Poster session 1 (12:45-1:45pm, Day 1)

While we perceive our world as continuous, the human perceptual system is limited in processing incoming information. We will see discrete picture as continuous, if the presentation time between pictures is short enough. The neural mechanisms of temporal perception, however, are not well understood. A candidate mechanism are alpha oscillations. Alpha oscillations have been shown to correlate with visual perception. Using near threshold stimuli, results suggest an influence of prestimulus alpha phase and our detection ability.

In this study, we investigated how prestimulus alpha oscillations influence visual temporal perception. Specifically, we were interested whether prestimulus alpha phase influences behavioural accuracy and shapes integration windows and/or modulates perceptual processes. In an MEG study, participants had to integrate two visual stimuli, separated by a stimulus onset asynchrony, and report the position of a missing element. Source reconstruction was used to evaluate the location of phase effects by calculating the quantitative difference between phase angles values of correctly and incorrectly integrated trials.

We found a cluster between -0.8 and -0.5 s (relative to stimulus presentation) and between 8-20 Hz in parieto-occipital regions in which phases differed between correct and incorrect perceptions. Additionally, behavioural performance decreased when deviating from the participants’ individually preferred phase. We also found a correlation between early event related components (N100) and phase. Again, the amplitude of the N100 decreased when deviating from the subject specific preferred phase. In contrast, we found no evidence that prestimulus phase modulates temporal integration windows. We conclude that, that stimulus processes is optimal at a certain phase and therefore results in increased behavioural performance.

Transcranial Alternating Current Stimulation (tACS) is a non-invasive brain stimulation technique in which alternating sinusoidal currents are applied to the scalp. In turn, the electric field generated by the stimulation is presumed to entrain endogenous brain oscillations in a frequency-specific manner. Some studies report that applying tACS at alpha frequencies induce changes that persist beyond the stimulation period. However, there is no consensus on the ideal stimulation protocol to elicit long-lasting aftereffects. Stimulation parameters such as frequency, intensity and duration of applied current, and electrode montage differ widely between studies, thereby leading to several contradictory results. The origin of the aftereffects of stimulation also remains a matter of contention as tACS effects may be mediated by transcutaneous stimulation of peripheral nerves. In this study, we aimed to elucidate paradoxical reports by modulating alpha power in the right somatosensory cortex while controlling for peripheral nerve stimulation. We were interested in testing whether comparably short periods of tACS are able to entrain neuronal oscillations. To this end, during simultaneous acquisition of MEG, we administered tACS at individual alpha frequency in an intermittent on/off pattern. tACS was applied in 10 or 30 second trains in two separate blocks. Changes in alpha power in the post-stimulation intervals were compared to resting-state baseline as well as sham. The results show focal enhancements of alpha power at stimulation frequency relative to baseline, and this power modulation was more substantial than sham. These findings provide evidence for direct transcranial modulation of alpha power in the somatosensory cortex and point to the importance of combining tACS with electrophysiological recordings to probe the causal role of alpha oscillations.

Background: Our previous fMRI work showed that a frontoparietal network posited to play a critical role in attentional control (the multiple-demand, MD, network) selectively codes information at the intersection of spatial and feature attention. Further, our MEG work showed this tuning arises first in higher-order brain regions, which in turn Granger-causally influence coding in visual cortex. We thus propose that the highly selective task- relevant representations of MD regions act as a source of bias, directing task-related representations to be similarly precise in domain-specific regions. Using an exploratory case series approach, we ask how and when this selective prioritisation changes after damage to the MD system.

Methods: We recruited six patients with unilateral, primarily parietal, chronic lesions of mixed aetiology (three male, three female; three left, three right hemisphere; age range: 55 to 75), and ten age-, sex- and education- matched controls (six male, four female; age range: 53 to 76). Participants covertly attended to one of two objects, presented left and right of a fixation cross (spatial attention), and reported the attended object's colour ("red" or "green") or shape ("X-shaped" or "flat") (feature attention) via button press while MEG data were recorded. Attention was also assessed behaviourally using standard neuropsychological tests. In ongoing analyses, we are using multivariate pattern analysis in sensor and volumetric source space to quantify the dynamic representation of attended and unattended stimulus information.

Objectives: We aim to examine whether chronic parietal lesions, typically associated with subtle spatial attention deficits, affect the specificity and/or timecourse of task-related representations. Moreover, we will explore whether task-related coding in patients relates to performance on behavioural measures of attentional bias. Through this work, we aim to elucidate how attentional prioritisation is achieved after brain damage.

Cognitive ageing in humans is marked by declines in vision and hearing, and manifest as delays in sensory-evoked potentials, likely reflecting a general slowing of information processing with age. There has been evidence that different mechanisms may drive age-related delays in distinct ways—for instance, visual-evoked responses are marked by a constant delay—suggesting reduction in transmission mediated by white matter tracts; while auditory-evoked responses are marked by a delay that accumulates over time—likely mediated by deficits in local processing in grey matter [1].

In the proposed study, we investigate the causal role of these mechanisms underlying age-related delays in visual and auditory evoked responses using dynamic causal modelling (DCM) of magnetoencephalography (MEG) data from a large (N=630) healthy population-derived sample from the Cambridge Centre for Ageing & Neuroscience (Cam-CAN) [2]. Data was acquired during an audiovisual task in which participants passively experienced 120 trials with either a visual stimulus consisting of two circular checkerboards, or binaural auditory tone.

We apply DCM to model evoked responses for the visual and audio trial from MEG data [3]. We model the extrastriate and primary auditory cortices as generators of visual and auditory evoked responses respectively. We hypothesize that age-related delays in transmission will result in reduced strength of inputs to the generators, while age-related deficits in local processing will result in reduced strength of intrinsic connections in the generators. We test these predictions using Bayesian model comparison within the Parametric Empirical Bayes (PEB) framework. Our findings will help elucidate the different neural mechanisms that could result in distinct age-related delays in different sensory systems.

Ref: 1. D. Price et al, Nat Commun. 8, 15671 (2017)
2. J. R. Taylor et al, NeuroImage. 144, 262–269 (2017)
3. O. David et al, NeuroImage. 30, 1255–1272 (2006)

Background: MEG studies typically rely on simple or abstract experiments to carefully probe cognition. Whilst powerful, these paradigms do not resemble how we use these faculties in everyday life, and so we need experiments which closer reflect the real world, rather than the laboratory. With the advent of MEG systems which are both wearable and allow the subject to around move freely, MEG is in a unique position generate high-quality data from complex paradigms.

Aim: To derive the organization of the motor cortex from spontaneous human movement.

Methods: 4 participants underwent 2 experiments; the first a simple block design experiment where a participant was visually cued to move one of their 4 limbs, and a second where they danced to an audio recording the “Hokey Cokey”. All participants wore custom-fitting helmets containing an array of multi-axis optically pumped magnetometers (OPMs) and underwent simultaneous recordings of MEG and video of them performing the experiments. The video was processed with a deep-learning framework to extract key points (wrist, elbow, etc.) from the body and key point velocities were partitioned with a hidden Markov model to detect states of movement.

Results: Hidden Markov Model states corresponding to the 4 limbs were recovered from the video in both experiments. In the block design motor experiment, the video states closely matched stimulus timings (AUC: 0.912). Source reconstruction of beta-band activity based on either the block design timings and video-derived timings showed similar spatial organisation, with the video-derived timings generating higher magnitude statistical images. States derived from the dancing paradigm and their associated cortical maps also closely resemble the cortical homunculus.

Conclusion: We can partition and analyse MEG data to recover motor function based on video-derived subject behaviour, opening the possibility of more complex naturalistic studies in the future.

Background
Theta-gamma phase-amplitude coupling (PAC) is thought to be an important brain mechanism that allows coordination of neural activity within and between distinct cortical regions to control behaviour. Here, we investigated oscillatory dynamics in the spoken language network, with a focus on theta-gamma PAC.

