Intracranial electrocorticographic correlates of intrinsic brain neetworks

Analyzing patterns of intracranial electroencephalographic (EEG) recordings can provide insight into how temporal and spatial components of brain activity are related on a trial-by-trial basis. Research on fMRI resting state networks has clarified the role of the default mode network (DMN) in internally directed cognition (e.g., mind-wandering), and the frontoparietal network (FPN) in externally directed cognition (e.g., working memory tasks); however, the relation of these networks to similar functional properties of neuroelectric activity is not yet understood. Recently, it has been suggested that the scalp EEG phenomena of alpha event-related desynchronization (ERD) and theta-band eventrelated synchronization (ERS) during simple cognitive tasks might represent neuroelectric parallels of DMN down-regulation and FPN activity, respectively. This study addressed this question by investigating task-related patterns of alpha ERD and theta ERS at intracranial sites within the DMN and FPN networks while participants completed multiple trials of the Multi-Source Interference Task (MSIT), containing three levels of task difficulty. The intracranial EEG patterns of 8 patients who underwent presurgical mapping for treatment of epilepsy were examined. Alpha ERD was expected to be preferentially observed at sites corresponding to the DMN while theta ERS was expected to be preferentially observed at FPN sites, dependent upon task difficulty. Additionally, higher task difficulty was expected to be related to slower reaction time. In support of these hypotheses, increased task difficulty was positively related to rection time. Contrary to these hypotheses, electrodes within the default mode and frontoparietal networks showed no significant differences in theta or alpha power; however, exploratory findings direct future work to investigate potential theta and alpha power differences within and between the frontoparietal, default mode, visual, and dorsal attention networks. Specific hypotheses for alpha and theta ERD/ERS within these networks are discussed. These results indicate that intracranial electroencephalography can be used to analyze the spatial and temporal patterns of neural networks. With replication, it is likely that increased task difficulty will elicit differing mechanisms of cognitive engagement within these four networks, supported by the found ERD/ERS trends of alpha and theta power, dependent upon task difficulty. These findings focus the exploration of cognitive information processing towards the dorsal attention and visual networks, strengthen the drive to understand the distinct classification of the frontoparietal and dorsal attention networks, and support new ways to study these networks using neuroelectric recordings.

All Rights Reserved ii ABSTRACT Analyzing patterns of intracranial electroencephalographic (EEG) recordings can provide insight into how temporal and spatial components of brain activity are related on a trial-by-trial basis. Research on fMRI resting state networks has clarified the role of the default mode network (DMN) in internally directed cognition (e.g., mind-wandering), and the frontoparietal network (FPN) in externally directed cognition (e.g., working memory tasks); however, the relation of these networks to similar functional properties of neuroelectric activity is not yet understood. Recently, it has been suggested that the scalp EEG phenomena of alpha event-related desynchronization (ERD) and theta-band eventrelated synchronization (ERS) during simple cognitive tasks might represent neuroelectric parallels of DMN down-regulation and FPN activity, respectively. This study addressed this question by investigating task-related patterns of alpha ERD and theta ERS at intracranial sites within the DMN and FPN networks while participants completed multiple trials of the Multi-Source Interference Task (MSIT), containing three levels of task difficulty. The intracranial EEG patterns of 8 patients who underwent presurgical mapping for treatment of epilepsy were examined. Alpha ERD was expected to be preferentially observed at sites corresponding to the DMN while theta ERS was expected to be preferentially observed at FPN sites, dependent upon task difficulty. Additionally, higher task difficulty was expected to be related to slower reaction time. In support of these hypotheses, increased task difficulty was positively related to rection time. Contrary to these hypotheses, electrodes within the default mode and frontoparietal networks showed no significant differences in theta or alpha power; however, exploratory findings direct iii future work to investigate potential theta and alpha power differences within and between the frontoparietal, default mode, visual, and dorsal attention networks. Specific hypotheses for alpha and theta ERD/ERS within these networks are discussed. These results indicate that intracranial electroencephalography can be used to analyze the spatial and temporal patterns of neural networks. With replication, it is likely that increased task difficulty will elicit differing mechanisms of cognitive engagement within these four networks, supported by the found ERD/ERS trends of alpha and theta power, dependent upon task difficulty. These findings focus the exploration of cognitive information processing towards the dorsal attention and visual networks, strengthen the drive to understand the distinct classification of the frontoparietal and dorsal attention networks, and support new ways to study these networks using neuroelectric recordings. iv TABLE OF CONTENTS ABSTRACT ii INTRODUCTION 1 METHODS 5 RESULTS 11 DISCUSSION 21 REFERENCES 31 v Brain activity was studied anatomically up until the early 2000s when the discovery of a baseline system known as the default mode network changed our understanding of cognitive functioning (Raichle et al., 2001). Since this discovery, recent research has shined a light on the role of many neural networks in relation to cognitive functioning and human behavior. Rather than thinking about the brain modularly (i.e., this part of the brain is responsible for this function), we are now starting to view brain activity more holistically. Neural activity is correlated less with individual neurons firing, and more with complexes of interconnected neurons that form what are now termed neural networks. This thesis will analyze the connection between neural networks and electrophysiological measures to better understand