Biology subjects

Momi, D.

Publications and source records attributed to Momi, D..

18 recordsLinked to original sources

bioRxiv2026

Inferring structural complexity from dynamical measures

Complexity, commonly defined as the coexistence of integration and differentiation, is a fundamental property of thalamocortical networks and is regarded as a prerequisite for adaptive brain function and consciousness. A variety of metrics have been developed to estimate neural complexity from recorded activity and are increasingly employed in both research and clinical settings. Yet, the extent to which these measures reflect specific architectural features of the underlying circuits remains poorly understood. Here, we used Wilson-Cowan mean-field equations to simulate neural dynamics across networks spanning a broad range of architectures. By systematically rewiring connectivity, we generated graded variations in the structural integration-differentiation balance and assessed a panel of complexity measures derived either from spontaneous activity (observational metrics) or from responses to external perturbations (perturbational metrics). Observational measures exhibited markedly heterogeneous behavior: some were primarily sensitive to network integration, others to differentiation, whereas some showed limited sensitivity to architectural manipulations. In contrast, perturbational measures robustly peaked at intermediate rewiring steps and scaled with small-worldness, consistent with architectures that combine local recurrence and distributed communication. Moreover, perturbational metrics proved robust to common-driver confounds and were highest in a wake-like regime near critical instability, declining as sleep-like dynamics became more widespread across the network. Taken together, these findings map widely used empirical complexity measures onto specific network architectures, reveal dynamical factors that can bias observational estimates, and provide guidance for inferring complexity in systems whose internal organization is inaccessible, unconventional, or only partially known, including brain organoids, non-human nervous systems, and synthetic neural networks.

neuroscience↗

bioRxiv2026

Brain-wide reconfiguration of burst firing by psilocybin reveals 5-HT2A-dependent circuit dynamics

Psilocybin produces rapid and lasting therapeutic effects, yet how 5-HT2A receptor activation reshapes brain-wide circuit dynamics during acute drug administration remains poorly understood. Using simultaneous multi-region Neuropixels recordings of 46,360 single units from 35 mice, together with scalp electroencephalography (EEG), pupillometry, and locomotion monitoring, we provide a brain-wide, single-unit and field-potential characterization of psilocybin's acute effects, with pharmacological dissection using the 5-HT2A antagonist ketanserin. Psilocybin selectively reconfigured burst coding, rather than mean firing rate, across cortical, thalamic, and hippocampal circuits: burst firing decreased in hippocampal CA1-CA3 and was bidirectionally modulated in the thalamus, with the reticular nucleus bursting more and first-order geniculate nuclei bursting less. Critically, most of these burst effects were abolished by ketanserin, consistent with at least partial 5-HT2A receptor dependence. These data suggest that the psychedelic state is not simply a matter of how much neurons fire, but of how they fire, pointing to a region-specific, 5-HT2A-associated reconfiguration of burst coding that may underlie the acute phenomenology of the psilocybin experience.

neuroscience↗

bioRxiv2026

A Whole-Brain Dynamical Framework Linking Resting-State Activity to TMS-Evoked Responses

A major challenge in systems neuroscience is understanding how external perturbations interact with ongoing brain activity. Transcranial magnetic stimulation (TMS), increasingly used in both basic and clinical neuroscience and often combined with electroencephalography (EEG), provides a unique opportunity to probe this interaction. However, how intrinsic dynamics constrain the propagation of TMS-evoked activity remains poorly understood. In particular, effective connectivity (EC)--capturing directed, state-dependent interactions between brain regions--is thought to critically shape perturbational spread, yet remains difficult to estimate at the whole-brain EEG level. Here we introduce an analytically tractable, generative whole-brain model that links spontaneous EEG activity to cortical responses under perturbation. By deriving a closed-form expression for the models cross-spectral density, we directly fit empirical resting-state EEG spectra and infer biophysically interpretable local dynamical parameters without time-domain simulations. We then estimate stimulation-site-specific EC using only a small fraction of the TMS-EEG trials. The resulting model accurately predicts the spatiotemporal structure of TMS-evoked potentials (TEPs) in unseen trials. Moreover, even without subject-specific refitting, group-level EC templates capture canonical site-specific propagation motifs underlying single-subject early TMS responses. Together, our results establish an analytical framework for individualized whole-brain modeling of TMS-EEG with potential applicability to model-based neuromodulation.

