Programme

Abstract:
All animals including humans have to cope with immediate threat to survive and reproduce. Ample evidence shows that non-human animals behave in sophisticated and apparently goal-directed ways. Rapid decisions between these actions, without much leeway for cognitive or motor errors, poses a formidable computational problem. In my talk, I will give an overview of our research that aims to elucidate the neural mechanisms of these decisions. First, our virtual reality (VR) platform allows simulating immediate threat situations in a safe manner. Second, results from a series of behavioural experiments suggest that human behaviour under threat, while following a standard motor sequence, is flexibly adapted and exhibits many characteristics of reflective planning. Third, we developed a VR head mounted display (HMD) that can be used together with optically pumped magnotmeters. Our data suggest that this HMD allows recording meaningful MEG signals across the entire brain. Together, our results pave the way towards an investigation of naturalistic threat-related decisions with MEG.

Co-authors: Yali Pan, Ana Pesquita, Ole Jensen, Katrien Segaert, Hyojin Park
Abstract:
In challenging listening environments, visual cues such as lip movements can enhance speech comprehension. Here, we hypothesise that the external modulation of visual speech signals using non-invasive rhythmic stimulation can harnessed to improve speech understanding. We directly tested the hypothesis using a novel paradigm using Rapid Invisible Frequency Tagging (RIFT). RIFT is a technique that modulates visual stimuli at specific frequencies below the threshold of conscious perception to influence neural processing. We manipulated visual speech signals – using RIFT – to influence the integration of visual and auditory information, and measured brain responses using magnetoencephalography (MEG) alongside speech comprehension performance. 40 participants viewed naturalistic speech videos under dichotic listening conditions. One ear was presented with speech that matched the visual speech information (task relevant) while the other was presented with speech that did not (task irrelevant). Both streams of auditory speech were tagged at 40Hz. The visual flicker (55Hz) was implemented on the area whereby participants derive visual speech information: the speaker’s mouth and was modulated by either task relevant or irrelevant speech amplitude envelopes. When modulated by relevant speech information, RIFT significantly enhanced performance in behavioural measures of speech comprehension. The MEG results showed significant effects of auditory and visual tagging in their respective sensory cortices across all experimental conditions. The visual tagging response was significantly stronger when the amplitude was modulated by relevant speech. This stronger tagging response predicted speech comprehension performance. These results suggest that modulating visual input with relevant auditory speech rhythms can facilitate the excitability of visual cortex perhaps leading to enhanced crossmodal integration. Non-invasive sensory stimulation through RIFT may therefore serve as a promising tool for improving speech intelligibility in complex listening environments with multiple competing speakers, particularly for populations such as older adults, individuals with hearing impairments, or those with auditory processing disorders.

Co-authors: Joseph Gibson1, Ryan M. Hill1,2, Niall Holmes1,2, Lukas Rier1,2, Jessikah Fildes1, Matias Ison3, Alan Kirby, Vishal Shah4, Elena Boto2,1, Richard Bowtell1,2, Matthew J. Brookes1,2 1Sir Peter Mansfield Imaging Centre, School of Physics and Astronomy, University of Nottingham, Nottingham, United Kingdom 2Cerca Magnetics Limited, 7-8 Castlebridge Office Village, Kirtley Drive, Nottingham, United Kingdom 3School of Psychology, University of Nottingham, Nottingham, United Kingdom 4QuSpin Inc. 331 South 104th Street, Suite 130, Louisville, Colorado, USA
Abstract:
Background: Functional neuroimaging seeks to characterise how neural activity underpins cognition. Most modalities require participants to remain still in enclosed spaces, preventing the implementation of naturalistic tasks that accurately depict real-world environments [1]. Here, we exploited a wearable OPM-MEG system to assess motor network dynamics whilst subjects learnt to play a musical instrument; a complex, naturalistic motor learning task that elicits rich brain activity [2]. Methods: 22 participants were scanned twice using a 192-channel OPM-MEG system [3] while playing ‘Twinkle Twinkle Little Star’ on the violin. Before the first scan, they viewed a video of an expert playing the tune; between scans, they received a 25-minute violin lesson from the same expert. Participants played the tune 30 times in both runs. An IR tracking system (Optitrack, NaturalPoint Inc., Corvallis) was used to monitor participant movement and allowed participants to self-pace the trials. Audio data of each trial were also recorded. OPM-optimised Spatiotemporal Signal Space Separation (tSSS) [4] was used to reduce the effects of motion artefacts. MEG data were source-localised using an LCMV beamformer and beta-band (13-30Hz) dynamics were examined. Results: Large head movements (max (140±100)mm translation / (47±50)° rotation) were observed during the task, however tSSS reduced movement related artefacts (from ~0.9nT to ~0.4pT, shielding factor 2250). The expected movement related beta decrease (MRBD) was observed in all participants during playing of the tune and, compared to rest, localised to the sensorimotor regions. We observed a trend for a greater MRBD (p=0.067) following the lesson compared to before the lesson, in the right dorsolateral motor region. Discussion: We successfully collected a rich naturalistic dataset with 22 participants comprising MEG, audio and motion tracking data. We have shown that despite the presence of large motion we are able to reliably measure high-fidelity MEG data, and we have seen preliminary evidence of differences in motor dynamics following the lesson with an expert. Future analyses will combine measures of trial rating to investigate if brain activity changes with success and explore functional connectivity measures. References: [1] Maselli et al. 2023, 10.1016/j.plrev.2023.07.006 [2] Gelding et al. 2019, 10.1038/s41598-019-53260-9 [3] Schofield et al. 2024, 10.1162/imag_a_00283 [4] Holmes et al. 2024, 10.1109/TBME.2024.3465654