cognitive performance. Cognitive functioning and task-based performance have been linked to the activity of several neural networks (Anticevic et al., 2010). A parcellation provided by Yeo and colleagues (2011) indicates the following seven networks: visual, somatomotor, dorsal attention, ventral attention, limbic, frontoparietal, and default mode. Two networks relevant to the study of active information processing and working memory are the default mode network (DMN) and the frontoparietal network (FPN). Default Mode and Frontoparietal Networks The DMN is correlated with brain regions including the medial temporal lobe, the medial prefrontal cortex, and the posterior cingulate cortex (Yeo et al., 2011). This network is known to be involved in non-task specific activity, or more colloquially, mind wandering. Non-task specific activity, or mind wandering, refers to when the brain is not actively engaged in a specific task such as remembering a phone number, but rather left to daydream or engage in “shower thoughts”. Past literature has found a link between deactivation of the DMN and goal-driven behavior such as working memory (Anticevic et al., 2010). The temporal and spatial patterns of DMN activation and deactivation provide us with insight on what the brain is doing while not engaged in task-based behavior and how this connects to information processing. Deactivation of the DMN likely gives room for resources and energy to be utilized in other networks during taskdriven behavior for increased performance and processing. The FPN is associated with brain regions including the rostro- and dorsolateral prefrontal cortex, the anterior insula, the dorsal anterior cingulate cortex, and the anterior inferior parietal lobule (Yeo et al., 2011). In contrast to the DMN, the circuitry within these regions has been correlated with functions such as visuospatial analogical reasoning and active information processing (Watson & Chatterjee, 2012). The FPN activity is related to goal-driven behavior (Marek & Dosenbach, 2018), meaning that tasks involving more attention and active information processing, such as working memory tasks, will likely correspond with this network. The contrast between the FPN and DMN in neural network activation and associated task engagement could be relevant to mechanisms of cognitive performance. For this study, a Multi-Source Interference Task (MSIT) was used to elicit goaldriven behavior and FPN activation. The MSIT is a visuospatial working memory task that combines multiple dimensions of cognitive interference into three levels of difficulty, and it has been related to dorsal Anterior Cingulate Cortex (dACC) activation (Bush et al., 2003; Sheth et al., 2012), a brain module associated with the FPN. The connection between the MSIT and the dACC, a brain module studied in many lines of research 2 involving attention and cognitive processing, indicates that this task likely elicits goaldriven behavior and cognitive effort in requiring participants to accurately answer questions regarding a sequence of numbers despite modes of cognitive interference. Finally, acknowledging that Sheth and others (2012) found the dACC to be involved in behavioral adaptation (e.g., modifying behavior and cognitive processing mechanisms in response to changes in stimuli), it was predicted that task difficulty would act as a moderator of cognitive processing and show interaction effects with cognitive performance. Using electrophysiological measures of neural networks Electrophysiological processes have three possible dimensions of interest: frequency, time, and space. For example, an EEG signal in the theta-band could show a frequency level of around 5 Hz, a time component that peaks around 300-400 milliseconds following a stimulus, and a spatial characterization of belonging to an electrode located in the medial frontal region of the brain. Event-Related Synchronization (ERS) and Desynchronization (ERD) combine the time component with the frequency component to allow for non-phase-locked analyses (Pfurtscheller & Da Silva, 1999), meaning that we can see what is happening without locking our analysis in to a short time window when a stimulus is presented. This was accomplished here by analyzing changes in band-specific power, or amplitude, from pre- to post-stimulus, where stimulus represents the presentation of a single trial for a given cognitive task. Alpha ERD is symbolized by a negative change in alpha power, and theta ERS is symbolized with a positive change in theta power. 3 Research on fMRI resting state networks has clarified the role of the DMN in internally directed cognition (e.g., mind-wandering), and the FPN in active focused cognition; however, the relation of these networks to similar functional properties of neuroelectric activity is not yet understood. This study will analyze whether EEG phenomena of alpha event-related desynchronization (ERD) and theta-band event-related synchronization (ERS) during task performance represent neuroelectric parallels of DMN down-regulation and FPN activity, respectively. The construct of cognitive control has great potential for analyzing neural network activity through electrophysiology by allowing us to relate theta synchronization with FPN activity and alpha desynchronization with DMN activity (Scheeringa et al., 2009). This thesis will examine the intracranial EEG patterns of 8 epilepsy patients who underwent presurgical mapping for treatment of epilepsy. Patterns of desynchronization and synchronization will be compared to anatomical representations of the DMN and FPN to examine whether there are spatial preferences for these neural networks. The hypothesized spatial preferences are an alpha-band desynchronization pre-stimulus in DMN areas of the brain, and a theta-band synchronization post-stimulus in FPN areas of the brain. Understanding the parallels between electrophysiology and neural networks will allow for a greater understanding of cognitive processing at the neural level. In utilizing technology with better spatial resolution (Kim et al., 1997), constructs such as cognitive performance can be more accurately characterized in the brain through the dimensions of frequency, time, and space. Finally, understanding how these processes occur can result in valuable insight on potential classification and treatment for certain clinical diagnoses 4 such as Epilepsy, Alzheimer’s, or ADHD. For example, a decline in overall theta power within the frontoparietal network could represent a biomarker for the onset of Alzheimer’s Disease if these variables were found to be related. Understanding the neural mechanisms associated with clinical diagnoses could provide new insight for improvement of classification and treatment practices. METHODS Participants This study will examine the intracranial EEG patterns of 8 patients who underwent presurgical mapping for treatment of epilepsy. Participants were recruited from Columbia University in New York and Massachusetts General Hospital in Boston, and were screened for participation based on three factors. Only patients who were undergoing neurosurgical monitoring for treatment of medically refractory epilepsy, over 18 years old, and judged to have an IQ above 70 by a neuropsychologist were eligible for participation. Roughly half of the recruited participants were female (42%), and half male (58%). All participants had grid or depth electrodes implanted surgically for specific clinical purposes (see Figure 1). 