Aim
How do principle neurophysiological mechanisms support language functioning. Methods 27 healthy older adults (≥50 years) completed an overt picture naming task during an MEG scan. Stimuli consisted of black and white line drawings of objects randomised to a clear (naming) or scrambled (control) condition. Stimuli were presented on the screen for 2s, preceded by a 2.5s fixation period. Participants were instructed to name clear stimuli and respond ‘no’ when scrambled stimuli appeared. The scrambled condition controlled for low-level visual input and overt speech output but did not involve lexical retrieval processes. MEG analyses were conducted using SPM12 on response-locked epochs. Sources of oscillatory activity were reconstructed using an LCMV beamformer and contrasted between naming and control conditions at a statistical threshold of FWE-c p< 0.05 at voxel level. Theta-gamma PAC was estimated using KL divergence (Tort et al., 2010). Cluster-based-permutation testing was used to assess PAC differences between conditions.

Results
Picture naming, relative to control, induced theta (4-6Hz) synchronisation followed by an alpha-beta (8–20Hz) desynchronization during word retrieval. Theta activity localised to left superior and inferior frontal gyri and left superior and medial temporal gyri. These same regions also showed significant theta-gamma PAC that was only present during naming.

Discussion
Theta-gamma PAC appears to facilitate functional integration between frontal and temporal regions of the language network during spoken naming. Future work will explore whether theta-gamma PAC is altered in aphasia and if it can be modulated to improve anomia.

Speech communication is a fundamental aspect of human experience, enabling us to express thoughts and emotions. This study investigates the possibility of measuring the neural processes underlying social speech perception with electroencephalography (EEG). Past research has contributed significantly to our understanding of how speech is transformed into meaning by our brain. While previous work typically relied on simplified speech listening tasks, such as listening to sequences of isolated syllables, recent developments have shifted towards employing more naturalistic paradigms in ecologically-valid settings: the perception of continuous speech. In particular, EEG and magnetoencephalography demonstrated a robust relationship between speech inputs and the corresponding neural signal. This enabled researchers to probe the neural encoding of speech and language properties at various levels of abstraction. However, that work focused on how the human brain processes speech monologues, either in quiet or in noise, while ignoring one of the foundational roles of speech: social communication. Here we present an EEG experiment where we take the first step in that direction. Participants were presented with a dialogue between two interlocutors from conversations in a podcast-style setting. The multivariate Temporal Response Function (mTRF) methodology was used to measure the EEG encoding of the sound envelope and lexical predictions. We will present results supporting the hypothesis that the analytical mTRF framework for studying monologue listening also applies to third- person perception of dialogues. This social scenario presents challenges that have yet to be addressed in the literature, such as the adaptation of speech and language for their use in social contexts. We will discuss how the outcomes of this study could render a novel avenue for speech neurophysiology, enabling the investigation of social communication in more realistic scenarios involving natural speech listening.

Background
Optically-pumped magnetometers (OPMs) promise to make MEG technology more clinically accessible, particularly for epilepsy evaluation. However, it is critical that this new technology is validated against the current clinical standard – EEG. Here, we use a 128-channel OPM-MEG system (Cerca Magnetics Ltd) and a whole-head 64-channel EEG (Brain Vision LLC, NC, USA), to test the feasibility of simultaneous OPM-MEG and EEG measurements.

Methods
12 healthy adults (mean age 41 ± 13yrs, 8 female), undertook a right index finger abduction task (50 trials) and an eyes open/eyes closed task (5 trials). Tasks were completed 3 times: with simultaneous EEG/MEG, EEG only, and MEG only, allowing comparison of MEG signal quality in the presence of EEG and vice versa. Before each scanning session, the remnant magnetic field was nulled by combining optical tracking (OptiTrack, NaturalPoint Inc., USA) with magnetometer data from sensors in the helmet (Rea et al., 2021).

Results
For MEG and EEG at the sensor level, there is a clear increase in alpha with eyes closed in comparison with eyes open. For the finger abduction task, trials were time locked to the point of movement offset and averaged over trials for each participant, revealing a task-modulated beta response in EEG and MEG. The sensor with the highest SNR was selected separately for each participant and the SNR was calculated for EEG alone, MEG alone, and each in the presence of the other. There was no significant difference in the SNR of MEG signals with and without EEG (p=0.4) or EEG with and without MEG (p=0.5). The presence of EEG also did not affect the quality of residual field nulling.

Conclusion
It is possible to measure whole-head EEG and OPM-MEG together with good quality data. This paves the way for future clinical assessment of patients with epilepsy using the two modalities combined.

Background: As population ageing continues to surge globally, a major imperative exists to identify mechanisms of cognitive decline associated with aging. Working memory (WM) and decision-making (DM) are fundamental building blocks of cognition that deteriorate with age. While these processes are typically studied in isolation, recent computational and empirical studies indicate that a common neural circuit configuration is capable of maintaining (for WM) and integrating (for DM) information over time through shared attractor dynamics, and that this circuit is subject to shared sources of noise and bias that shape both WM and DM behaviour.

Methods: The present study leveraged this emerging, consolidative framework for understanding WM and DM to interrogate sources of age-related decline in both functions. Young and older adults (N=33 in each group) completed psychophysical tasks designed to parse sources of shared and unique variance in WM and DM behaviour while high-density scalp EEG was recorded.

Results: Results from both modalities, informed by analyses of noise and bias in WM and DM reports and decoding of task variables from EEG signals, converged to suggest a leading locus of age-related dysfunction – degraded sensory encoding – that gives rise to a specific pattern of decline across both domains.

Conclusions: These findings provide fundamental insights into the neural basis of these functions and their susceptibility to the deleterious effects of aging. More generally, we hope that the integrative approach to understanding WM and DM developed here will be of merit for pinpointing loci of dysfunction in mental disorders characterised by deficits in both functions.

Background: Alzheimer’s disease (AD) and behavioral variant frontotemporal dementia (bvFTD) lack well-understood characterization in diverse, non-stereotypical, and underrepresented populations. Electroencephalography (EEG) is a high-resolution, cost-effective technique for studying dementia globally, but lacks mechanistic models and produce non-replicable results.

Methods: We developed a generative whole-brain model that combines EEG source-level metaconnectivity, anatomical priors, and a perturbational approach. This model was applied to Global South participants (AD, bvFTD, and healthy controls).

Results: Metaconnectivity outperformed pairwise connectivity and revealed more viscous dynamics in patients, with altered metaconnectivity patterns associated with multimodal disease progression. The biophysical model showed that connectome disintegration and hypoexcitability triggered the altered metaconnectivity dynamics, and identified critical regions for brain stimulation. We replicated main results in a second subset of participants for validation with unharmonized, heterogeneous recording settings.

Implications: The results provide a novel agenda for developing diagnostic methods and model-inspired therapies in clinical, translational, and computational neuroscience.

Background: Major depressive disorder (MDD) carries a significant risk of suicide, particularly among individuals with a history of suicide attempts (SA), which may be associated with deficits in inhibitory control. This study aimed to investigate the role of abnormal neuronal oscillations in the inhibitory function deficits observed in SA patients.

Methods: A total of 111 participants, including 74 MDD patients with SA and 37 controls, underwent magnetoencephalography recordings while performing a GO/NOGO task. Time-frequency representations and phase-amplitude coupling were analyzed for the brain circuits involved in inhibitory function. Phase-slope indexes were calculated to determine power flow direction between regions.

Results: The SA group exhibited significantly increased reaction time and decreased judgment accuracy compared to other groups. During the perception stage of the GO task (around 125 ms), the SA group showed elevated alpha power in the ventral prefrontal cortex (VPFC) and reduced beta power in the dorsal anterior cingulate (dACC) compared to the other groups (p < 0.01). In the processing stage of the NOGO task (around 300 ms), the SA group displayed decreased beta power in the VPFC and increased alpha power in the dACC (p < 0.01). Alpha-beta decoupling was observed in the SA group during both tasks. Moreover, the decoupling from VPFC to dACC during the NOGO task significantly correlated with suicide risk level.

Conclusion: Our findings suggest that dysregulated oscillatory activities in the dACC and VPFC are associated with deficits in execution and inhibition functions, contributing to an increased risk of suicide in SA patients. The alpha-beta decoupling from the VPFC to the dACC holds promise as a neuro-electrophysiological biomarker for identifying individuals at potential suicide risk.

