythms engaged across a broad range of frequencies, cortical anatomy, and phases of memory encoding.
2. Materials and methods
2.1. Subjects
The dataset of EEG recordings from a total of 164 participants undergoing surgical evaluation of drug-resistant epilepsy was taken from a large multicenter collaborative study (all de-identified data are available at http://memory.psych.upenn.edu/Electrophysiological_Data).
The recordings were collected from the following centers: Mayo Clinic, Thomas Jefferson University Hospital, Hospital of the University of Pennsylvania, Dartmouth-Hitchcock Medical Center, Emory University Hospital, University of Texas Southwestern Medical Center, and Columbia University Hospital. One common research protocol was approved by the respective Institutional Review Board at each center, and informed consent was obtained from each participant. For the 139 patients included in the final analysis, information regarding the number of word trials complete and seizure onset zone are summarized in Suppl. Table 1. Seizure onset was localized to the right hemisphere in 46 patients, the left hemisphere in 55, bilaterally in 26, and undetermined in 12. Demographic information such as age and sex were only availed- able for 85 patients (mean age 35.6 years with standard deviation 11.6 years; 43 Females to 42 Males, Suppl. Table 1) due to limited access to the sensitive healthcare information from multiple clinical centers that collected the aggregate dataset. Electrophysiological recordings were collected from standard clinical subdural and depth electrodes (AdTech Inc., PMT Inc.) implanted on the cortical surface and into the brain parenchyma, respectively. Subdural electrode contacts were arranged either in a grid or a strip configuration with 10 mm separation, and depth electrode contacts were separated by 5 to 10 mm. The placement of electrodes was determined by a clinical team to localize seizure foci for possible epilepsy respective surgery or implantation of a device for therapeutic electrical brain stimulation.
2.2. Anatomic localization and brain surface mapping
Cortical surface parcellations were generated for each participant from pre-implant MRI scans (volumetric T1-weighted sequences)- ing Freesurfer software (RRID: SCR_001847). The hippocampus and sur- rounding cortical regions were delineated separately based on an additional 2 mm thick coronal T2-weighted scan using the Automatic Segmentation of Hippocampal Subfields (ASHS) multi-atlas segmentation method. Electrode contact coordinates derived from co-registered post-implant CT scans were then mapped to the pre-implant MRI scans to determine their anatomic locations. For subdural strips and grids, the electrode contacts were additionally projected to the cortical surface using an energy minimization algorithm to account for postoperative brain shift. Contact locations were reviewed and confirmed on surfaces and cross-sectional images by a neuroradiologist. The T1-weighted MRI scans were also registered to the MNI152 standard brain to enable the comparison of recording sites in a common space across subjects. Anatomic locations of the recording sites, including Brodmann areas, were derived by converting MNI coordinates to Talairach space and querying the Ta- Talairach daemon (www.talairach.org).
2.3. Electrophysiological recordings
The EEG signals were recorded using one of the following clinical electrophysiological acquisition systems (dependent on the institution for data collection): Nihon Kohden EEG-1200, Natus XLTek EMU 128, or Grass Aura-LTM64. Depending on the acquisition system and the preference of the clinical team, the signals were sampled at either 500, 1000, or 1600 Hz and were referenced to a common contact placed either in- intracranially, on the scalp, or the mastoid process. For analysis, all recordings using higher sampling rates were filtered by an antialiasing filter and down-sampled to 500 Hz. A common bipolar montage was calculated post hoc for each subject by subtracting measured voltage time series on all pairs of spatially adjacent contacts. This resulted in N-1 bipolar signals in the case of the penetrating and the strip electrodes, and N = (i − 1)∗j + (j − 1)∗I, where i and j are the numbers of contacts in the vertical and horizontal dimensions of the grid. For the data analysis of this study, one “electrode” refers to the bipolar signal from one bipolar pair of contacts. It is important to note that contacts may be used for multiple electrodes, so not all signals are truly independent.