neuroscience↗

bioRxiv2026

Ibogaine is associated with reorganization of high-beta brain networks in veterans with post-traumatic stress disorder

Post-traumatic stress disorder (PTSD) is a debilitating condition that affects millions of veterans. A recent observational study in 30 veterans showed that a single dose of the atypical psychedelic ibogaine can be highly effective at treating PTSD up to twelve months later. Although a prior study demonstrated that ibogaine transiently alters electroencephalography (EEG) power in various frequency bands, the long-term, network-level neural mechanisms targeted by ibogaine are unclear. Here, we investigated whether ibogaine-related clinical improvements are associated with the reorganization of certain brain networks. We applied a novel framework, FREQuency-resolved brain Network Estimation via Source Separation (FREQ-NESS), to identify frequency-specific brain networks in resting-state EEG data acquired at baseline, three to four days after treatment (immediate-post), and one month after ibogaine. At both the immediate-post and one month-post timepoints, high-beta (24 and 25 Hz) networks shifted away from frontal areas and towards posterior regions, an effect that was replicated in an independent EEG dataset on ibogaine treatment for opioid use disorder. This posterior shift was significantly correlated with improvements in PTSD symptoms at both timepoints. Neural field modeling demonstrated that these posterior high-beta shifts are associated with increases in corticocortical, but not corticothalamic, connectivity. Our results are consistent with prior evidence implicating aberrant frontal beta-band activity in PTSD. Overall, we demonstrate that the reconfiguration of high-beta brain networks could be a robust biomarker for ibogaines therapeutic effects.

neuroscience↗

bioRxiv2025

Pre-stimulus Brain States Predict and Control Variability in Stimulation Responses

Does the ongoing brain state determine how it responds to localized electrical stimulation? This fundamental question has major implications for neuroscience and medicine, as stimulation outcomes remain highly variable even when identical parameters (how) and target sites (where) are used. Such unpredictability undermines reproducibility, limits clinical reliability, and forces current protocols to rely on empirical trial-and-error rather than principled, evidence-based strategies. Mounting evidence suggests that the brains state prior to stimulation-a crucial when factor-shapes responses, yet the most reliable predictive markers remain unknown. Here, we systematically characterize the links between pre-stimulus (spontaneous) activity and post-stimulus (evoked) responses using simultaneous high-density EEG and stereotactic EEG from 36 epilepsy patients across[~] 320 sessions (>10,000 individual stimulations). We show that large-scale neural dynamics robustly predict stimulation outcomes, with a subset of measures-particularly network synchronization, functional connectivity, and spatiotemporal signal diversity-consistently forecasting responses across sessions. Whole-brain activity enhanced prediction compared to local assessments, and predictability varied across networks, being strongest in sensorimotor and visual regions. These findings establish a quantitative framework for state-dependent brain stimulation: by timing interventions to optimal pre-stimulus states, variability can be reduced and reproducibility enhanced. Our results directly address the fundamental question of where and when to stimulate, providing a pathway toward evidence-based protocols with improved therapeutic precision.

neuroscience↗

bioRxiv2025

Corticothalamic circuit mechanisms underlying brain region and ageing variations in resting-state alpha activity

Understanding the neural mechanisms underlying oscillations in resting-state brain activity, which exhibit substantial spatial and age-related variations, remains a significant challenge. This study aims to characterize the contributions of neural circuits to the mechanisms governing resting-state alpha oscillations, which are crucial for various neurocognitive processes and pathologies. Using the Cam-CAN dataset, source-space MEG analyses revealed a pronounced posterior-anterior gradient in alpha frequency, alpha power, and aperiodic components, alongside notable age-related changes. Through neurophysiological modelling, we uncover strong corticothalamic interactions in occipital regions, contrasting with predominantly corticocortical interactions in frontal areas. Ageing is associated with reduced intrathalamic activity and increased corticothalamic delay in occipital regions, while fronto-central regions exhibit increased intrathalamic activity. These findings establish how different circuits shape alpha oscillations across posterior-anterior axis and age, providing a mechanistic foundation for targeted clinical interventions and offering benchmarks for future studies in patient populations.