5 Figure 1. The above image displays the distribution of electrodes for a single participant. The colored brain regions indicate a neural network parcellation (Yeo et al., 2011). Procedure With electrodes implanted in the brain for presurgical mapping of epilepsy patients, all participants consented to complete a simple cognitive task under electroencephalographic (EEG) recording after pre-screening as described above. Participants completed the Multisource Interference Task (MSIT), a visuospatial cognitive task described in more detail below. In total, recordings were obtained from 1899 electrocorticography sites or depth electrodes. Measures Task Participants completed the MSIT (similar to the Stroop task), which manipulates spatial and numerical interference to produce three levels of task difficulty (Bush & Shin, 6 2006). For each trial, participants were asked to identify a target based on a stimulus of a sequence of three numbers. To answer correctly, the participant had to press one of three buttons corresponding to the target identity: 1, 2, or 3. Some trials provided spatial or numeric interference by having incongruence between the target identity and the spatial position or numeric identity. For example, if the prompt asked for which number was different, both sequences of “112” and “121” would be followed by a correct response of pushing the second button, indicating the number “2”. In this scenario, the first sequence has spatial interference in that the number “2” is in the third position. An example of numeric interference would be a prompt asking for the number in the third position of either sequence above. In combining different combinations of these two modes of spatial interference, this task produces three main levels of difficulty: type 0, type 1a and 1b, and type 2. All participants completed at least one experimental session, where some participants completed multiple sessions. The average number of trials per session was 285 trials, ranging from 134 to 500 trials. In total, data were collected from 1899 electrodes distributed among 19 patients. The MSIT is a visuospatial working memory task that has been related to dorsal Anterior Cingulate Cortex (dACC) activation and frontoparietal network activity (Bush et al., 2003; Sheth et al., 2012). Knowing that the dACC has been studied in many lines of attention and cognitive processing research, it is likely that the connection between the dACC and the MSIT indicate this task elicits goal-driven behavior and cognitive effort. Previous research has validated the use of this interference task to elicit frontoparietal network activity (Bush et al., 2006). 7 ECoG Data Cleaning and Signal Processing Before analyses were run, raw EEG data was pre-processed to remove noise. First, noisy channels were identified and removed by visual inspection. Next, line noise was removed with a band stop filter (at 60 Hz), and all trials were demeaned, meaning that the mean voltage over whole intervals was subtracted from each time-point. Following, demeaned trials were re-referenced with a common average. Finally, noisy trials were identified and removed by visual inspection. After the data were pre-processed as described above, signal processing steps were taken to produce final measures of theta- and alpha-band power change. First, cleaned epochs were convolved with a series of morlet wavelets – there were a total of 23 wavelets ranging from 3-14 Hz in 0.5 Hz intervals and each wavelet was fixed to five cycles. The result of the convolution was a complex time-frequency representation of the brain activity associated with each trial from -3000 ms pre-stimulus to +3000 ms poststimulus and from 3 to 14 Hz. Each of these complex matrices was then multiplied by its complex conjugate to convert to power (amplitude squared). For each trial, we then averaged over the frequency bands comprising the conventional theta (3-7 Hz) and alpha (8-13 Hz), producing theta and alpha power for each trial. After the raw data was transformed into power, we then extracted pre-and poststimulus data, where a stimulus represents the timepoint during which a single trial of the task was presented. The interval from -1.0 to -0.5 seconds was defined as the prestimulus interval, and +0.5 to +2.0 seconds was defined as the post-stimulus interval. The post-stimulus interval corresponds to the period of time during which participants were considering and making a response, while the pre-stimulus interval corresponds to the 8 period of time just before the stimulus of the trial was presented. Theta-band task-related power change (TRPC) was calculated as the average voltage over the post-stimulus period minus the average voltage in the pre-stimulus period for each trial. The same process was completed for calculation of alpha-band TRPC. The final data were single values for theta- and alpha-band TRPC for each trial, where positive values indicate ERS, and negative values indicate ERD. Analysis Plan Analysis of EEG patterns related alpha ERD and theta ERS measures to DMN and FPN activation on a trial-by-trial basis. To complete this analysis, spatial locations of all electrodes in the brain for each patient were first obtained. This was done by mapping structural MRI data onto a template brain to standardize the coordinates for each electrode into Montreal Neurological Institute (MNI) space (Chau & McIntosh, 2005). Then, these electrode coordinates were matched to that of the closest cortical region within the YEO parcellation (Yeo et al., 2011), existing in the same space. This provided the network identity and