Multivariate pattern classification (MVPC) plays an important role in many areas of cognitive research including visual perception and memory. Many studies begin with a classifier training task to acquire data that a classifier can be trained on before subsequently being applied to the main experimental data. Classification can occur at multiple levels of analysis, for example, at the category or the exemplar level. Here we ask whether the task used during classifier training and the level of analysis at which it requires participants to process visual stimuli can influence subsequent classifier fidelity. Participants performed three tasks during MEG: a category discrimination task, in which participants needed to focus on category level information, an exemplar discrimination task, in which the participants needed to focus on exemplar level information, and an oddball task, which acted as a baseline in which neither category nor exemplar level information were required. In each task participants were presented with exemplars from eight categories: faces, plants, dogs, insects, buildings, cars, tools, and furniture. We applied MVPC to the data to determine whether the level of processing required in each task could influence classifier fidelity when classifying the stimuli at either the category or the exemplar level. The results from this experiment will help to inform the field on how to best structure pre-experimental classifier training tasks so as to optimise MEG classification accuracy.

With an increasing proportion of elderly people, there is a pressing need to disambiguate between healthy and pathological ageing of the brain. Here, we examine the impact of healthy ageing on networks of oscillatory activity using resting-state MEG recordings from 612 participants (18-88 years old). We source reconstruct this data to 52 regions of interest and estimate static and dynamic networks (power spectra, spatial power maps and coherence networks). Dynamic (transient) networks are identified using a Time-Delay Embedded Hidden Markov Model (TDE-HMM). We observe clear trajectories for changes in oscillatory properties as a function of age. For static networks we see a slowing of occipital alpha (8-13 Hz) activity into the theta band (4-7 Hz), increase in temporal alpha activity, decrease in whole-brain delta (1-4 Hz) activity and increase in frontal and sensorimotor beta (13-22 Hz) activity. We see that the coherence of each frequency band can sometimes change in the opposite direction to power; e.g., there is an increase in whole-brain delta coherence with age. Underlying insights into the causes of these static changes can be found by looking at the dynamics of networks identified using the TDE-HMM. We find that significant differences between young and old participants can be seen in metrics that summarise the “activity” time courses of each dynamic network (e.g., lifetime, interval, fractional occupancy). In particular, the strongest effect is an increase in the lifetime of a temporal alpha state which explains the increase in static temporal alpha power. Additionally, we see a lower fractional occupancy of an occipital alpha state, which explains the decrease in static occipital alpha with age. Understanding age-related changes to oscillatory activity is invaluable as a normative model. In this work, we characterise the expected functional networks at rest using a large population. These networks can be used as a basis set for smaller boutique studies.

Background: Magnetoencephalography (MEG) systems based on optically pumped magnetometers (OPMs) typically employ large, external, electromagnetic coils for active magnetic shielding [1] as well as sensor benchmarking and calibration [2]. Whilst coil systems formed from distributed windings can be designed to produce accurate magnetic field patterns [3], they are difficult to manufacture, occupy space inside the magnetically shielded room (MSR) and only operate over fixed regions [4]. By contrast, matrix coil systems (formed from a series of small simple unit coils, each with an individually controlled current) can effectively ‘re-design’ themselves to generate desired magnetic field patterns over multiple volumes [5].

Methods: Here we outline the construction, calibration and operation of a matrix coil system and its associated low-noise (<25 nV/√Hz@5Hz) coil drive system. Through the integration of the matrix coil with optical tracking with OPM data acquisition, field changes induced by participant movement are cancelled with low latency (25 ms).

Results: Beta-band modulation relating to a button pressing task was clearly observed in sensorimotor areas despite the presence of large (65 cm translations and 270° rotations) ambulatory participant movements during the recording.

Discussion: This work shows the capabilities of OPM-MEG to provide a platform for previously unrealisable studies of movement disorders, including Parkinson’s disease and gait ataxia, and exciting neuroscientific studies of spatial navigation and social interaction.

References:
[1] Boto, E.; et al. Nature 2018, doi:10.1038/nature26147. [2] Iivanainen, J.; et al. Sensors 2022, doi:10.3390/s22083059. [3] Zetter, R.; et al. J. Appl. Phys. 2020, doi:10.1063/5.0016087. [4] Holmes, N.; et al. Neuroimage 2018, doi:10.1016/j.neuroimage.2018.07.028. [5] Holmes, N.; et al. Neuroimage 2023, doi:10.1016/j.neuroimage.2023.120157.

Background: The adoption of MEG for clinical applications is partially hindered by its reliance on cryogenic sensors housed in non-life-span compliant, one-size-fits-all systems. Recently developed wearable systems using lightweight optically pumped magnetometers (OPMs) have shown promise, but the optimal design of paediatric OPM-MEG systems has not been established and normative paediatric OPM-MEG data is not widely available. We test the capability of a novel 186-channel OPM-MEG system to map the sensorimotor cortex, construct whole-brain functional connectomes in a paediatric cohort and compare results with data collected in adults.

Methods: MEG data were recorded in 26 children aged 2-13 years and 26 adults while they watched television and received tactile stimulation from braille stimulators. In each trial, 0.5s of tactile stimulation to the tips of the index or little finger was followed by 3s rest. Participants wore a size-matched, rigid helmet populated with 63 triaxial OPM sensors to give whole-head coverage. The OPM-MEG system was housed in a magnetically shielded room with active magnetic field compensation. Pseudo-T statistical maps of activation for index- and little-finger trials were generated and time-courses of activity in 78 cortical regions were estimated using beamforming. Source modelling was achieved using template MRIs, warped to fit individual head shapes estimated using 3D structured light scans.

Results: We observed the expected changes in oscillatory power from finger-tip stimulation and sensorimotor and visual networks in both groups. Children exhibited lower task-induced modulation in the beta band and connectomes revealed age-related changes.

Conclusion: OPM-MEG is an exciting new platform to study neurodevelopmental trajectories in both health and disease with the capability to detect network dynamics in children.

Background/Aims:
Agents can optimise behaviour by attributing undesired outcomes to different underlying causes, which require different actions: For example, a misjudgement (internal error), requires the alteration of an internal state to improve performance, e.g. slowing down to improve accuracy. Alternatively, a change in the world (external error), may require an agent to adjust their world model, e.g. learning new stimulus- response mappings. A candidate neural mechanism for integrating these distinct sources of error is sensorimotor beta oscillations, which consist of two components: A pre-movement desynchronisation, and a post-movement rebound.

Methods:
We combined EEG with a novel orientation judgement task where volunteers had to determine the direction of rotation of two gratings, one of which was associated with a reward. The absence of reward could be due to either a perceptual error (incorrect rotation direction) or a side error (volunteer reported the rotation of the unrewarded stimulus).

Results:
We show a cognitive double dissociation between beta desynchronization and rebound. Beta desynchronisation was associated specifically with residual reaction time; the degree to which a participants’ response was faster or slower than predicted by the difficulty of the rotation task, i.e., an index of internal state regarding vigilance, attention etc. In contrast, beta rebound was specifically associated with proximity to a decision to switch sides, indicating that it may relate to a participants’ internal model of the world and belief that that model needs updating.

Discussion/conclusions/implications:
The novel conclusion of this study is that beta desynchronisation and rebound are not merely ‘two sides of the same coin’. Rather, they index two distinct cognitive processes by which an agent may integrate two distinct causes of an undesirable outcome, respectively, updating internal state or adjusting an internal model.

Speech perception is an active process where our brains build expectations that are compared with the actual speech input. This phenomenon has been extensively studied with controlled experiments involving simplified listening tasks (e.g., single words, single sentences), indicating distinctive neural changes due to violations of linguistic expectations, such as lexical violations when hearing a word that is out of context. Recent work has demonstrated that the neural signature of that process can also be measured in response to continuous natural speech. Methodologies such as the multivariate temporal response function (TRF) enables the isolation of such prediction processes. Specifically, it was shown that neural signals co-vary with information such as semantic dissimilarity and lexical surprisal. This work is also relevant to the recent breakthroughs in large language models, as those precise and cohesive models of language (e.g., GPT) can be used for estimating expectations at multiple linguistic levels at once, which are then related with the neural signals. While previous TRF results showed significant relationships between neurophysiology recordings and such word-level information, the literature shows effects that are typically weak, leaving considerable uncertainty on whether the neural activity tracks that information only coarsely or at a fine-grained level, hampering the applicability of those metrics in basic and applied research. For example, given two possible models of lexical processing, could these measurements be used to determine which model is most physiologically plausible? In this study, we identified an improved TRF metric that is more sensitive to the effect of word-level information on neurophysiology signals. First, we present the new metric by showing its effectiveness on synthesised EEG data. Second, we will discuss how the new assessment metric performs on EEG and MEG responses to natural speech monologues.