neuroscience↗

bioRxiv2025

Alpha rhythm subharmonics underlie responsiveness to theta burst stimulation via calcium metaplasticity

AO_SCPLOWBSTRACTC_SCPLOWRepetitive transcranial magnetic stimulation (rTMS) is a non-invasive technique to modulate brain activity, often used in treating Major Depressive Disorder (MDD) by targeting fronto-limbic circuitry. Despite its clinical utility, optimizing rTMS protocols remains challenging due to the complex and variable effects of stimulation parameter changes on synaptic plasticity. Oscillatory brain activity, measurable via Electroencephalography (EEG), serves as a biomarker for functional circuits and treatment response. To better understand the impact of rTMS on brain oscillations and connectivity, we used computational modeling of corticothalamic circuits to explore the mechanisms of stimulus-induced plasticity. We integrated calcium-dependent plasticity (CaDP) with Bienenstock-Cooper-Munro (BCM) metaplasticity formulations in a neural population model of resting-state EEG. By varying protocol parameters, we simulated iTBS effects on spectral power, synaptic efficacy, and calcium concentrations. Our findings highlight a resonance between theta stimulation and individual resting-state alpha rhythms, enhancing incoming excitatory long-term depression (LTD) and inhibitory long-term potentiation (LTP), leading to corticothalamic feed-forward inhibition (FFI). Induced effects were encapsulated by a weakening of corticothalamic loops and enhancement of intrathalamic loops. This work offers a novel paradigm for individualizing iTBS treatments, provides insights into the neurophysiological basis of clinical responsiveness, and offers a framework with which to derive tailored protocols.

neuroscience↗

bioRxiv2025

Excitation-Inhibition Balance and Fronto-Limbic Connectivity Drive TMS Treatment Outcomes in Refractory Depression

Depression affects over 350 million people worldwide, with treatment resistance occurring in up to 30% of cases. Intermittent theta burst stimulation (iTBS) targeting the left dorsolateral prefrontal cortex (DLPFC) has emerged as a promising intervention, yet the neurophysiological mechanisms determining which patients will respond remain poorly understood. Here, we combined transcranial magnetic stimulation with electroencephalography and whole-brain computational modeling to uncover the mechanistic basis of treatment efficacy in 90 patients with treatment-resistant depression. We identified two distinct neurophysiological signatures that differentiate responders from non-responders: (1) post-treatment shifts in excitation-inhibition balance toward greater inhibitory control, and (2) a pre-treatment brain state characterized by anticorrelated dynamics between subgenual anterior cingulate cortex and DLPFC. These features were significantly correlated with clinical improvement and could not be explained by non-specific factors. Our findings provide a neurophysiologically-informed framework for developing personalized and optimized neuromodulation approaches in treatment-resistant depression.

neuroscience↗

bioRxiv2025

State-dependent relationship between lower and higher order networks in fMRI brain dynamics

Resting-state functional magnetic resonance imaging (fMRI) is a powerful tool for exploring the brains functional organization. Functional connectivity (FC) is a commonly studied feature of fMRI data defined as the temporal correlation between activity patterns in pairs of brain regions. A major discovery of the past two decades of FC research has been the identification of consistent modular groupings in brain region time series correlations, commonly known as resting-state networks (RSNs). A second major discovery is the observation that RSNs in cortex are organized spatially along a functional/anatomical gradient. At one end of this gradient are lower order networks (LONs), predominantly specialized for unimodal information processing. At the other end are higher order networks (HONs), responsible for integrating multimodal information. Unlike the stable structural connectivity (SC) based on fixed anatomical links, FC fluctuates over time, and varies across brain regions. FC coordination within RSNs depends on SC, forming interconnected networks that regulate cognition, emotion, and behavior. The aim of the present study was to understand better how RSNs interact and communicate, based on their underlying SC. We used a whole-brain connectome-based neural mass modelling approach to study resting-state and task-based fMRI FC data. Following virtual SC lesions in the model, we characterized the FC changes within and between RSNs, and observed how these changes varied across different cognitive states. Our findings reveal how FC dynamics depend on underlying SC, highlighting the flexibility of these interactions across different brain states. LON lesions generally decrease FC within and between other LONs, and vice-versa for HON lesions. At rest, we observed a mutual antagonism between LONs and HONs, which was reversed during task conditions, with certain tasks increasing coordination between LONs and HONs. These results highlight the dynamic nature of brain network interactions, influenced by brain states and task demands. Our findings also have implications for clinical practice, offering insights into conditions such as brain tumors and stroke. HighlightsO_LIThe reduced Wong-Wang neural mass model was implemented in a PyTorch environment and used to understand what drives resting-state network functional connectivity (FC). C_LIO_LIStructural connectivity (SC) lesions were used to isolate functional networks to see how functional networks interact in various brain states. C_LIO_LIThere is a mutual antagonism between LONs and HONs in the resting-state. C_LIO_LIThis antagonism is seen to reverse in tasks with higher cognitive load (working memory, language). C_LIO_LITasks with a lower cognitive load (motor) share a similar network interactivity profile to the resting-state following network SC lesions. C_LI