electrode label information, e.g., an electrode with X, Y, and Z coordinates was matched to electrode ITG6, which is in the DMN. After network identities were assigned to each electrode, signals of alpha (8-13 Hz) and theta (3-7 Hz) TRPC were obtained for each electrode and trial. In condensing this information with a wavelet transformation and pre-processing as described above, we found that the final data revealed power changes in range of alpha and theta bands. Because electrodes were implanted for clinical purposes, we obtained incomprehensive coverage of all seven neural networks. Due to a lack of consistency in network coverage, 9 data from electrodes within the somatomotor, ventral attention, and limbic networks were excluded. Additionally, participants who did not complete MSIT trials at all three levels of difficulty were excluded as well. The final dataset included eight participants with electrodes in four neural networks. The first analysis revealed the increase in task difficulty as a potential predictor of task performance in terms of reaction time. Behavioral measures of reaction time (in milliseconds) for the eight participants were recorded at each MSIT trial and then averaged for each condition of task difficulty. Then, a repeated measures one-way Analysis of Variance (ANOVA), assessed within subjects, examined the effect of the three-level condition of task difficulty (low, medium, and high) on reaction time. A follow-up pairwise comparison with a Bonferroni correction revealed which levels of task difficulty contributed to the main effect on reaction time. The result of this analysis was compared to theta power change as described below to better understand how behavioral and cognitive measures are related. The second set of analyses examined electrophysiological measures, network identities, and task difficulty. All electrodes were placed in a network category (e.g., DMN). Then, each trial was categorized further by difficulty level. Changes in theta- and alpha-band power from pre- to post-stimulus were measured for each trial and then averaged within each difficulty level and neural network. The final measures of power change were analyzed via a two-way repeated measures ANOVA, with two withinsubjects factors – neural network identity (4 levels) and task difficulty (low, medium, high). The hypothesized results were a decrease in alpha power for DMN electrodes and an increase in theta power for FPN electrodes, and a potential interaction effect of neural 10 network identity with task difficulty. Post-hoc exploratory pairwise comparison tests were conducted to investigate potential level effects of neural network identity and task condition on change in theta and alpha power for development of future hypotheses. RESULTS Design and Analysis To better understand the role of neural networks in cognitive information processing, we measured the electrophysiologic patterns occurring in the brain as participants completed the MSIT, a visuospatial cognitive task. Specifically, we recorded network identities according to electrode locations, reaction times according to trial difficulty, and the change in frequency-band specific power from pre- to post-stimulus, averaged across neural networks and conditions of task difficulty. We were most interested in the default mode and frontoparietal networks, as they have been linked to temporal patterns of theta and alpha rhythms in past literature (Scheeringa et al., 2009). See Figure 2 for an example of electrode distribution and theta-band power change. We predicted that task performance, in terms of reaction time, would show a positive relationship with theta-band synchronization. Additionally, we hypothesized that theta-band ERS, defined as a positive change in theta power, would be seen in the frontoparietal network (FPN), and post-stimulus alpha-band ERD, defined as a decrease in alpha power from pre- to post-stimulus, would be seen in the default mode network (DMN). We conducted several repeated measures analyses of variance (ANOVA) to evaluate the factors of neural network identity and task difficulty, and then conducted 11 exploratory follow-up analyses to investigate whether there was evidence for specific effects of task difficulty within four neural networks to be replicated in future work. Figure 2. The above image displays the distribution of electrodes for a single participant. Red spheres indicate a negative theta power change (ERD) and blue spheres indicate a positive theta power change (ERS). The size of the sphere indicates the degree of power change. The colored brain regions illustrate a parcellation of neural networks (Yeo et al., 2011). Behavioral Measures In comparing average reaction times at different levels of task difficulty, we found a positive relationship as shown in Figure 3. The lowest difficulty level had a mean reaction time of 1163.1 ms (SD = 308.9 ms), the middle difficulty level had a mean reaction time of 1219.7 ms (SD = 313.4 ms), and the highest difficulty level had a mean reaction time of 1549.1 ms (SD = 379.4 ms). To better understand this relationship, we conducted a one-way repeated measures ANOVA with our three-level factor of task difficulty as a within-subjects factor. Because Mauchly’s sphericity assumption was not 12 met, we used a Greenhouse-Geisser correction and found a significant effect of task difficulty on reaction time [F(1.15, 8.05) = 16.11, p = 0.003, η2 = 0.697]. This result supports the positive relationship shown below. To better characterize this effect, we conducted a follow-up test with a Bonferroni correction for multiple comparisons to determine which levels of difficulty were significantly different from one another. We found that reaction time for the highest difficulty level was significantly greater than the middle difficulty level (p = 0.028, 95% CI for mean difference [38.81, 619.94]) and the lowest difficulty level (p = 0.006, 95% CI for mean difference [130.92, 641.08]). The middle task difficulty level was not found to significantly differ from the lowest level in terms of reaction time (p = 0.303, 95% CI for mean difference [-37.22, 150.47]). These results indicate that increased task difficulty is related to increased reaction time, particularly between the highest task difficulty level and the lower two difficulty levels. 