2.4. Free recall task
A classic paradigm for probing formation of verbal episodic memory was employed (Kahana, 2014), in which subjects were shown lists of words for subsequent free recall. Participants were instructed to study lists of words presented on a laptop computer screen for a delayed test of vocal recall. Lists were composed of 12 words chosen at random and without replacement from a pool of high-frequency nouns in the subject’s native language (either English or Spanish; http://memory.psych.upenn.edu/WordPools). Each word appeared on the screen for 1600 ms, followed by a random jitter of 750 to 1000 ms blank interval between stimuli. At the end of each word list, a distractor task was performed by the subject. This task lasted for 20 s and was composed of a series of simple, arithmetic problems of the format A + B + C, where A, B, and C were random, single-digit integers between 1 and 9. After the distractor task, participants were given 30 s in which to recall as many of the words from the list as possible in any order (Fig. 1). Vocal responses were recorded digitally by the laptop and were later scored manually for analysis. Only subjects who recalled at least 15% of words and completed at least 12 lists of the task were included in further analysis (Long et al., 2014). This left 139 of the 164 original subjects for a total of 14,219 electrodes used in this study. Electrolysis-logical recordings were synchronized to stimulus appearance on the screen through an electric pulse generator operated by the task laptop, which sent pulses to a designated event channel in the clinical acquisition system. The events were timestamped after the recording session using custom-written MATLAB codes and were used to extract specific epochs of interest around word presentation. Each recorded epoch was 3000 ms long and included 1600 ms of word presentation on the screen with 700 ms of a blank screen of the interstimulus interval before and after each word presentation.
Trial-averaged spectral power of EEG signals from epochs of word presentation (one word per trial) was integrated across the epoch and used as a feature to classify active electrodes that record from brain regions engaged in verbal memory encoding. Active electrodes were clas- sified independently in six frequency bands of the EEG spectrum using normalized estimates of the power change (z-score transform). The clas- sification procedure used an unsupervised method based on the Gaussian Mixture Model (Saboo et al., 2019). Notice induced activity across all six frequency bands in the example spectrogram plotted from trial- averaged activity from one electrode localized in the occipital cortex.
2.5. Data analysis
We analyzed the iEEG recording epochs during word presentation for memory encoding. Each presentation epoch was FIR filtered (2000- order Barlett-Hanning with zero-phase distortion filtering, bandpass with frequency limits specified for each frequency band) before being spectrally decomposed, normalized, and binned independently into dis- tinct frequency bands between 2 and 120 Hz: low theta (2–4 Hz), high theta (5–9 Hz), alpha (10–15 Hz), beta (16–25 Hz), low gamma (25– 55 Hz), and high gamma (65–115 Hz). The boundaries for these frequency bands were determined based on non-overlapping and continuous ranges growing in width along the frequency spectrum. The upper limit of 115 Hz was chosen to ensure at least 4 samples per cycle given the sampling frequency of 500 Hz and to avoid the harmonic of the 60 Hz line noise. To convert filtered signals to a spectrogram with a
2 Hz frequency bin resolution from 2 Hz to 115 Hz, we used 500 ms sliding windows with 5 ms slide lengths and the multi-taper method with 1 taper from the chronic toolbox in MATLAB (http://chronux.org/; Mitra, 2007). Spectrograms of the word presentation epoch were avered- aged together after log-normalization and a z-score transformation of each time-frequency point. Z-scoring was performed according to the following formula:
x(t, f) − μ
z(t, f)=σ
where x is the signal, t is the time bin, f is the frequency bin, f is the mean, and f is the standard deviation for the given frequency. Each trial was normalized separately. Log normalization was implemented to correct for the 1/f-power law effect of lower frequencies on estimating power in the higher frequency bands, and the z-score transformation was performed to provide a normalized scale of power change for comparison of signals from different electrodes, sessions, and participants (Kucewicz et al., 2019, 2017; Saboo et al., 2019). This z-score normalization produces positive and negative power changes relative to the mean power within any one-word epoch instead of using a “baseline” period. Hence, even small positive or negative deviations of absolute power would effectively be augmented relative to the signal mean, as described before (Kucewicz et al., 2019; Alotaiby et al., 2015). Edge artifacts in the power estimate were eliminated by clipping two-time bins from either end of the final spectrogram for all frequencies. For each electrode and frequency band, we determined the average spectral power at each time bin across all word presentation epochs. The overall induced power in a given frequency band during the word presentation period was quantified as the area under the absolute value of the corresponding time-series curve, as shown inFig. 1. Electrodes were then clas- sified into “active” and “inactive” in each specific band using our unsupervised clustering method based on Gaussian Mixture Modeling (GMM) to identify those that showed task-induced power changes (Saboo et al.,2019). This method essentially performs a binary machine-learning classsification of active and inactive channels based on the induced power value of each channel. The classification was done separately for each frequency band of interest. The entire procedure is summarized inFig. 1.