Co-authors: Tim M. Tierney, James Osborne, Meaghan E. Spedden, Robert A. Seymour, Nicholas A. Alexander, Cody Doyle, David Bobela, George C. O’Neill, Katarzyna Rudzka, Sahitya Puvvada, Maike Schmidt, Vishal Shah, Gareth R. Barnes
Abstract:
The latest generation of Optically Pumped Magnetometer (OPM)-based Magnetoencephalography (OP-MEG) systems are increasingly portable and lightweight, making the system easier to wear while undertaking large, whole-body translational and rotational movements. This comes at the expense of added interference due to this movement, and necessitates careful consideration of how non-stationary environmental interference will be addressed. To demonstrate the potential of such setups, we recorded OP-MEG from three participants during auditory stimulation, while the participants walked around a magnetically shielded room, moving across an area of at least 1.7 m2 and rotating by at least 180-degrees. All sensor-specific electronics were housed within a wearable backpack, minimising movement restrictions which would otherwise be imposed by sensor cabling. In this experiment, the maximum peak-to-peak range of the observed magnetic field recorded on a single channel was 25.2 nT, considerably beyond the approximately 4 nT range of many previous OP-MEG systems. We show that adaptive multipole modelling (AMM) with the temporal extension can be used to separate the interference and neuromagnetic signals, despite the large degree of movement and close proximity of the sensor electronics. We observed significant auditory evoked responses at the sensor-level, without the need for source localisation approaches for interference suppression. These findings demonstrate the capability of OP-MEG for conducting naturalistic experiments involving movement.

Co-authors: Matias J. Ison, Joaquin Gonzalez, Damian Care, Aditi Jain, Ryan Hill, Joseph Gibson, Paul McGraw, Matthew J. Brookes & Juan E. Kamienkowski
Abstract:
Eye movements are fundamental to a range of daily activities, from reading to driving. Although behavioural patterns of eye movements during real-world tasks are well-characterised, the neural mechanisms supporting these processes remain elusive, partly due to the substantial artifacts they introduce in M-EEG recordings, often eclipsing neural signals. We present a set of experiments combining eye tracking with EEG, MEG and OPM-MEG recordings in various naturalistic tasks involving eye movements. Participants completed a hybrid visual and memory search task (EEG: N=42; MEG: N=21), in which they searched for one of several items held in memory, and a simulated driving task (OPM-MEG: N=10) using an optically pumped magnetometer system. In the EEG study, we demonstrate how deconvolution methods applied to fixation-related potentials (FRPs) can effectively disentangle temporally overlapping neural events, revealing distinct components related to target detection, task progression, and memory load. MEG source reconstruction of fixation-related activity allowed us to identify a visually evoked lambda response originating in primary visual cortex (V1). Target-related activity was observed in a distributed P3m component and further confirmed with a functional connectivity analysis. Time-frequency analyses revealed neural signatures of memory encoding, retention, and visual search. Applying similar strategies to the driving task enabled to characterization of fixation-aligned visual and attentional processing in dynamic, realistic environments. Altogether, these findings allow us to understand how the neural mechanisms underlying the interaction of memory, attention and visual processing emerge in complex dynamic tasks involving eye movements, offering insights toward more ecologically valid models of cognition.

Abstract:
Young children develop in complex, stimulus-rich environments. Understanding the developing brain in childhood requires conducting naturalistic studies in real-world environments. In this talk, we will discuss various approaches to extending neuroscientific research beyond controlled laboratory settings. Specifically, we will present several naturalistic functional near-infrared spectroscopy (fNIRS) studies conducted at Birkbeck’s ToddlerLab. We will describe a set of hyperscanning studies, in which we use wearable fNIRS with multiple participants simultaneously, to investigate the neural correlates of collaborative problem-solving between preschoolers. We will review the optimisation of our analytical pipelines to reduce the influence of physiological noise in naturalistic signal measurement. We will also discuss how techniques such as diffuse optical tomography (DOT) and virtual reality can be used to simulate everyday experiences within controlled lab environments, particularly in studies involving neurodivergent children. Finally, using attention as a case study, we will present new approaches for exploring naturalistic parent–infant interactions with fNIRS technologies. We will reflect on the lessons learned and the challenges encountered when working with different age groups, populations, and experimental settings.