neuroscience↗

bioRxiv2024

Assessing the Validity and Reliability of HD-DOT TD-fNIRS Resting-State Measurements in Rapid Succession Data Collection Settings

AO_SCPLOWBSTRACTC_SCPLOWFunctional magnetic resonance imaging (fMRI) has long been a cornerstone in the study of brain activity, but its high operational costs, limited availability, and restricted practical applicability have led researchers to seek alternative neuroimaging technologies. Recent advancements in high-density diffuse optical tomography (HD-DOT) and Time-Domain (TD) functional near-infrared spectroscopy (fNIRS) have emerged as promising solutions, offering the ability to generate detailed tomographic maps of hemodynamic fluctuations associated with neural activity. In this study, using the Kernel Flow device, we assess the performance of HD-DOT TD-fNIRS in terms of signal validity and reliability, particularly in rapid succession data collection settings. We conducted a multiple test-retest experiment involving fNIRS recordings from three participants across 20 ten-minute sessions of eyes-open resting-state brain activity over six days. Our findings indicate that HD-DOT TD-fNIRS systems like the Kernel Flow can reproduce hemodynamic patterns identified by fMRI, albeit with less spatial detail, and can detect resting state networks overall, though some individual network detections are not significant. The system performs consistently over days, with more variability within the time of day, and can capture subject-specific patterns with high accuracy as identified through FC fingerprinting analysis. We conclude that the new generation of HD-DOT TD-fNIRS systems holds significant promise for enhancing and expanding the measurement of functional brain data in both clinical and more naturalistic research settings. This study represents an important step towards a comprehensive understanding of the data quality and consistency achievable with these innovative neuroimaging devices.