13 (ANOVA) with both factors assessed within-subjects. Because Mauchly’s sphericity assumption was not met, we used a Greenhouse-Geisser correction. The results showed that there was no main effect of neural network [F(1.65, 11.54) = 1.52, p = 0.256], task difficulty [F(1.30, 9.10) = 4.27, p = 0.061], or an interaction between the two factors [F(2.50, 17.47) = 1.50, p = 0.251]. These results indicate that there was no effect of either neural network or task difficulty on change in theta power while participants performed the MSIT. Although there were no overall effects of either task difficulty or neural network on change in theta power, we conducted exploratory pairwise comparisons without correction for multiple comparisons to investigate potential effects of specific levels with a goal of developing future hypotheses. Figure 5 shows the mean change in theta power according to specific levels of neural network and task difficulty. We found that within the frontoparietal network, the middle task difficulty level showed a 23.48 (ɥV)2 greater decrease in theta power than the lowest task difficulty level (p = 0.048, 95% CI for mean difference [-46.73, -0.24]). In addition, within the dorsal attention network, there was a non-significant trend of decreased theta ERS with higher task difficulty. The highest task difficulty level showed a 49.30 (ɥV)2 greater decrease in theta power than the lowest task difficulty level (p = 0.056, 95% CI for mean difference [-100.31, 1.72]), and the middle difficulty level showed a 22.95 (ɥV)2 greater decrease in theta power than the lowest difficulty level (p = 0.053, 95% CI for mean difference [-46.27, 0.38]). These exploratory findings suggest that within the frontoparietal and dorsal attention networks, there may be a trend of 16 decreasing theta ERS with increased task difficulty; however, replication in an independent dataset with a larger sample size is needed to confirm this hypothesis. Figure 5. The above bar graph displays the mean change in theta power from pre- to post-stimulus according to neural network and task difficulty. Error bars indicate a single standard deviation. Positive values indicate theta event-related synchronization (ERS), and negative values indicate event-related desynchronization (ERD). Alpha-Band Results Similar to the distribution of change in theta power across the different neural networks, the distribution of changes in alpha power from pre- to post-stimulus during MSIT trials showed extreme outliers, leading us to use measures of median and 17 ANOVA with factors of neural network identity and task condition containing four and three levels, accordingly. Again, Mauchly’s sphericity assumption was not met, so a Greenhouse-Geisser correction was used. The results showed that there was no main effect of network identity [F(1.30, 9.13) = 1.46, p = 0.268], task difficulty [F(1.12, 7.84) = 4.17, p = 0.073], or an interaction between the two factors [F(2.00, 14.02) = 2.46, p = 0.122]. Similar to the theta-band results, although there were no observed effects of either task difficulty or neural network, we conducted exploratory pairwise comparisons without correction for multiple comparisons to investigate whether evidence exists for specific level effects to be replicated in future work. Figure 7 shows the mean change in alpha power from pre- to post-stimulus according to the specific levels of neural network and task difficulty. In looking at the effects of task difficulty dependent on neural network, we found that within the frontoparietal network, the lowest difficulty level had a mean change in alpha power that was 17.459 (ɥV)2 greater than the middle difficulty level (p = 0.035, 95% CI for mean difference [1.588, 33.330]), and 21.102 (ɥV)2 greater than the highest difficulty level (p = 0.025, 95% CI for mean difference [3.539, 38.665). With replication, this result suggests that within the frontoparietal network, increased task difficulty is related to a greater decrease in alpha power, or alpha event-related desynchronization (ERD). In addition, we found that within the dorsal attention network, the lowest difficulty level had a mean change in alpha power that was 26.772 (ɥV)2 greater than the middle difficulty level (p = 0.022, 95% CI for mean difference [5.101, 48.444]). This results also suggests that greater task difficulty is related to a greater decrease in alpha 19 power, but positive mean values of change in alpha power suggest an overall alpha ERS within the dorsal attention network. Although there is no evidence for an effect of neural network or task difficulty alone on change in alpha power, the frontoparietal network shows an alpha event-related desynchronization (ERD) within-subjects, increasing with greater task difficulty, and the dorsal attention network shows an alpha event-related synchronization (ERS), decreasing with greater task difficulty. While we hypothesized that alpha ERD would be seen in the default mode network as compared to the frontoparietal network, these findings suggest greater differences in alpha synchronization between extrinsic and goal-oriented networks such as the frontoparietal and dorsal attention networks; however, replication with a larger independent sample is needed. 20 Figure 7. The above bar graph displays the mean change in alpha power from pre- to post-stimulus according to neural network and task difficulty. Positive values indicate event-related synchronization (ERS), and negative values indicate event-related desynchronization (ERD). Error bars indicate a single standard deviation. DISCUSSION We sought to determine whether the intracranial EEG phenomena of alpha eventrelated desynchronization (ERD) and theta-band event-related synchronization (ERS) during simple cognitive tasks might represent neuroelectric parallels of DMN downregulation and FPN activity, respectively. This was accomplished by investigating taskrelated patterns of alpha ERD and theta ERS at intracranial sites within the DMN and 21 FPN networks while participants completed a Multi-Source Interference Task (MSIT). The results did not support our specific hypotheses regarding expected power differences between the default mode and frontoparietal networks. However, exploratory analyses suggest there may be significant differences between levels of task difficulty within the default mode, frontoparietal, dorsal attention, and visual networks in replication with a larger sample size. Our initial hypotheses were driven by pervious findings on the characteristics and function of several neural networks. The default mode network (DMN) has been correlated with brain regions including the medial temporal lobe, the medial prefrontal cortex, and the posterior cingulate cortex (Yeo et al., 2011). This