Co-authors: James Dowsett, Inés Martín Muñoz, Paul Taylor
Abstract:
We are developing a new method for generating Steady State Visually Evoked Potentials (SSVEPs) of real-world environments in mobile EEG/MEG with LCD flicker glasses (Dowsett et. al. Journal of Neuroscience Methods, 2020). LCD glasses go dark, like sunglasses, when voltage is applied across the glass. The timing can be accurately controlled, allowing the glasses to flicker at any frequency. We previously demonstrated robust SSVEP responses from single channel EEG whilst participants were walking, overcoming the problems of motion artefacts without requiring high numbers of sensors. Using this method, we can apply a high frequency visual flicker to whatever the participant is looking at in the real-world; unlike traditional SSVEP paradigms which are limited to pictures on screens. In the current study we applied this method to decoding real-world environments. Specifically, when participants stood in 6 unique locations and fixated on a point in the distance, while the glasses flickered at 10 Hz, a unique SSVEP waveform shape was generated. We found that SSVEP responses from real world scenes are surprisingly complex and have distinct shapes: they differ markedly across scenes and participants but are consistent within individuals, even across sessions. This unique SSVEP could be reproduced at a later time and also on a separate day, even if various aspects of the visual scene had changed such as overall luminance or cloud cover. The SSVEPs were significantly unique and consistent that the correct scene could be identified with over 90% accuracy. This decoding works with a single electrode, with any of the electrodes tested, and even with a few second’s of data. The decoding was successful with both 10 Hz, 1 Hz and 40 Hz visual flicker; 40 Hz is particularly promising for future research in naturalistic neuroscience as the visual flicker at this frequency is barely noticeable and would not interfere with normal daily activities. We band-pass filtered the SSVEP at various harmonics to investigate the contribution of different frequency bands to decoding accuracy; for all flicker frequencies tested the SSVEP contained information at harmonics of the flicker frequency, and in all cases the gamma band (40 Hz) contained the maximal amount of information about the visual scene. We propose that this is a highly promising method for naturalistic EEG and MEG.

Co-authors: K. Lee, B. Blanco, R. Cooper, A. Edwards, J. Hebden, K. Pammenter, J. Uchitel, T. Austin
Abstract:
Preterm birth has been associated with cognitive, social, and sleep difficulties later in life, outcomes that may be exacerbated by affected sleep (Gao, 2017, Stangenes, 2017). However, the relationship between sleep states, gestational age (GA), and functional brain development remains poorly understood. Studying the role of protected sleep in the NICU may reveal neuroprotective benefits and improve long-term clinical outcomes. High-Density Diffuse Optical Tomography (HD-DOT), a functional near-infrared spectroscopy (fNIRS) technology, has been used to investigate static resting-state functional connectivity (FC) during active sleep (AS) and quiet sleep (QS) states in term-aged infants (Uchitel, 2023). Dynamic FC analysis investigates time-varying patterns in brain activity to shed light on the non-stationary nature of resting state brain functionality. One method proposed for this objective identifies recurring co-activation patterns (CAPs) using clustering algorithms which capture instantaneous brain configurations (Liu, 2018). This study examines dynamic FC in term and preterm infants during sleep to better understand the functional relationship between sleep states and early brain connectivity. HD-DOT data were acquired from sleeping newborns at the Rosie Hospital, Cambridge UK (term cohort: n = 44, GA = 40+0 weeks (median), 38+1 – 42+1 weeks (range); preterm cohort: n = 26, GA = 35+0 weeks (median), 29+1 – 36+6 weeks (range)). Sleep state was labelled as AS/QS using synchronized video or electroencephalography. Frames were sorted by seed activity for three regions of interest (ROI), frontal, central, and parietal regions, and the top 15% frames were selected for k-means clustering. This threshold was chosen because the average of the top 15% of seed-selected frames strongly correlated with the seed-based correlation maps from the static analysis, validating the CAP procedure (see Figure 1). The clustered frames were averaged to create the CAP maps. Dynamic FC was compared across sleep states by calculating CAP consistency, in-participant fraction, dwell time, and transition likelihood for the term cohort. Additionally, regional bilateral activation was compared across sleep states within each CAP using a two proportion Z-test. The post-clustering analysis has been performed for the term cohort and will be applied to the preterm cohort for comparison.