neuroscience↗

bioRxiv2024

The Role of Inhibition in fMRI Resting-State Negative Correlations

Resting-state brain activity, as observed via functional magnetic resonance imaging (fMRI), displays non-random fluctuations whose functional connectivity (FC) is commonly parsed into spatial patterns of positive and negative correlations (PCs and NCs). Mapping NC patterns for certain key seed regions has shown considerable promise in recent years as a tool for enhancing neuro-navigated targeting and clinical outcomes of repetitive transcranial magnetic stimulation (rTMS) therapies in psychiatry. These successes bring to the fore several major outstanding questions about the neurophysiological origins of fMRI NCs. In this work, we studied candidate mechanisms for the emergence of fMRI NCs using connectome-based brain network modeling. Simulations of fMRI data under manipulation of inhibitory parameters WI and {lambda}, representing local and network-mediated inhibition respectively, were explored, focusing on the impact of inhibition levels on the emergence of NCs. Despite the considerable difference in time scales between GABAergic neuronal inhibition and fMRI FC, a clear relationship was observed, whereby the greater levels of overall inhibition led to significantly greater magnitude and spatial extent of NCs. We show that this effect is due to a leftward shift in the FC correlation distribution, leading to a reduction in the number of PCs and a concomitant increase in the number of NCs. Relatedly, we observed that those connections available for recruitment as NCs were precisely those with the weakest corresponding structural connectivity. Relative to nominally default values for the models used, greater levels of inhibition also improved, quantitatively and qualitatively, single-subject fits of simulated to empirical FC matrices. Our results provide new insights into how individual variability in anatomical connectivity strengths and neuronal inhibition levels may determine individualized expression of NCs in fMRI data. These, in turn, may offer new directions for optimization and personalization of rTMS therapies and other clinical applications of fMRI NC patterns. Author SummaryResting-state brain activity, as detected through functional magnetic resonance imaging (fMRI), demonstrates non-random fluctuations in its covariance structure, often characterized as functional connectivity (FC), which is further divided into spatial patterns of positive and negative correlations (NCs). Mapping patterns of NCs of specific key seed regions have demonstrated significant potential as a method for improving the precision of neuro-navigated targeting and enhancing clinical outcomes in the application of repetitive transcranial magnetic stimulation therapies within the field of psychiatry. In our study, we employed the reduced Wong-Wang neural mass model to investigate the physiological underpinnings of NCs observed in resting-state fMRI (rs-fMRI). Our simulated data partially captures the dynamics of empirical rs-fMRI data, revealing that increased inhibition levels correlate with a higher number of NCs. We also observed differential effects on model stability and NCs with varying levels of excitation and inhibition. These findings shed light on the complex interplay between neural dynamics and rs-fMRI connectivity patterns. Importantly, our work contributes to refining model parameters and offers insights for future validation with empirical clinical data. Understanding the factors influencing NCs in rs-fMRI FC has implications for optimizing therapeutic interventions and advancing our understanding of brain function.

neuroscience↗

bioRxiv2024

A comprehensive investigation of intracortical and corticothalamic models of alpha rhythms

The electroencephalographic alpha rhythm is one of the most robustly observed and widely studied empirical phenomena in all of neuroscience. However, despite its extensive implication in a wide range of cognitive processes and clinical pathologies, the mechanisms underlying alpha generation in neural circuits remain poorly understood. In this paper we offer a renewed foundation for research on this question, by undertaking a systematic comparison and synthesis of the most prominent theoretical models of alpha rhythmogenesis in the published literature. We focus on four models, each studied intensively by multiple authors over the past three decades: i) Jansen-Rit, ii) Moran-David-Friston, iii) Robinson-Rennie-Wright, and iv) Liley-Wright. Several common elements are identified, such as the use of second-order differential equations and sigmoidal potential-to-rate operators to represent population-level neural activity. Major differences are seen in other features such as wiring topologies and conduction delays. Through a series of mathematical analyses and numerical simulations, we nevertheless demonstrate that the selected models can be meaningfully compared, by associating parameters and circuit motifs of analogous biological significance. With this established, we conduct explorations of rate constant and synaptic connectivity parameter spaces, with the aim of identifying common patterns in key behaviours, such as the role of excitatory-inhibitory interactions in the generation of oscillations. Finally, using linear stability analysis we identify two qualitatively different alpha-generating dynamical regimes across the models: i) noise-driven fluctuations, and ii) self-sustained limit-cycle oscillations, emerging due to an Andronov-Hopf bifurcation. The comprehensive survey and synthesis developed here can, we suggest, be used to help guide future theoretical and experimental work aimed at disambiguating these and other candidate theories of alpha rhythmogenesis.

neuroscience↗

bioRxiv2024

Corticothalamic modelling of sleep neurophysiology with applications to mobile EEG