network is known to be involved in intrinsic brain activity, or mind wandering, referring to when the brain is not actively engaged in a specific task. Past literature has found a negative relationship between deactivation of the DMN and goal-driven behavior such as working memory processes (Anticevic et al., 2010). In terms of electrophysiology, the DMN has been found to consistently decrease its activity during active information processing as opposed to relaxed non-task states of cognitive functioning, supporting an intrinsic understanding of neural function (Raichle, 2015). The frontoparietal network (FPN) is associated with brain regions including the rostro- and dorsolateral prefrontal cortex, the anterior insula, the dorsal anterior cingulate cortex, and the anterior inferior parietal lobule (Yeo et al., 2011). In contrast to the DMN, the circuitry within these regions has been correlated with functions such as visuospatial analogical reasoning and active information processing (Watson & Chatterjee, 2012). According to Marek and Dosenbach (2018), the FPN is critical for our ability to 22 coordinate behavior in a rapid, accurate, and flexible goal-driven manner. The FPN has been outlined as a distinct control network, and theta-band oscillatory synchronization within this network has been associated with interference control (Cooper et al., 2015). Implications of Reaction Time Findings We predicted that exerted cognitive performance, measured by reaction times, would show a positive trend with task difficulty. We found that reaction time increased with task difficulty and pairwise comparisons suggested that this was especially the case between the greatest and both lower levels of task difficulty, supporting our hypothesis. On the other hand, we found the two lower levels of task difficulty to not differ in terms of reaction time, suggesting that the relationship between task difficulty and reaction time is not linear. Upon further investigation and replication, these findings could lead us to conclude that as task difficulty increases, reaction time changes according to a threshold or exponential model. However, future work can explore whether this pattern is observed during other cognitive tasks to exclude the possibility that these results only apply to the MSIT. An alternate explanation for the observed effects of task difficulty is that the lower two difficulty levels of the MSIT were not manipulated distinctly enough to produce differential change in reaction time within subjects. Implications of Theta-Band Findings for Each Neural Network The specific hypothesis related to theta-band frequency measures was that ERS would show preferential spatial patterns within the frontoparietal network over the 23 default mode network — in other words, there would be a greater increase in theta power from pre- to post-stimulus within the FPN compared to DMN. Previous research on neural networks indicates that FPN activity is related to goal driven behavior (Marek & Dosenbach, 2018), meaning that tasks involving greater cognitive effort, such as working memory tasks, will likely correspond with this network. Due to the Multi-Source Interference Task (MSIT) being a visuospatial working memory task associated with regions within the frontoparietal network (Bush et al., 2003; Sheth et al., 2012), we predicted participants would show an increase in theta power after the presentation of a visual stimulus. Contrary to this hypothesis, we found no significant difference between changes in theta power within the FPN and DMN. However, exploratory pairwise comparisons revealed a potential trend of increasing theta ERD within the frontoparietal network and decreasing theta ERS within the dorsal attention network with increasing task difficulty. Additionally, comparing mean change in theta power across conditions of neural network and task difficulty shows a non-significant trend of decreased theta power with increased task difficulty for every neural network. These results could suggest that the difficulty of the cognitive task being performed is related to mechanisms of cognitive engagement. Also, considering our previous findings on reaction time leads us to conclude that cognitive engagement may be reflected in cognitive performance and information processing speed. However, replication with a larger sample size is needed to confirm these exploratory findings. Although neural network identity did not show an effect on changes in theta power overall, exploratory pairwise comparisons suggest differing mechanisms of 24 cognitive engagement between the dorsal attention and frontoparietal networks. The dorsal attention (DAN) and default mode networks (DMN) are referred to as antagonistic networks in that the former represents extrinsic functioning (externally oriented) and the latter intrinsic functioning (Spreng et al., 2013). Moreover, the frontoparietal network (FPN) has been found to modulate the relationship between these two networks (Spreng et al., 2013), and to be anatomically positioned to integrate information between the DAN and DMN (Vincent et al., 2008). Since the construct of cognitive control is related to the frontoparietal network (Cooper et al., 2015), there may be more reason for analyzing the FPN as an integration network between the DAN and DMN during cognitive interference tasks than as an independent neural network. Implications of Alpha-Band Findings for Each Neural Network We hypothesized that alpha-band frequency measures would show desynchronization within the default mode network as opposed to the frontoparietal network, meaning there would be a greater decrease in alpha power within the DMN than the FPN. This hypothesis was founded on past literature which revealed a link between deactivation of the DMN and goal-driven behavior such as working memory (Anticevic et al., 2010). A decrease in alpha-band power within the DMN could indicate a mechanism of resources and energy being redistributed to other networks during taskdriven behavior for increased performance and processing. Contrary to this hypothesis, we found no main effect of neural network or task difficulty on change in alpha power. However, exploratory pairwise comparison tests revealed an apparent trend of greater alpha ERD within the frontoparietal network with 25 increasing task difficulty. With replication, this finding could suggest a mechanism of inhibition as tasks get more difficult in regions associated with cognitive control, such as the frontoparietal network (Marek & Dosenbach, 