Co-authors: Christine M Embury, Zelekha Seedat, Kelly St. Pier, Caroline Scott, Friederike Moeller, Krishna Das, Tim Tierney, Gareth Barnes, Matthew Walker, Umesh Vivekananda, J. Helen Cross
Abstract:
Epilepsy impacts more than 100k children in the UK alone. Curative treatments in focal lesional epilepsies are largely dependent on precision mapping of epileptogenic activity coupled with structural imaging to determine surgical targets. Previous studies demonstrate the improved mapping of epileptic activity in magnetoencephalography (MEG) relative to electroencephalography (EEG), but these benefits are likely not realised in those ill-suited for the static, one-size-fits-all set-up of traditional cryogenic MEG. The next generation of the technique, optically-pumped magnetometer (OPM)-MEG promises adaptability and improvement in signal detection by bringing the sensors close to the scalp and arranging them flexibly to better fit all head sizes and shapes, particularly advantageous in children. To examine the capabilities of OPM-MEG in detecting epileptic activity in children with epilepsy, we scanned 12 children, six with focal and six with generalised, for 15 minutes to 1 hour while resting or performing tasks. Our OPM-MEG array consisted of 64 QuSpin dual axis sensors (128 channels) housed in child-sized helmets within a light MuRoom coupled with static active shielding (Cerca Magnetics Ltd., Nottingham, England, UK). Data were pre-processed in BESA Research (version 7.1, BESA GmbH, Gräfelfing, Germany) to reduce the influence of cardiac activity. Spikes were marked by experienced clinical scientists. Spike counts were determined per data file and compiled for each participant ranging from 3 detected spikes to more than 70 in the time they were able to complete in the scanner (up to an hour). For those with focal presentations, equivalent current dipoles were mapped on the half-rise of spikes to determine the ability to precisely delineate epileptic foci from interictal activity detected. Overall, we demonstrate the ability of the technique to detect epileptiform activity in children with focal and generalised epilepsies, and with concordance with clinical presentation. This investigation lays the groundwork for a wider use case for the technique, combining the adaptable set-up and increased tolerability with increased sensitivity and precision of the equipment over available clinical tools. OPM-MEG demonstrates an advantageous potential leap for diagnostic capabilities as well as presurgical workup in paediatric epilepsy.

Co-authors: Joachim Gross, Robin A. A. Ince
Abstract:
Recently, sensory neuroscience has embraced “naturalistic” experimental conditions in which brain activity is recorded during movie watching, natural scene observation or audiobook listening. In combination with modern video-, image- or audio-processing models, this has yielded unprecedentedly predictive stimulus-computable models of brain activity whose validity is hoped to exceed that of models developed from simplistic artificial stimuli. However, the field is now facing a new set of challenges: With stimulus material full of uncontrolled correlations, it remains unclear what features actually cause response variance. The interpretability is further obscured by the complexity of competitive stimulus processing models. Moreover, linear “encoding models” that relate model representations to brain responses are overly flexible, leaving high degrees of freedom such that many different extracted feature spaces predict response variance to the same degree. It becomes difficult to adjudicate between algorithmically diverse hypotheses. Here, we address these challenges with an approach that aims to generalise the performance of a linear encoding model predicting understudied power time courses as recorded with MEG in response to speech listening. Specifically, we find that such encoding models, trained on passive audiobook listening data, fail to generalise to simple and interpretable but out-of-distribution controlled conditions known from the literature. We diagnose this problem to stem from the largely unconstrained and nonlinear phase responses of the encoding models and devise a regularisation penalty to tackle this. While this effectively reduces the degrees of freedom of the encoding models, some of them achieve competitive performance not only on the passive audiobook listening data, but also on the controlled experiment. However, other models, even when constrained, still fail to generalise to the controlled experiment. This highlights how the consideration of a simplistic but controlled experiment points out dispensable model degrees of freedom and affords an improved and interpretable capacity to adjudicate between models that are equiperformant in the naturalistic condition. Taken together, we subscribe to an integrative perspective on sensory neurosciences that attempts to bridge rich naturalistic datasets to controlled experiments, and specifically considers evidence readily available from existing literature.

Abstract:
Measuring detailed information on how people move, see, and think during realistic social situations can be a powerful method in studying social behaviour and cognition. However, measurement-driven research can be limited by the available technology, with bulky equipment and rigid constraints often confining such work to the laboratory, thus limiting the ecological validity of any findings. In this talk, I will discuss some of the studies on live performance I’ve been involved with, using techniques like wearable EEG hyperscanning, eye-tracking, and motion capture. The work aims to explore the use of live performance and theatre as a laboratory for real-world neuroscience, while developing new measurement techniques using wearable sensors.