Background:
Accurate modelling of functional brain networks is essential in the quest for understanding how the brain produces cognition. With recent studies, data driven models like the Hidden Markov Model (HMM, Baker et al), Dynamic Network Modes (DyNeMo, Gohil et al) are getting more attention due to their ability to infer fast temporal dynamics in dynamic functional networks in an unsupervised manner. However, these dynamic network models are limited by only giving a group level description, e.g., of the brain regions and spectral content in each brain network. While it is possible to post-hoc estimate subject-specific estimates of these networks, using so-called dual-estimation, this does not allow for the model to discover and benefit from subject-wise structure in the population, e.g. sub-groupings of subjects.

Methods:
We propose an extension to the current approaches (HMM, DyNeMo) that incorporates subject embedding (analogous to word embedding in Natural Language processing) into the group model. This effectively infers a “fingerprint”for each subject, which can group together subjects with similar spectra/networks.

Results:
Using simulated data, we show that the approach can reliably recover the underlying population structure and infer subject-specific estimates more accurately than post-hoc dual-estimation. Results on real data show that the model can distinguish data from different datasets and subjects with different demographics.

Conclusion:
We propose an approach that models subject variability in a principled way and have shown its advantages over traditional post-hoc dual-estimation on simulated data. On real data, we show its utility in uncovering underlying population structure. The more accurate subject-specific estimates can also benefit downstream prediction tasks.

Background:
Brain network changes across the adult lifespan are observable in electrophysiological recordings of human brain activity. Here, we identify markers of ageing in neuronal oscillations across four large, open access MEG/EEG datasets and explore the extent to which they are robust to differences in physiology, acquisition and recording modality.

Methods:
We analyse sensor space power spectra using a temporal General Linear Model (GLM) to estimate a power spectrum at the individual subject level, by using the output from a sliding window Fourier Transform at each frequency separately as the dependent variable while controlling for covariates such as blinking and scan duration at the individual subject level. A group GLM then then models between- subject factors such as head-size, sex, acquisition site and age affect the power spectrum.

Results:
A consistent set of age-effects are found. Ageing is associated with decreased low frequency power , increased low alpha power, decreased high alpha power and increased beta power. These effects are robust across different datasets, sensor types, MEG systems, and source reconstruction. The two alpha effects constitute a slowing of alpha peak frequency which is accompanied by a spatial shift from a strong peak in occipital pole in young adults to a more diffuse occipital/parietal/temporal distribution in older adults.

Discussion:
The effect of ageing in neuronal power spectra across healthy ageing is robust and reproducible at the group level.

Aim
Brain activity contains recurrent oscillatory activity in large scale cortical networks, but their temporal evolution is unclear. Our study introduces a new method for analyzing the long-term dynamics of these states and demonstrate that there is a tendency for the brain to cycle through the states in a particular order.

Methods
We used hidden Markov modelling to identify 12 network states from three large MEG datasets (MEGUK, CamCAN, HCP), and studied their temporal dynamics via a new method called Temporal Interval Network Density Analysis (TINDA). TINDA computes a measure for each pair of states indicating how much a state occurs in the first, versus the second, half of the time between subsequent visits to another state. This indicates if there is a tendency for a state to follow, or precede, another state over long time-scales. We then globally organize the states into a circle such that we minimize violations of the pairwise ordering learnt by TINDA.

Results
Applying TINDA to the MEGUK resting MEG data, we found unique pairwise state orderings for each network state. Globally ordering the states revealed an unmistakable cyclical pattern with the brain moving through network states in a consistent direction. The strength of the cycling was stronger than expected by chance, especially for longer intervals. We replicated these results in the CamCAN and HCP datasets and showed that cycle strength and cycle rate are related to age/sex, and cycle rate is heritable. Moreover, the cycle’s phase incorporated the resting states’ spatio-spectral features, progressing from high power/synchrony states to low frequency cognitive states, low power/synchrony states, and finally high frequency perceptual/attentional states.

Conclusion
We have developed a new method to study temporal dynamics in brain networks and revealed an intrinsically cyclical organization of network activity, which was replicated in three datasets and is predictive of personal traits.

Background/Aims:
The striking right-hemisphere dominance of neglect, plus observations of pseudo-neglect in healthy population (leftward attention in line bisection), support the widely accepted view that the right hemisphere of the brain is dominant for spatial attention. However, regions in the right hemisphere that cause neglect overlap with the ventral attention system – which activates for involuntary/ exogenous attention, and brain damage to regions that are part of the dorsal attention network– involved in the voluntary control of attention – do not produce neglect. Furthermore, brain stimulation to left parieto-occipital areas (not only right brain areas) have shown to modulate attention shifts. Even though the general dominance of the right hemisphere in attentional processes cannot be dismissed, evidence for its dominance in the voluntary control of spatial attention is scarce. Therefore, the question still stands: Is the right hemisphere dominant for spatial attention also when voluntarily controlled?

Methods: 32 participants undertook an attention-shift task with concurrent Magneto- and Electro-encephalogram recordings. Our experimental conditions control for general attentional processes, isolating the spatial component. We analyse individual hemispheric contribution to control of spatial attention at sensor and source-level.

Results:
We hypothesise that attentional control for spatial orienting arises from symmetric interhemispheric interaction.
Discussion:
This study will present results at sensor and source level, using complementary EEG and MEG data, relying on changes in alpha-amplitude as a reliable measure of hemispheric engagement during spatial attention deployment. We look to discuss our results in light of past literature on alpha-modulation that has used diverse experimental designs, to provide an answer to the question if the right hemisphere is dominant during voluntarily controlled spatial attention.

Background/Aims:
Electrooculography (EOG) and Electroencephalography (EEG) are integral to neurophysiological research and diagnostics. However, the presence of EOG artefacts in EEG data can complicate data interpretation. This research aimed to develop a computer vision-based system, using OpenCV, for non-invasive EOG data inference from video, and to explore its potential in enhancing EEG data interpretation.

Methods:
OpenCV was leveraged to track eye movements from video data and to infer corresponding EOG signals. This system was then extensively tested for accuracy and reliability against traditional EOG data. Further, the inferred EOG signals were integrated with EEG data for improved interpretation. The procedure involved identifying EEG artefacts related to eye movements, annotating these periods in the EEG data, and exploring the enriched understanding of brain activity patterns.

Results:
The developed system demonstrated strong agreement with traditional EOG data despite minor limitations related to resolution and precision. The application of inferred EOG signals to EEG data resulted in the enhanced understanding of brain activity, improved artifact identification, and EEG data annotation.

Discussion/Conclusions/Implications:
This research presents a promising approach for non-invasive, comfortable EOG data acquisition, and potential improvement in EEG data interpretation. It has significant implications for neuroscience, biomedical signal processing, and human-computer interaction. Future research can further optimize this system, extend its application to other types of artifacts, and explore automatic artifact tagging and removal.

Background:
Magnetoencephalography (MEG) is a non-invasive neuroimaging technique that measures the magnetic fields generated by neuronal electrical activity in the brain. MEG provides high temporal resolution and has become a valuable tool in neuroscience and clinical research. One of the challenges in analyzing MEG data is the non-stationarity of brain signals. The brain’s activity fluctuates over time, making it difficult to assume that the statistical properties of the data remain constant throughout the recording. Non-stationarity can arise from various factors, including changes in attention, arousal, or task demands. Dealing with non-stationarity is crucial for the accurate interpretation of MEG data. Researchers employ various strategies such as segmenting the MEG data into smaller time intervals, applying adaptive filtering, and time-frequency analysis for feature extraction.

Method:
In this work, we have analysed non-stationarity (i.e. covariate shift) on the publically available MEG data acquired from 17 healthy subjects and developed MEGNet for single-trial classification to decode or classify brain states or cognitive processes from individual trials of MEG data. MEGNet network is inspired by EEGNet (a deep learning-based architecture), which helps to avoid doing manual feature extraction and lets the network learn the hidden pattern in the data for pair-wise binary classification. Robust single-trial classification can be useful in various applications, including brain-computer interfaces (BCI), cognitive load estimation, and clinical diagnosis.

Result:
We got 69% accuracy H-W (i.e. hand vs word) class pair.