We chose a GMM-based method to ensure that different numbers reflect- trades would be allowed for the clusters of active electrodes and the inactive electrodes. The number of clusters was set to two because the active-inactive categorization of electrodes is binary: change in the spectral power can either be induced (increased or decreased) or not during the word presentation period. The use of a GMM also enabled classicification without a requirement for ground truth data. Electrodes were pooled across subjects and then clustered to identify active electrodes.
2.6. Induced power mapping
Most of the active electrodes were localized in the cortical regions associated with processing visual and semantic information, as well as with declarative memory and executive functions. Out of 39 Brod- Mann Areas, the majority of electrode locations grouped into 9 brain re- regions each comprising Brodmann Areas: visual (Brodmann areas 18 and 19), inferior temporal (Brodmann areas 20 and 37), precuneus (Brod- main areas 30 and 31), lateral parietal (Brodmann areas 7 and 39), mesial temporal (Brodmann area 28 and Hippocampus), lateral temporal (Brodmann areas 21 and 22), Broca’s areas (Brodmann areas 44 and 45), lateral prefrontal (Brodmann areas 9 and 46), and frontal pole (Brodmann areas 10 and 11). Unless stated otherwise, all further analysis was focused on the recalled word epochs, i.e. epochs with successfully encoded words that were subsequently freely recalled, in these nine cortical regions.
Induced power changes in each frequency band were plotted from all active electrodes onto an average brain surface for each time point of the word epoch using a custom MATLAB code for an electrode-location-dependent weighing technique. The brain surface was created as a mesh, where each element of the mesh was assigned a color dependent on the induced power of the surrounding electrodes weighed by the relative distance to the element (greater weight to more proximal electrodes). For each mesh element of the brain surface plotted, only the effects of electrodes within a threshold of 10 mm from the element center point were considered. Hence, the induced power was averaged from all electrodes within the threshold radius that were linearly weighted by their proximity according to the following formula:
IP1∗ (threshold− r1) + IP2∗ (threshold− r2) + …+ IP ∗ (threshold− rn )
(threshold− r1 )+ (threshold− r2 )+ …+(threshold− rn )
where IP is the induced power change value for an electrode, and r is the distance between the electrode and the chosen element of the brain surface mesh.
2.7. Statistics
ANOVA tests with multiple comparison corrections for the effects of the frequency band, brain region (Brodmann Area), and hemisphere were used to analyze the percentage of electrodes in an area labeled as “ac- time (Fig. 2). Temporal differences between hemispheres were quantified using a two-tailed, unpaired t-test statistic with Bonferroni correction (Fig.3A). We then broke each word epoch into pre-encoding, early encoding, late-encoding, and post-encoding phases (Burke et al., 2014; Kucewicz et al., 2019) and used repeated measures ANOVA (encoding phase as the within-subjects factor) with posthoc Tukey-Kramer test to analyze the effects of the frequency band, brain region, and hemisphere on induced power during each phase of the verbal memory task (Fig.3B). Peak latency for each frequency band and area was determined as the time bin with the maximum positive peak of the induced power. Latin- cities were compared using repeated measures ANOVA (encoding phase as the within-subjects factor) and post-hoc Tukey-Kramer multiple comparison tests (Fig. 4). Pearson’s correlation test was used to assess the effect of position along each axis of the brain on the time of peak power latency (Fig.5B). The alpha value for all tests was set at 0.05.