2018). In addition, although no main effect of neural network was found, the great variability in alpha power change, particularly within the visual network, may contribute to this result. The mean change in alpha power appears to be the greatest in the visual network, followed by the dorsal attention network, and indicates alpha ERS in both networks. However, great overlap in error bars, owing to a small sample size and specific limitations discussed below, may have led to non-significant results. In replication with a greater independent sample, specific effects of neural network and task difficulty may be evident. Limitations The most significant limitations of this thesis were the distribution of intracranial electrodes and lack of control over the selection of participants. First, because all participants had intracranial electrodes placed for clinical purposes, we were unable to manipulate where these electrodes were located and what neural networks they represented. Further, because all participants had unique brain anatomy and clinical concerns, there was little overlap in electrode locations between subjects. These factors likely contributed to great variability in frequency-band results owing to limited spatial coverage of neural networks. Second, because all participants were recruited based on having intracranial electrodes implanted for the purpose of presurgical mapping for treatment of epilepsy, the 26 generalizability of these findings and implications to follow may be limited. The effect of the clinical diagnosis of epilepsy on neural mechanisms of information processing is unclear. Future research could explore the connection between psychiatric conditions and neural processes within the brain to better understand this potential limitation (Mormann et al., 2000). Implications for Utilizing Intracranial EEG Technology Although the results of this study are limited by variability between participants and lack of experimental control over electrode placement, the exploratory findings regarding theta and alpha power suggest that there may be important information to be learned about cognitive performance within the default mode, frontoparietal, visual and dorsal attention networks through intracranial technology. More precisely, the overall increase in alpha power within the visual network across the presentation of a visual stimulus in combination with the overall increase in theta power within the dorsal attention network could suggest that fluctuations in brain frequency-band power during cognitive tasks are more present in these networks than the frontoparietal network. Despite the lack of observed main effects of neural network or task difficulty, this thesis supports the method of using intracranial encephalography as a tool to study neural network activity. Not only do these findings provide evidence for an effective method of understanding neural patterns of information processing, but also a potential solution to the problem of spatial and temporal resolution (Kim et al., 1997). Intracranial EEG is advantageous in that electrodes are placed directly into the brain, leading to fewer obstacles associated with the skin and skull barrier. The support for using intracranial 27 data to understand neural network activity, which is more commonly studied with imaging technology such as fMRI, has important implications for improvement of data resolution in future research and clinical practices. Ultimately, the findings presented in this study could also be pertinent to clinical practice. With a better understanding of how frequency band-specific patterns represent mechanisms of neural networks, scientists could be provided with additional resources to diagnose and treat psychiatric conditions and have a better understanding of the underlying neural mechanisms involved with these conditions. For example, decreased theta synchronization within control networks such as the dorsal attention network could provide valuable information about the onset of a clinical diagnosis such as Alzheimer’s Disease. Opportunities for Future Research Although the results of this study do not support the initial frequency band hypotheses surrounding the frontoparietal and default mode networks, it not yet clear whether this is owing to a lack of effect within these neural networks or the limitations of the participant sample. An opportunity for future research could be to analyze more detailed maps of network anatomy for the spatial and temporal patterns of cognitive function. Also, these frequency-band specific hypotheses could be tested within specific brain regions or nodes of the frontoparietal or default mode network, rather than networks more broadly. This could lead to more specific findings regarding the regional fluctuations in theta and alpha rhythms. 28 Another area of future study is to learn about the spatial and temporal patterns within the broad range of frequency instead of restricting hypotheses to band-specific predictions. This exploratory research could lead to a better understanding of how broadband rhythms and spontaneous activity are related to neural processes (Myers et al., 2014). Better understanding how slow delta and high gamma rhythms contribute to trialto-trial variability could allow for more explanatory power of the underlying neural mechanisms involved in cognitive functioning and information processing. The relationship between neural processes within the brain and human behavior is poorly understood. A great opportunity to learn more about this relationship is to investigate how the behavioral measures of cognitive performance correlate with neural mechanisms. In this thesis, we found that increased task difficulty was related to decreased cognitive performance in terms of reaction time, and comparison of means revealed a potential overall increase in theta power within the dorsal attention network, suggesting a connection between cognitive and behavioral measures of engagement. However, the overlap of error bars in the distribution of theta power change between subjects suggests great individual differences. Future research could provide a better understanding of how inter-subject variability fits into the broader literature on cognitive and behavioral measures of exerted effort. Finally, the task presented in this study has been shown to elicit working memory and visuospatial processes (Bush & Shin, 2006), but it is not clear whether the effects found here could be evident with participation in other cognitive tasks. A future direction