Discussion:
MEG is a powerful neuroimaging technique that enables the study of brain activity with high temporal resolution. However, the non-stationarity of brain signals poses challenges in data analysis. By employing appropriate methods to address non-stationarity and using deep learning we can adapt to these non-stationary changes.

Recalling our past experiences, autobiographical memories, is a reconstructive process that involves many cognitive skills. Previous research has implicated a set of brain regions in autobiographical memory retrieval, including the hippocampus and medial prefrontal cortex (mPFC). Here we investigated whether a similar network of brain areas was involved in the retrieval of memories that were formed in a naturalistic virtual reality (VR) environment. Specifically, we created a VR town through which participants (N=12; 6 female; aged 23-34) virtually navigated using a walk-in-place method, within the confines of a magnetically shielded room. This involved a range of different naturalistic experiences – see the Alexander et al. poster on encoding. One day later they returned to recall autobiographical memories of the virtual experiences which involved a period of silent retrieval of each memory, followed by recollection of each memory out loud. Neuromagnetic fields were recorded throughout encoding and retrieval using a whole-head OPM-MEG system (120-142 channels). First, we will show that participants recalled the virtual experiences similarly to real-world autobiographical memories, with good accuracy and vividness. Next, we will present preliminary analyses of whole brain oscillatory power changes in the theta (4-8 Hz) and alpha (8-12 Hz) bands during silent memory retrieval versus a control task. In addition, we will show time-frequency spectrograms from regions of interest including the hippocampus and mPFC. Finally, we will share preliminary data from the out loud recall period to assess whether memory-related neuromagnetic fields can be identified despite the artefacts from speech production. Overall this retrieval study represents an important step forwards in showing that VR is a proxy for actual lived experiences which, when combined with OPM-MEG, opens up a host of opportunities for future naturalistic studies across a range of neurocognitive domains.

Introduction:
The basal ganglia and their interactions with the cortex play an important role in movement, as shown by animal- and non-invasive human neuroimaging studies. In humans, one can investigate cortico- basal ganglia interactions when patients undergo deep brain stimulation (DBS) surgery, as researchers can directly record local field potentials (LFP) from the DBS target in the basal ganglia. Combining LFP recordings with MEG or EEG affords the characterization of large-scale brain networks associated with movement. Previous studies mostly focused on MEG due to the presence of externalised DBS wires which hinder the placement of an EEG cap. However, MEG limits the amount of movement and is heavily affected by DBS artefacts. The new generation of DBS technology, the Percept PC from Medtronic, offers wireless LFP recording in patients with chronically implanted stimulators, opening new possibilities for movement neuroscience.

Methods:
We measured simultaneous EEG and internal Globus Pallidus LFPs in 8 patients with dystonia on and off DBS, during several tasks. Our movement tests were passive and active movement, walking, writing, pouring, hand posing and speaking. We also included non-movement tasks such as resting- state, sensory stimulation and sensory tricks, a classic hallmark of dystonia. Simultaneously, we acquired movement tracking data.

Results:
We developed a pipeline for data synchronisation and analysis and identified and solved several of the challenges involved in recording from the new Medtronic device. Our analysis of the resting state shows a significant EEG-pallidal coherence in the theta and alpha band which is significantly suppressed by DBS in the posterior parietal area. During movement, significant EEG-pallidal coherence is observed which is suppressed by DBS in the beta band, and increased in the low gamma band.

Conclusion:
Our results are broadly consistent with the previously reported suppression of pallidal low-frequency activity by DBS in dystonia. Our study has several limitations, among which are the small population and heterogeneity of the patient group, and the short withdrawal time between the DBS on and off conditions. Nevertheless, we believe our results offer important and novel insights into the cortico- basal ganglia interactions in dystonia.

Our life experiences are encoded in autobiographical memories. To date, neuroimaging research has mostly focussed on the retrieval of autobiographical memories, because examining their formation is precluded in traditional head-immobilising brain scanners such as MRI and SQUID MEG. Moreover, in the absence of information from the time of encoding, the accuracy of retrieval could not be reliably assessed in most extant studies. Here we created a virtual reality environment through which participants (N=12; 6 female; aged 23-34) moved using a walk-in-place method while in a magnetically shielded room. During their tour of this virtual town, they had a range of different naturalistic experiences, such as visiting a museum or observing street musicians, that varied in duration and content. Throughout the town tour we collected neural data using a ~130 channel OPM-based MEG system. The next day, and also during OPM-MEG, participants were asked to recall and describe their memories of experiences from the town tour – see the Seymour et al. poster for further details on retrieval. We first applied our recently-established pre-processing pipeline for ambulatory OPM-MEG which allowed us to reliably identify brain signals during large movements (~50 cm) caused by walking-in-place and looking around the virtual town. We then undertook, and will present, a preliminary characterisation of the neural signatures associated with forming naturalistic memories, including those that go on to be remembered or degraded (where now the ground truth is known). Based on the extant literature, we also examined the boundaries, particularly those at the end, of experiences. Our main focus was on theta activity (4-7 Hz) in the hippocampus and medial prefrontal cortex. Overall our study shows that it is possible to leverage the combination of virtual reality and OPM-MEG to examine how the brain functions in something more akin to everyday life which, until now, has eluded detailed scrutiny.

Background:
Short-term memory (STM) sustains information temporarily, while working memory (WM) is defined as goal-directed manipulation of this maintained information. The neural background of memory-related processes has been widely studied but many previous studies of visual WM maintenance have operationalized it in a way that is more consistent with STM, i.e., maintenance of information over a short delay, with no additional processing demands. These studies found both local and interareal oscillatory changes in several frequency ranges during the maintenance period.

Objectives:
Our objective was to identify oscillatory mechanisms dissociating WM and STM, utilizing a task where participants had to either mentally manipulate the maintained information (WM) or not (STM). We hypothesized that increased top-down modulation in the WM condition would be associated with increased theta (4-8 Hz) and alpha (8-14 Hz) activity.

Methods:
We recorded brain activity with simultaneous MEG-EEG during a retro-cued delayed- match-to-sample memory task from 30 healthy young adults (ongoing, target n = 40). Participants were retro-cued to maintain either all probe stimuli (shapes and gratings) during a delay period (STM condition) or only stimuli belonging to one category (shapes or gratings), while ignoring non-cued stimuli (WM condition).

Results:
Source localized oscillatory power in the high-frequency alpha (10-14 Hz) band in parieto- occipital regions was increased during the maintenance (post-retrocue) period for WM compared to STM trials, while prefrontal theta power was decreased. Preliminary analyses of interareal synchrony revealed hubs differentiating WM and STM trials in similar regions and frequency ranges.

Discussion:
Our results suggest that high-alpha band oscillations reflect top-down selective attention acting on internal mental representations during maintenance, while frontal theta might be linked to increased resource allocation in conditions with higher load.

Background and Aims:
When planning for epilepsy surgery, multiple potential sites for resection may be identified through anatomical imaging. The likely seizure onset zone is frequently confirmed using intracranial EEG. Non-invasive magnetoencephalography using optically pumped sensors (OP-MEG) could potentially refine or replace this invasive recording. Here, we test the utility of a-priori information from anatomically identified potential lesion sites in simulation.

Methods:
We simulated OP-MEG recordings for 1309 potential lesion sites identified from anatomical images in the Multi-centre Epilepsy Lesion Detection (MELD) project. To localise the simulated data, we used the empirical Bayesian beamformer (EBB), as well as a restricted version of Multiple Sparse Priors (MSP) with only the patient’s potential lesion locations as priors. We added errors to both the sensors (e.g. gain and co-registration) and the source model (e.g. lesion extent).

Results:
Knowledge of the candidate lesion zones made the inversion extremely robust to random errors in sensor gain, orientation and even location. For sensor position, orientation and gain errors of 5 mm, 10⁰ and 5%, 98% of the correct sites were identified. By contrast, imprecise source models undermined the utility of the a-priori information. When the edge of the lesion was simulated as the epileptogenic source but the model prior assumed that only the centre of the lesion was active, accuracy dropped to 60% (chance level 18%).

Discussion and Conclusions:
Anatomical lesion mapping data could be used in conjunction with flexible OP-MEG helmets to overcome limitations due to imprecise array geometry. However, consideration should be given to the source modelling assumptions, which can have a considerable impact on reconstruction accuracy.

