Localized estimation of event-related neural source activity from simultaneous MEG-EEG with a recurrent neural network.

Estimating intracranial current sources underlying the electromagnetic signals observed from extracranial sensors is a perennial challenge in non-invasive neuroimaging. Established solutions to this inverse problem treat time samples independently without considering the temporal dynamics of event-related brain processes. This paper describes current source estimation from simultaneously recorded magneto- and electro-encephalography (MEEG) using a recurrent neural network (RNN) that learns sequential relationships from neural data. The RNN was trained in two phases: (1) pre-training and (2) transfer learning with L1 regularization applied to the source estimation layer. Performance of using scaled labels derived from MEEG, magnetoencephalography (MEG), or electroencephalography (EEG) were compared, as were results from volumetric source space with free dipole orientation and surface source space with fixed dipole orientation. Exact low-resolution electromagnetic tomography (eLORETA) and mixed-norm L1/L2 (MxNE) source estimation methods were also applied to these data for comparison with the RNN method. The RNN approach outperformed other methods in terms of output signal-to-noise ratio, correlation and mean-squared error metrics evaluated against reference event-related field (ERF) and event-related potential (ERP) waveforms. Using MEEG labels with fixed-orientation surface sources produced the most consistent estimates. To estimate sources of ERF and ERP waveforms, the RNN generates temporal dynamics within its internal computational units, driven by sequential structure in neural data used as training labels. It thus provides a data-driven model of computational transformations from psychophysiological events into corresponding event-related neural signals, which is unique among MEEG source reconstruction solutions.

Neurodesk: an accessible, flexible and portable data analysis environment for reproducible neuroimaging.

Neuroimaging research requires purpose-built analysis software, which is challenging to install and may produce different results across computing environments. The community-oriented, open-source Neurodesk platform ( https://www.neurodesk.org/ ) harnesses a comprehensive and growing suite of neuroimaging software containers. Neurodesk includes a browser-accessible virtual desktop, command-line interface and computational notebook compatibility, allowing for accessible, flexible, portable and fully reproducible neuroimaging analysis on personal workstations, high-performance computers and the cloud.

Localized estimation of electromagnetic sources underlying event-related fields using recurrent neural networks.

Objective. To use a recurrent neural network (RNN) to reconstruct neural activity responsible for generating noninvasively measured electromagnetic signals.Approach. Output weights of an RNN were fixed as the lead field matrix from volumetric source space computed using the boundary element method with co-registered structural magnetic resonance images and magnetoencephalography (MEG). Initially, the network was trained to minimise mean-squared-error loss between its outputs and MEG signals, causing activations in the penultimate layer to converge towards putative neural source activations. Subsequently, L1 regularisation was applied to the final hidden layer, and the model was fine-tuned, causing it to favour more focused activations. Estimated source signals were then obtained from the outputs of the last hidden layer. We developed and validated this approach with simulations before applying it to real MEG data, comparing performance with beamformers, minimum-norm estimate, and mixed-norm estimate source reconstruction methods.Main results. The proposed RNN method had higher output signal-to-noise ratios and comparable correlation and error between estimated and simulated sources. Reconstructed MEG signals were also equal or superior to the other methods regarding their similarity to ground-truth. When applied to MEG data recorded during an auditory roving oddball experiment, source signals estimated with the RNN were generally biophysically plausible and consistent with expectations from the literature.Significance. This work builds on recent developments of RNNs for modelling event-related neural responses by incorporating biophysical constraints from the forward model, thus taking a significant step towards greater biological realism and introducing the possibility of exploring how input manipulations may influence localised neural activity.

Neurodesk: An accessible, flexible, and portable data analysis environment for reproducible neuroimaging.