3. Results
3.1. Spectral activities induced during memory encoding are heterogeneously distributed across the human cortex
We first identified a subset of electrodes from cortical areas activated during memory encoding. For this purpose, spectral power-in-band in six non-overlapping frequency ranges (low theta: 2–4 Hz, high theta: 5–9 Hz, alpha: 10–15 Hz, beta: 16–25 Hz, low gamma: 25–55 Hz, high gamma: 65–115 Hz) was used as features for automated classification of active electrodes (Saboo et al., 2019) independently in each band (Fig. 1). Out of the total of 14,219 electrodes implanted across all participants, 4738 (33.3%) showed the induced spectral power in at least one of the frequency bands. Hence, a typical electrode site showed induced power in a defined range of either low, intermediate, or high-frequency bands (Fig.2A). In agreement with our previous results (Kucewicz et al., 2019; Alotaiby et al., 2015), these active electrodes were widely distributed across 39 Brodmann areas that were defined in all cortical lobes (Fig.2B). The relative proportion of electrodes in each Brodmann area that were actively varied from 4.3% to 75%, depending on location within the limbic, frontal, prefrontal, parietal, temporal, or occipital lobes (ANOVA F = 99.21, p < 0.001, df = 5), with significantly more active electrodes found in the sensory visual areas of the occipital and the parietal cortex (posthoc Tukey Kramer, p < 0.05). Within each area, similar proportions were observed in each frequency band, including the alpha and the beta band, with no significant differences across the spectrum (N-way ANOVA F = 0.17, p = 0.9747, df = 5). In summary, regardless of the frequency band analyzed, at least 4% of electrode sites within any of the cortical areas analyzed revealed significantly induced power during the task.
To investigate whether these various low and high-frequency spectral activities were induced on the same electrodes, we summarized the over-lap across the six frequency bands for nine cortical regions of interest (ROI; each composed of two Brodmann areas with >18 active electrodes from >8 participants; see table 1). An electrode can be active in one or more bands. We found that the highest proportion of electrodes was active in only one (43.12%) or two (22.94%) bands, and proportionally- ally fewer in more than two bands with only a minority (5.4%) active in all six (Fig. 2C). The highest overlap of induced spectral power in more than two bands was observed for the electrodes in the posterior sensory visual areas of the occipital cortex (Fig. 2C; red), where 4–5 bands were activated on average per electrode during word encoding. Other visual areas of the temporal and parietal cortex also showed relatively high distributions of electrodes with more than two bands activated, in contrast to the anterior higher-order processing areas in the temporal and prefrontal cortex, where electrodes were active in 1–2 frequency bands (Fig. 2C; black). Overall, the majority of electrode sites showed the in- reduced spectral power in one or two specific frequency bands, suggesting that the various activities were spatially segregated in the cortex.