The measurement of electrophysiology is of critical importance to our understanding of brain function. However, current non-invasive measurements, including electroencephalography (EEG) and magnetoencephalography (MEG) have limited sensitivity, particularly compared to invasive recordings. Optically-pumped magnetometers (OPMs) are a new type of magnetic field sensor which ostensibly promise MEG systems with higher sensitivity; however, the noise floor of current sensors remains high compared to cryogenic instrumentation and this has proven limiting. Here, we question how sensor array design affects sensitivity, and whether judicious sensor placement could compensate for the higher noise floor. Through theoretical analyses, simulations, and experiments we show that increasing the total signal measured by an OPM array – either by increasing the number of sensors and channels, or by changing the placement of those sensors – affords a linearly proportional increase in signal-to-noise ratio (SNR). Our experimental measurements confirm this finding, showing that by changing sensor locations in a 90-channel array, we could increase the SNR of visual gamma oscillations from 4.8 to 10.5. Using a 180-channel OPM-array, we capture broadband gamma oscillations induced by a naturalistic visual paradigm, with an SNR of 3; a value that compares favourably to similar measures made using conventional MEG. Our findings add to the growing argument that OPMs are the sensor of choice for MEG system construction. They are also important for the design of future OPM based instrumentation, and most importantly, they show how non-invasive imaging technologies can be optimised to provide non-invasive measurements of human brain electrophysiology with the highest possible sensitivity.

Background/aims:
Optically pumped magnetometers (OPMs) have revolutionised neuroimaging, enabling a new generation of MEG system which allows movement during scanning. However, most existing OPM-arrays require complex cabling to control each OPM, making ambulatory motion challenging. Here we trial a new OPM array design with miniaturised control and data acquisition electronics integrated into a wearable “backpack”. We compare system performance to an established OPM-MEG system and determine its viability for MEG measurements.

Methods:
The system comprised 64 triaxial OPMs (QuSpin) (192 independent MEG channels) mounted in a 3D printed helmet. The OPMs are connected to a miniaturised digital control electronics and data acquisition system (QuSpin) housed inside a backpack. The electronics can be positioned a maximum of 80 cm from the OPMs. Firstly, noise recordings were taken in an empty magnetically shielded room. We contrasted this with an established system comprising 57 triaxial OPMs each controlled by its own electronics. Then, we measured MEG signals from an individual performing a button press task. Data were processed by a beamformer to generate functional images depicting the spatial signature of beta modulation.

Results:
For both systems, the noise floors were 15 fT/√(Hz) or lower. The presence of the system electronics in close proximity to the sensors had little effect on system noise. Beta modulation induced by a button press was seen, with primary effects in sensorimotor cortex and the expected oscillatory signature (movement induced beta desynchronisation followed by a post movement rebound) clearly delineated.

Discussion:
Our results show clearly that the new system is capable of collecting high quality MEG data. Minimised cabling means new opportunities for ambulatory experimentation and the removal of long cabling to electronics outside the MSR means the system is less vulnerable to electronic interference.

Background/Aims:
Enhanced gamma activity (30-100Hz) coincides with successful episodic memory retrieval, but it remains unknown whether this oscillatory activity is a cause or a consequence of the retrieval process. We aim to address this question of causality.

Methods: 70 human participants completed a paired associates memory task whilst undergoing sensory stimulation (at 65Hz, 43.3Hz and 32.5Hz) during memory retrieval. To understand the neural effects of stimulation, we built pyramidal-interneuronal network gamma (PING) models and stimulated them using the same protocol as the behavioural task. Two ongoing MEG and EEG studies will aim to identify the neural locus of these behavioural and computational effects.

Results:
Both 65Hz and 32.5Hz sensory stimulation enhanced memory recall above a baseline condition where no sensory stimulation was applied. Only a small proportion of participants (~10%) could perceive the 65Hz visual flicker, suggesting 65Hz sensory stimulation is imperceptible. The behavioural results could be reproduced by stimulating a PING model with an endogenous ~32Hz oscillation, but not in a PING model with an endogenous ~65Hz oscillation, suggesting 65Hz sensory stimulation enhances recall by harmonically entraining an endogenous ~32Hz oscillation. We anticipate that the M/EEG results with verify this “harmonic entrainment”, with 65Hz and 32.5Hz sensory stimulation both enhancing visual cortical “slow” gamma (~32.5Hz) activity.

Discussion:
These results suggest imperceptible sensory stimulation enhances recall, providing a novel and entirely unintrusive means of tackling mnemonic issues. Furthermore, these results show that harmonic entrainment can impact behaviour, highlighting the non-linear interactions between exogenous stimulation and endogenous neural activity. Lastly, if the M/EEG results complement these findings, we would propose that “slow” gamma oscillations play a causal role in episodic memory retrieval.

Background/Aims:
“mTBI-predict” is a large, multi-site, multi-vendor, longitudinal prospective cohort study on patients with mild traumatic brain injury (mTBI) which aims to evaluate the accuracy and precision of prognostic biomarkers. As a precursor, we will undertake a baseline variability study to assess the reproducibility of our data collection and analysis methods.

Methods:
The study involves 3 sites. 20 controls will complete 6 scanning days (4 at one site, 1 at each of the other sites) and 20 mTBI patients complete 4 scanning days (1 site only). Half of the subjects are civilians and half military. On each day, participants complete MEG and MRI scans. The MEG session involves two 5-minute resting-state scans and 3 tasks: choice reaction task (CRT), spatial attention task (covert attention) and implicit face viewing task (happy, neutral, or angry faces).

Analysis:
Control data will be used to characterise how much pre-defined primary outcomes, outlined here, vary across sites and sessions. Spatial attention data will be used to assess hemispheric lateralisation (cue-target alpha power and target gamma power) as a diagnostic marker of mTBI-related attention problems. Greater induced gamma (60-90Hz) power associated with angry face stimuli has been found in PTSD, so this measure will be used to identify patients with PTSD symptoms. Resting-state delta (2-8Hz) power and functional connectivity will be examined as a diagnostic marker of mTBI. CRT data will be used to compare measures from different modalities (e.g., fMRI).

Implications:
This study will enable quantitative assessment of candidate biomarker variability across scanning sessions and sites. Achieving high reproducibility and repeatability in brain metrics across centres will determine whether these measures can be used for the main project. This large, cross-site MEG study will provide valuable lessons about optimising high-intensity cross-site working to achieve high statistical power in MEG research.

Background/Aims:
Doubt is typically considered a hallmark of obsessive-compulsive disorder (OCD), yet its mechanistic account remains unclear. Here we aimed to investigate individual differences in the components of information gathering and their neural correlates.

Methods:
We collected MEG data from a sample including OCD patients and high compulsive non-patients (N = 113) using a newly developed information gathering task, where participants indicated which of two possible stimuli was more plentiful. Subjects had the choice to either continue sampling additional information or stop and make a binary choice between the two stimuli. We experimentally varied the amount of current and past information, the maximum sampling time before making a decision and the evidence strength to assess their impact on information gathering and the relationship with OCD traits. We used MEG decoding to analyse the time course of these experimental factors and investigated how the underlying activation patterns (Haufe et al., 2014) related to OC traits. Behavioural analyses will be verified in a large-scale online sample completed the Obsessive-Compulsive Inventory-Revised (N = 4375).

Results:
In our behavioural analysis, we found that individuals with higher OC traits were less sensitive to the difference in evidence for the two stimuli when making a decision. When analysing the MEG data, we find a timely separation of all experimental factors, suggesting a cascade in information processing. We also find altered decodability in people with high OCD traits, driven by altered late activation patterns in centroparietal areas.

Discussion/conclusions/implications:
Together, this suggests that indecisiveness across the OC spectrum stems from differences in evidence sensitivity and thus belief updating during information search. Our new task provides a deeper understanding of the underlying mechanisms, which could enable more targeted interventions in future.