Neuroimaging data analysis often requires purpose-built software, which can be challenging to install and may produce different results across computing environments. Beyond being a roadblock to neuroscientists, these issues of accessibility and portability can hamper the reproducibility of neuroimaging data analysis pipelines. Here, we introduce the Neurodesk platform, which harnesses software containers to support a comprehensive and growing suite of neuroimaging software (https://www.neurodesk.org/). Neurodesk includes a browser-accessible virtual desktop environment and a command line interface, mediating access to containerized neuroimaging software libraries on various computing platforms, including personal and high-performance computers, cloud computing and Jupyter Notebooks. This community-oriented, open-source platform enables a paradigm shift for neuroimaging data analysis, allowing for accessible, flexible, fully reproducible, and portable data analysis pipelines.

Natural vs forced language switching: Free selection and consistent language use eliminate significant performance costs and cognitive demands in the brain.

Bilinguals are known to switch language spontaneously in everyday conversations, even if there are no external requirements to do so. However, in the laboratory setting, language control is often investigated using forced switching tasks, which result in significant performance costs. The present study assessed whether switching would be less costly when performed in a more natural fashion, and what factors might account for this. Mandarin-English bilinguals engaged in language switching under three different contexts with varied task demands. We examined two factors which may be characteristic of natural switching: (i) freedom of language selection; (ii) consistency of language used to name each item. Participants' brain activities were recorded using magnetoencephalography (MEG), along with behavioural measures of reaction speed and accuracy. The natural context (with both free selection and consistent language use for each item) produced better performance overall, showing reduced mixing cost and no significant switch cost. The neural effect of language mixing was also reversed in this context, suggesting that freely mixing two languages was easier than staying in a single language. Further, while switching in the forced context elicited increased brain activity in the right inferior frontal gyrus, this switch effect disappeared when the language used to name each item was consistent. Together, these findings demonstrate that the two factors above conjointly contribute to eliminating significant performance costs and cognitive demands associated with language switching and mixing. Such evidence aligns with lexical selection models which do not assume bilingual production to be inherently effortful.

Whole-Language and Item-Specific Inhibition in Bilingual Language Switching: The Role of Domain-General Inhibitory Control.

A prominent theory of bilingual speech production holds that appropriate language selection is achieved via inhibitory control. Such inhibition may operate on the whole-language and/or item-specific level. In this study, we examined these two levels of control in parallel, by introducing a novel element into the traditional cued language switching paradigm: half of the stimuli were univalent (each required naming in the same language every time it appeared), and the other half were bivalent (each required naming in different languages on different trials). Contrasting switch and stay trials provided an index for whole-language inhibition, while contrasting bivalent and univalent stimuli provided an index for item-specific inhibition. We then investigated the involvement of domain-general brain mechanisms in these two levels of language control. Neuroimaging studies report activation of the pre-supplementary motor area (pre-SMA), a key region in the executive control brain network, during language switching tasks. However, it is unclear whether or not the pre-SMA plays a causal role in language control, and at which level it exerts control. Using repetitive transcranial magnetic stimulation (TMS) to transiently disrupt the pre-SMA, we observed an essential role of this brain region in general speech execution, while evidence for its specific involvement in each level of inhibition remains inconclusive.

Neuro-dynamics of executive control in bilingual language switching: An MEG study.

Bilinguals have a remarkable ability to juggle two languages. A central question in the field is concerned with the control mechanisms that enable bilinguals to switch language with ease. Theoretical models and neuroimaging evidence suggest that a range of control processes are at play during language switching, and their underlying neural mechanisms are closely related to executive function. What remains unclear is when these control processes are engaged in language switching. In this study, we used magnetoencephalography (MEG) to examine the brain activity while unbalanced Mandarin-English bilinguals performed a digit-naming task with cued language switching. Following presentation of the language cue, an asymmetrical switch effect was observed in the left inferior frontal gyrus (IFG), where switch-related increase in evoked brain activity was larger for switching into the non-dominant language. Following presentation of the naming target, evoked brain activity in the right IFG was larger when naming was required in the non-dominant language compared to the dominant language. We conclude that control processes take place in two stages during language switching, with the left IFG resolving interference following cue presentation and the right IFG inhibiting competing labels following target presentation.