Fig. 2. Spectral activities induced during verbal memory encoding are non-uniformly distributed across the human brain. (A) Representative examples of trial-averaged spectrograms during word encoding from active electrodes show the induced power changes confined independently to each of the three frequency groups from different anatomical locations. (B) Proportions of active electrodes in six frequency bands studied (low theta: 2–4 Hz, high theta: 5–9 Hz, alpha: 10–15 Hz) beta:
16–25 Hz, low gamma: 25–55 Hz, high gamma: 65–115 Hz) show significant differences (ANOVA) in the anatomical distribution (red asterisks; p < 0.05 post-hoc Tukey Kramer test) but not across the frequency bands as summarized in the inset box plots. (C) Distribution of the electrodes active in one or more of the frequency bands reveals the highest overlap in the posterior areas of the occipital and parietal visual cortex, which was gradually decreasing in the more anterior cortical areas as shown in the average brain surface plot and the cartoon summary of all active electrodes (dots colored according to the bar chart) and in the violin plots (black and red lines indicate the mean and median, resp. (Hoffmann,2015)). (D) Distribution of all electrodes active in the gamma (high frequency), alpha/beta (intermediate frequency), and theta (low frequency) bands differs across nine selected cortical regions of interest (V - visual; IT - inferior temporal; Pre - precuneus; Par - lateral parietal; MTL - mesial temporal lobe; LT - lateral temporal; Br - Broca’s area; PFC - prefrontal cortex; FP - frontal pole) with relatively more prefrontal electrodes active in the gamma bands (red) and more electrodes in the visual areas active across all the bands (black) as shown in the average brain surface plot, the cartoon summary, and the violin plot. Notice that most electrodes overall were active in just one or two frequency bands either of the low, intermediate, or high frequencies, except for the visual areas with more than two frequency bands active per anyone electrode (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.).
We next asked whether the spectral activities that were induced at any one electrode site overlapped in similar frequency ranges. We grouped the activities into low, intermediate, and high-frequency groups (theta, alpha/beta, and gamma, respectively) and compared proportions of active electrodes with different group combinations. More than half of all active electrodes showed induced power is exclusively in the low, intermediate, or high-frequency bands (Fig.2D). Electrodes with high overlap across all groups were predominantly localized in the posterior visual areas (Fig.2D; black). The remaining electrode sites that showed induced power in one of the three frequency groups were uniformly distributed across the cortex with similar pro- portions in each ROI, apart from the prefrontal areas where relatively more high-frequency gamma activities were present during memory encoding (Fig.2D; red). Our results showed that activities of a particular frequency range were induced mostly at specific cortical sites, forming a mosaic-like pattern of distribution consistent with the notion of "spectral fingerprints" (Siegel et al., 2012).
3.2. Laterality effect of the induced power is specific to particular frequencies and cortical areas
Our next question was if the total power-in-band induced during the time of word encoding (Fig. 1) was different depending on the anatomical location, the frequency band, or the hemisphere. Individual active electrodes can show similar magnitudes of the induced power, despite the heterogeneous distribution across the cortical regions (Fig. 2). Overall, there was a significant effect of the frequency band on the induced power (repeated measures ANOVA F = 1346.99,p < 0.001, df = 5). Brain region (ANOVA F = 38.21, p < 0.001, df = 37) and of the hemisphere location of the electrode (ANOVA F = 17.4, p < 0.001, df = 1). Post-hoc
Fig. 3. Temporal profiles of induced power during encoding re-veal laterality effects in specific frequencies and cortical areas. (A) Each panel shows trial-averaged power changes across the time of word presentation (gray background) estimated from all active electrodes in a given frequency band and brain region plotted separately for the left and right hemisphere (red and blue, resp.). The gray area plots in the background summarize the statistical difference (t-statistic) between the two hemispheres in 50 ms time bins with dark gray indicating significantly more (positive t values) or less (negative t values) power in the left hemisphere (horizontal dashed lines; Student t-test, p < 0.05). Notice that the significant laterality effect is localized in specific frequency bands in a given brain region. (B) Summary comparison of the total induced power(repeated measures ANOVA) from all frequency bands across the nine brain regions (see Fig. 2 for labels) and four stages of memory encoding – the within-subjects factor (red asterisks indicate significant difference; Tukey-Kramer test, p < 0.05). (C) Memory effect was the largest in the left prefrontal cortex (indicated by the black double arrow), whereas, MTL showed the largest effect in the right hemisphere areas (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.).
group analysis (Tukey-Kramer, p < 0.05) confirmed significantly smaller power with increasing frequencies of the bands, except between the low and high gamma bands (Supplementary Fig. 1). Electrodes within the occipital lobe showed significantly higher induced power than in any other area in agreement with our previous study of the high gamma activities (Kucewicz et al., 2019).