Kurtosis beamforming of MEG data can be used to spatially locate the epileptogenic zone (Hall et al., 2018) in patients with epilepsy. Here, we present and evaluate a method for enhanced semi-automatic detection and characterization of these zones, based on moving-window kurtosis and spatial clustering in atlas space, based on this kurtosis timeseries, using non-negative matrix factorization (NNMF). Here we demonstrate the approach on pre-surgical MEG data collected from 4 patients scheduled for inter-cranial EEG and subsequent re-section. LCMV beamformer of 10-minutes of resting-state data was performed at 20-70 Hz, the standard range for kurtosis spike estimation (Ishii et al., 2008; Kirsch et al., 2006), on a 2x2x2 mm grid throughout the brain. Excess kurtosis was calculated for each of these grid locations, which is commonly used to visualise candidate epileptogenic regions. In our proposed method, for computational tractability, brain regions-of-interest (ROIs) were first defined using a variant of the standard AAL90 atlas, in which the larger atlas regions have been sub- divided, resulting in 174 ROIs. Within each ROI, the voxel containing the maximum kurtosis was determined and the beamformer virtual-sensor timecourse estimated for this grid location. The next stage involved spike detection and characterization. It was expected that true interictal epileptic activity would be represented as brief, high-kurtosis events in this timeseries. For each of the 174 AAL regions, a moving window kurtosis timeseries was calculated using a 1s window and used to identify brief events in the timeseries. This allowed us to identify those ROIs with the most spikes and confirm their likely epileptiform status via visualization of the event timecourses. In the final stage, we used NNMF to group the 174 ROIs into spatial clusters based on their moving- window kurtosis timeseries. The resulting output was a spatial connectivity map of each cluster and accompanying representative timeseries, revealing the presence of different sub-networks with independent spiking profiles. The new method therefore allows both easier segregation of true, versus artefactual, epileptiform spiking networks, and the possible identification of multiple, and clinically relevant, epileptiform networks in some patients.

References
Hall, M. B. H., Nissen, I. A., van Straaten, E. C. W., Furlong, P. L., Witton, C., Foley, E., Seri, S., & Hillebrand, A. (2018). An evaluation of kurtosis beamforming in magnetoencephalography to localize the epileptogenic zone in drug resistant epilepsy patients. Clinical Neurophysiology, 129(6), 1221–1229. https://doi.org/10.1016/j.clinph.2017.12.040
Ishii, R., Canuet, L., Ochi, A., Xiang, J., Imai, K., Chan, D., Iwase, M., Takeda, M., Snead, O. C., & Otsubo, H. (2008). Spatially filtered magnetoencephalography compared with electrocorticography to identify intrinsically epileptogenic focal cortical dysplasia. Epilepsy Research, 81(2), 228–232. https://doi.org/10.1016/j.eplepsyres.2008.06.006
Kirsch, H. E., Robinson, S. E., Mantle, M., & Nagarajan, S. (2006). Automated localization of magnetoencephalographic interictal spikes by adaptive spatial filtering. Clinical Neurophysiology, 117(10), 2264–2271. https://doi.org/10.1016/j.clinph.2006.06.708

Background:
Traditionally, the brain and spinal cord have been studied as separate systems due to the challenges of simultaneously imaging their activity. However, optically pumped magnetometer (OPM)-based imaging is uniquely versatile, allowing flexible sensor placement on different parts of the body. Using OPMs, we have developed a novel system for concurrent imaging of brain and spinal cord activity. Our focus in this work is to explore the endogenous interactions between the brain, spinal cord, and muscles involved in sensorimotor control.

Methods:
Healthy participants performed simple voluntary movements with their right and left hands. We recorded brain and spinal cord activity using OPMs placed on the head and neck, while also recording electromyography (EMG) data from the thumb abductor muscle. The data were then reconstructed in source space, and canonical variate analysis was applied to identify maximally correlated components of brain-spinal cord and spinal cord-EMG activity.

Results:
Our results provide evidence for both linear and nonlinear oscillatory interactions between the brain, spinal cord, and muscles during voluntary movement. These functional connectivity patterns were similar for movements of the right and left hands and were consistent with known features of sensorimotor pathways.

Implications:
This research demonstrates the utility of OPMs in studying endogenous spinal cord activity. Our OPM-based system, allowing concurrent imaging of the brain and spinal cord, opens new possibilities for advancing our understanding of how communication is coordinated in the central nervous system, both in health and disease.

Background:
Parkinson’s Disease (PD) is a neurodegenerative disorder affecting the whole brain, leading to several motor and non-motor symptoms. In the past, it has been shown that PD alters resting state networks (RSN) in the brain. These networks are usually derived from fMRI BOLD signals. This study investigated RSN changes in PD patients based on maximum phase amplitude-coupling (PAC) throughout the cortex. We also tested the hypothesis that levodopa medication shifts network activity back toward a healthy state.

Methods:
We recorded 23 PD patients and 24 healthy age-matched participants for 30 minutes at rest with magnetoencephalography (MEG). PD patients were measured once in the dopaminergic medication ON and once in the medication OFF state. A T1-MRI brain scan was acquired from each participant for source reconstruction. After correcting the data for artifacts and performing source reconstruction using a linearly constrained minimum variance beamformer, we extracted visual, sensorimotor (SMN), and frontal RSNs based on PAC (Florin and Baillet 2015).

Results:
We found significant changes in all networks between healthy participants and PD patients in the medication OFF state. Levodopa had a significant effect on the SMN but not on the other networks. There was no significant change in the optimal PAC coupling frequencies between healthy participants and PD patients.

Discussion/Summary: Our results suggest that RSNs, based on PAC in different parts of the cortex, are altered in PD patients. Furthermore, levodopa significantly affects the SMN, reflecting the clinical alleviation of motor symptoms and leading to a network normalization compared to healthy controls.

Aims:
Our prior knowledge greatly influences how we perceive the world and helps us to create better predictions of future events. Previous studies have shown that the power and phase of alpha oscillations (8-12Hz) modulate perception. Here, we seek to identify whether sensory predictions in visual cortex oscillate and are coupled to specific phases of low frequency oscillations. Establishing a link between perceptual predictions and endogenous oscillations would shed light on the mechanisms by which the brain generates predictions.

Methods:
Healthy volunteers (n=32) performed a shape discrimination task while neuromagnetic activity was measured using MEG. Auditory cues predicted the identity of an upcoming abstract shape with 75/25% trial validity and were orthogonal to the discrimination task.

Results:
Using a time-resolved decoding analysis at the sensor level, we identified the neural representations of the presented abstract shapes. To test whether sensory predictions have an oscillatory nature, we performed a time-frequency analysis on the decoded data in the pre-stimulus period. The decoded prediction templates fluctuated at low frequencies, predominantly in the alpha band (8–12Hz). We further hypothesise that the magnitude of the same decoded prediction templates is modulated by the phase of the alpha oscillations. Phase modulation of the alpha oscillations will be calculated and related to the strength of the decoded predictions in visual cortex. To obtain individual neural representations of the predicted abstract shapes and spatial origin of the alpha oscillations, MEG data will be further analysed with generalised eigendecomposition at sensor and source level. Discussion: Our results indicate that top-down prediction signals are dominated by oscillatory activity in the alpha-band. These findings are potentially in line with alpha oscillations acting as carriers of the neural representations of predicted shapes from downstream regions to visual cortex.

Background:
Previous evidence suggests there are functional connectivity changes across multiple large-scale networks in Dementia with Lewy Bodies (DLB), Parkinson’s Disease Dementia (PDD), and Alzheimer’s Disease (AD) observed in fMRI and EEG. Although MEG offers many advantages over other imaging modalities, very little is known about large-scale networks in dementia, specifically in DLB, derived from MEG data.

Aims:
Hence, we are conducting the multimodal imaging in Lewy body disorders (MILOS) study. The main objective of this study is to investigate combined MEG/EEG “microstates” of functional connectivity in large-scale networks in DLB. We will explore network organisation, characterisation of temporal dynamic patterns, and fast oscillatory activities in both space and time. Furthermore, this study will also study correlations between connectivity alterations and clinical symptoms commonly found in DLB.

Methods:
Patients with Lewy body disorder and healthy controls undergo T1 structural MRI in addition to resting state MEG/EEG for source localisation. The source localised MEG data is first separated into 10-sec-long intervals to allow for computing a dynamic evolution of large-scale networks’ topography. The data is then separated into Delta, Theta, Alpha, Beta, and low and high Gamma frequencies to examine different patterns of functional connectivity that can help identify frequency-specific biomarkers. Then, Independent Component Analysis is used, and the resulting components are then matched with the templates of large-scale networks for each 10 second time window.