could be to uncover more about how the task itself relates to information processing by recording reaction times and power change under several validated cognitive tasks. This 29 could allow for a better understanding of how context factors in to measured patterns of cognitive engagement. Summary and Conclusions Neither the theta-band or alpha-band specific hypotheses regarding the frontoparietal and default mode networks were supported in the current results. However, additional exploratory analyses revealed interesting potential findings within the frontoparietal and dorsal attention networks, guiding future research to investigate the neural mechanisms of information processing within or between these networks. Additionally, although the greatest alpha ERS was seen in the visual network, this finding was not found to be significant likely because of great variability stemming from the limitations of the participant sample. Future research can characterize the change in alpha power within the visual network through replication with a larger sample size to better understand this potential effect. This study provides valuable insight into where scientists can focus their specific research questions surrounding information processing, both spatially and temporally. With these findings, it is likely that there are important mechanisms of information processing to be uncovered within the default mode, frontoparietal, visual and dorsal attention networks. Likewise, parallel change in reaction time and theta power as task difficulty increased suggests a link between cognitive and behavioral measures of engagement. With replication, these findings could permit a reinterpretation of previous literature on frequency-band synchronization and desynchronization in relation to neural 30 network activity and support new ways to study these networks using neuroelectric recordings. References Anticevic, A., Repovs, G., Shulman, G. L., & Barch, D. M. (2010). When less is more: TPJ and default network deactivation during encoding predicts working memory performance. Neuroimage, 49(3), 2638-2648. http://dx.doi.org/10.1016/j.neuroimage.2009.11.008 Bush, G., Shin, L. M., Holmes, J., Rosen, B. R., & Vogt, B. A. (2003). The Multi-Source Interference Task: validation study with fMRI in individual subjects. Molecular psychiatry, 8(1), 60. http://dx.doi.org/10.1038/sj.mp.4001217 Bush, G., & Shin, L. M. (2006). The Multi-Source Interference Task: an fMRI task that reliably activates the cingulo-frontal-parietal cognitive/attention network. Nature protocols, 1(1), 308-313. https://dx.doi.org/10.1038/nprot.2006.48 Cavanagh, J. F., & Frank, M. J. (2014). Frontal theta as a mechanism for cognitive control. Trends in cognitive sciences, 18(8), 414-421. https://dx.doi.org/10.1016/j.tics.2014.04.012 Chau, W., & McIntosh, A. R. (2005). The Talairach coordinate of a point in the MNI space: how to interpret it. Neuroimage, 25(2), 408-416. https://doi.org/10.1016/j.neuroimage.2004.12.007 Cooper, P. S., Wong, A. S., Fulham, W. R., Thienel, R., Mansfield, E., Michie, P. T., & 31 Karayanidis, F. (2015). Theta frontoparietal connectivity associated with proactive and reactive cognitive control processes. Neuroimage, 108, 354-363. https://dx.doi.org/10.1016/j.neuroimage.2014.12.028 Kim, S. G., Richter, W., & Uǧurbil, K. (1997). Limitations of temporal resolution in functional MRI. Magnetic resonance in medicine, 37(4), 631-636. https://dx.doi.org/10.1002/mrm.1910370427 Marek, S., & Dosenbach, N. U. (2018). The frontoparietal network: function, electrophysiology, and importance of individual precision mapping. Dialogues in clinical neuroscience, 20(2), 133. http://dx.doi.org/10.31887/DCNS.2018.20.2/smarek Mormann, F., Lehnertz, K., David, P., & Elger, C. E. (2000). Mean phase coherence as a measure for phase synchronization and its application to the EEG of epilepsy patients. Physica D: Nonlinear Phenomena, 144(3-4), 358-369. https://doi.org/10.1016/S0167-2789(00)00087-7 Myers, N. E., Stokes, M. G., Walther, L., & Nobre, A. C. (2014). Oscillatory brain state predicts variability in working memory. Journal of Neuroscience, 34(23), 77357743. https://dx.doi.org/10.1523/jneurosci.4741-13.2014 Pfurtscheller, G., & Da Silva, F. L. (1999). Event-related EEG/MEG synchronization and desynchronization: basic principles. Clinical neurophysiology, 110(11), 18421857. http://dx.doi.org/10.1016/s1388-2457(99)00141-8 Raichle, Marcus E., et al. "A default mode of brain function." Proceedings of the National Academy of Sciences 98.2 (2001): 676-682. https://doi.org/10.1073/pnas.98.2.676 32 Raichle, M. E. (2015). The brain's default mode network. Annual review of neuroscience, 38, 433-447. https://doi.org/10.1146/annurev-neuro-071013014030 Scheeringa, R., Petersson, K. M., Oostenveld, R., Norris, D. G., Hagoort, P., & Bastiaansen, M. C. (2009). Trial-by-trial coupling between EEG and BOLD identifies networks related to alpha and theta EEG power increases during working memory maintenance. Neuroimage, 44(3), 1224-1238. http://dx.doi.org/10.1016/j.neuroimage.2008.08.041 Sheth, S. A., Mian, M. K., Patel, S. R., Asaad, W. F., Williams, Z. M., Dougherty, D. D., ... & Eskandar, E. N. (2012). Human dorsal anterior cingulate cortex neurons mediate ongoing behavioural adaptation. Nature, 488(7410), 218-221. https://dx.doi.org/10.1038/nature11239 Spreng, R. N., Sepulcre, J., Turner, G. R., Stevens, W. D., & Schacter, D. L. (2013). Intrinsic architecture underlying the relations among the default, dorsal attention, and frontoparietal control networks of the human brain. Journal of cognitive neuroscience, 25(1), 74-86. https://dx.doi.org/10.1162/jocn_a_00281 Vincent, J. L., Kahn, I., Snyder, A. Z., Raichle, M. E., & Buckner, R. L. (2008). Evidence for a frontoparietal control system revealed by intrinsic functional connectivity. Journal of neurophysiology, 100(6), 3328-3342. https://dx.doi.org/10.1152/jn.90355.2008 Watson, C. E., & Chatterjee, A. (2012). A bilateral frontoparietal network underlies visuospatial analogical reasoning. Neuroimage, 59(3), 2831-2838. http://dx.doi.org/10.1016/j.neuroimage.2011.09.030 33 Yeo, B. T., Krienen, F. M., Sepulcre, J., Sabuncu, M. R., Lashkari, D., Hollinshead, M., ... & Fischl, B. (2011). The organization of the human cerebral cortex estimated by intrinsic functional connectivity. Journal of neurophysiology. http://dx.doi.org/10.1152/jn.00338.2011 34 Name of Candidate: Ashley Anderson Birth date: August 2, 1999 Birth place: Chicago, Illinois Address: 5129 Solar Heights Dr Eugene, OR, 97405 35