Results/Discussion:
Here, we will report preliminary findings by showing MEG/EEG microstates with large-scale networks and expand our understanding of temporal characteristics of functional connectivity in DLB. This study will therefore provide the results of using a different form of methodology to obtain MEG/EEG microstates of functional connectivity than what is currently available.

Background/Aims:
Traditionally, Phase I (first in human) clinical trials have focused on assessing safety and pharmacokinetics (effects of the body on the drug). However, the repeated, hugely expensive late (i.e. phase III) failure of many Central Nervous System (CNS)-targeted drugs has highlighted the need for a radical redesign of Phase I trials. Influential initiatives such as the NIMH Fast-Fail program show the critical need to include in-vivo measures of target engagement to accelerate CNS drug discovery. To address this, we are using pharmaco- MEG and fMRI as indices of target engagement in a Phase I clinical trial of a novel positive allosteric modulator (PAM) of the AMPA receptor developed at Cardiff University. AMPA PAMs increase the conductance of AMPA and co-localised NMDA receptors, and are proposed to improve cognition via an increase in synaptic plasticity, particularly in conditions such as schizophrenia where NMDA receptor function is impaired. Schizophrenia affects ~1 % of the global population and though positive symptoms (e.g., hallucinations) can be relatively well managed, there are currently no treatments for the cognitive impairments that severely impair quality of life. Studies targeting different receptors, including AMPARs, show that MEG can provide sensitive and time-resolved markers of pharmacological action.

Methods:
Here, MEG will be recorded from healthy participants for the following paradigms: 1) resting-state 2) visual-motor gamma 3) auditory oddball 4) 40 Hz-steady-state auditory. Both evoked and induced responses will be analysed. Resting-state MEG will be analysed using frequency specific measures of both oscillatory amplitude and connectivity. In addition to MEG, we will collect resting state fMRI, structural MRI and a battery of cognitive tests. The study is placebo-controlled and double blinded. MEG/MRI study design exploits the 14-day dosing part of the trial (a key test of safety) and an additional 3-way crossover study will explore neural effects from a single dose (placebo, lower- and higher-dose).

Discussion/conclusions/implications:
To our knowledge, this is the first time MEG has been used at this stage of drug development. Data collection is ongoing, but the research demonstrates the possibilities for MEG to be integrated into the intensive and highly regulated Phase 1 clinical trial environment.

Background/Aims: 
Many tasks require the skilled interaction of both hands, such as eating with a knife and fork or keyboard typing. However, our understanding of the behavioural and neurophysiological mechanisms underpinning bimanual motor learning is still sparse. Here, we aimed to address this by characterising learning-related changes of different levels of bimanual interaction.

Methods: 
To explore early bimanual motor learning, we designed a bimanual motor learning task. In the task, a force grip device held in each hand (controlling x- and y-axis separately) was used to move a cursor along a path of streets at different angles (0°, 22.5°, 45°, 67.5°, and 90°). Each street corresponded to specific force ratios between hands, which resulted in different levels of hand interaction, i.e., unimanual (Uni, i.e., 0°, 90°), bimanual with equal force (Bieq, 45°), and bimanual with unequal force (Biuneq 22.5°, 67.5°). 42 healthy participants performed the task for 100 trials, whereby each trial comprises six streets.

Results: 
On the behavioural level, the three conditions differed in their movement time and error. Further, we found that the novel task-induced improvements in movement time and error, with no trade-off between movement time and error, and with distinct patterns for the three levels of bimanual interaction. On the neural level, beta event-related desynchronisation and synchronisation show different spatial topographies for the unimanual conditions. Moreover, we found that learning-related changes are manifold and are explored in detail using the amplitude-envelope variant of the Hidden Markov Model.

Discussion/conclusions/implications:
Overall, this complex bimanual motor task allows us to characterise bimanual motor learning with different levels of bimanual interaction. This should pave the way for future neuroimaging studies to further investigate the underlying mechanism of bimanual motor learning.

Recent studies suggest that spontaneous cortical activity underlies numerous processes (e.g., disease, demographics, cognition), albeit the neurochemical bases of spectral changes at rest are not well characterized. For example, the widespread action of acetylcholine (ACh) on pyramidal cell function implicates its role in a myriad of behavioral states, including visual-motor processing and attention. Such behavioral modulation is associated with increases in task-relevant low and high frequency oscillations in model systems. However, the role of ACh in modulating resting neural dynamics and behavior is unknown.
In this study, 43 healthy men completed 5 minutes of eyes open rest during MEG following administration of an ACh muscarinic receptor antagonist, biperiden, or placebo. MEG data were imaged in the time-frequency domain using a beamformer to examine drug-dependent effects of ACh on large-scale resting networks across the canonical frequency bands. Peak vertex time series were extracted from drug-dependent statistical maps and related to behavior assessed outside the scanner.
Our results revealed spectrally and spatially-resolved changes in cortical oscillations following biperiden administration. Specifically, we observed elevations in low (~2-7 Hz) and high-frequency (>30 Hz) cortical activity in key hubs of visual, motor, and dorsal attention networks following biperiden administration, while the opposite trajectory was observed for resting alpha power. Moreover, we observed a drug-dependent modulation of better behavioral performance during basic motor, learning and decision making tasks by resting state networks resonating at lower (<4 Hz) and higher (>30 Hz) frequencies.
En masse, these data suggest that the disruption of cholinergic transmission leads to widespread changes in spontaneous cortical rhythms, which may relate to compensatory changes in domain-specific cognitive function observed under deviant neurochemical conditions in humans.

Since early development, the brain synchronizes to external rhythms, such as speech. To date, little is known about the role of acoustic experience in the development of neural tracking of speech. Cochlear-implanted (CI) children, having experienced a period of auditory deprivation, provide the opportunity to fill this gap.
By applying an encoding model on electrophysiological data, we measured the neural tracking of speech envelope and spectrogram in 3-18 years old hearing controls (HC; N=37) and bilateral profound deaf children who received CIs (N=32). CI group was equally divided into children with congenital deafness (CD) and delayed deafness (DD) onsets to estimate the specific role of auditory deprivation in the first phase of life.
Results revealed a defined auditory response function and a developmental trajectory in HC and CI children. However, neural tracking of speech envelope measured in CI significantly differed from HC. The earliest auditory response was delayed in CI, and the subsequent neural tracking was reduced. While no difference emerged between CD and DD, we found that the latency of the earliest auditory response in CI was associated with the age of cochlear implantation. Behavioural data showed a significant speech comprehension deficit in CI children. Importantly, CI’s neural tracking dynamic was significantly reduced at the latency [150-250ms] correlated with comprehension scores in HC children. Finally, multivariate encoding of the speech spectrogram unveiled the role of the low-frequency range in CI’s neural tracking alterations.
These findings revealed that neural tracking of speech emerges regardless of the absence of acoustic experience in the first phases of life. However, it is delayed and hampered in CI children. Although early implantation mitigates the delayed auditory response, alterations at higher stages of speech processing remain and might account for CI’s comprehension deficits.

Background:
Steady fixation is considered a necessary requirement in cognitive experiments that involve visual stimuli. It is the absence of macroscopic oculomotor events that is thought to ensure the validity of the interpretation of the acquired data in relation to the cognitive construct studied. That is, a putative relationship between eye movement control and the brain’s response to visual stimulation is not considered per definition. The present report explores the extent to which visual signals may be related to eye movements through an analysis of cortico-ocular coherence, akin to established observations in the motor system.

Methods:
We re-examined simultaneously acquired magnetoencephalographic and eye tracking data in the context of an inward moving grating experiment. During visual stimulation, gaze was directed towards central fixation – as instructed. We processed the horizontal eye gaze position data in order to be able to use it as a proxy for the EMG of the extraocular muscles. We first subtracted a median filtered version of the eye tracker signal, to remove saccades. This was followed by high pass filtering (cutoff frequency at 40 Hz, windowed sync FIR filter), and rectification.

Results:
Sensor level coherence analysis revealed two spectral peaks in the delta/theta frequency range (2-7 Hz), and alpha/beta frequency range (10-16 Hz). Source localization identified involvement of bilateral early visual cortical areas, bilateral cerebellum and possibly the superior colliculus.

Conclusion:
The results are discussed in light of the conjecture that coherence between subsaccadic movements and cortical rhythms is a manifestation of an efferent oculomotor process supporting active vision.