AO_SCPLOWBSTRACTC_SCPLOWRecent developments in mathematical modelling of EEG enable the tracking of otherwise-inaccessible neurophysiological parameters throughout sleep. Likewise, advancements in wearable electronics have enabled easy & affordable collection of sleep EEG at home. The convergence of these two advances, namely neurophysiological modelling for mobile sleep EEG, can boost preclinical and clinical assessments of sleep. However, this subject area has received limited attention in existing literature. To address this, we used an established model of the corticothalamic system to analyze EEG power spectra from 5 datasets, spanning from research-grade systems to at-home mobile EEG. In the present work, we compare the convergent and divergent features of the data and the estimated physiological model parameters. While data quality and characteristics differ considerably, key patterns consistent with previous theoretical and empirical work are observed. During the transition from lighter to deeper NREM, i) exponent of the aperiodic (1/f) spectral component is increased, ii) bottom-up thalamocortical drive is reduced, iii) corticocortical connection strengths are increased. This effect is observed in healthy subjects but is interestingly absent when taking SSRI antidepressants, suggesting possible effects of ascending neuromodulation on corticothalamic oscillations. We further show a month-long increase in REM% in one mobile EEG subject, associated with boosted high-frequency activity in spectra and higher thalamothalamic gains in the model, pointing to possible changes of thalamic inhibition in REM parasomnias. Our results provide a proof-of-principle for the utility and feasibility of this physiological modelling-based approach to analyzing mobile EEG data, providing a mechanistic measure of brain physiology during sleep. Statement of significanceWe employ a physiological model of the corticothalamic circuitry to model the EEG power spectra in sleep. We fit this model to 5 EEG datasets, and demonstrate that while mobile and non-mobile EEG recordings differ in their characteristics and quality, they can both robustly represent the changes along sleep stages using the aperiodic (1/f) component. We observe an increased corticocortical connection strength and decreased corticothalamic connection strength as the subject goes into deeper stages of NREM sleep; an effect that is, importantly, not observed in subjects taking SSRIs. This work provides a proof-of-concept for using mathematical modelling, working well for large mobile and non-mobile datasets providing valuable insight into the mechanisms generating sleep EEG.

neuroscience↗

bioRxiv2024

Stimulation mapping and whole-brain modeling reveal gradients of excitability and recurrence in cortical networks

The human brain exhibits a modular and hierarchical structure, spanning low-order sensorimotor to high-order cognitive/affective systems. What is the causal significance of this organization for brain dynamics and information processing properties? We investigated this question using rare simultaneous multimodal electrophysiology (stereotactic and scalp EEG) recordings in patients during presurgical intracerebral electrical stimulation (iES). Our analyses revealed an anatomical gradient of excitability across the cortex, with stronger iES-evoked EEG responses in high-order compared to low-order regions. Mathematical modeling further showed that this variation in excitability levels results from a differential dependence of recurrent feedback from non-stimulated regions across the anatomical hierarchy, and could be extinguished by suppressing those connections in-silico. High-order brain regions/networks thus show a more functionally integrated processing style than low-order ones, which manifests as a spatial gradient of excitability that is emergent from, and causally dependent on, the underlying hierarchical network structure.

neuroscience↗

bioRxiv2023

Reliability of resting-state EEG modulation by continuous and intermittent theta burst stimulation of the primary motor cortex: A sham-controlled study

Theta burst stimulation (TBS) is a form of repetitive transcranial magnetic stimulation designed to induce changes of cortical excitability that outlast the period of TBS application. In this study, we explored the effects of continuous TBS (cTBS) and intermittent TBS (iTBS) versus sham TBS stimulation, applied to the primary motor cortex, on modulation of resting state electroencephalography (rsEEG) power. We first conducted hypothesis-driven region-of-interest (ROI) analyses examining changes in alpha (8-12 Hz) and beta (13-21 Hz) bands over the left and right motor cortex. Additionally, we performed data-driven whole-brain analyses across a wide range of frequencies (1-50 Hz) and all electrodes. Finally, we assessed the reliability of TBS effects across two sessions approximately 1 month apart. None of the protocols produced significant group-level effects in the ROI. Whole-brain analysis revealed that cTBS significantly enhanced relative power between 19-43 Hz over multiple sites in both hemispheres. However, these results were not reliable across visits. There were no significant differences between EEG modulation by active and sham TBS protocols. Between-visit reliability of TBS-induced neuromodulatory effects was generally low-to-moderate. We discuss confounding factors and potential approaches for improving the reliability of TBS-induced rsEEG modulation.

neuroscience↗

bioRxiv2022

TMS-evoked responses are driven by recurrent large-scale network dynamics

AO_SCPLOWBSTRACTC_SCPLOWA major question in systems and cognitive neuroscience is to what extent neurostimulation responses are driven by recurrent activity. This question finds sharp relief in the case of TMS-EEG evoked potentials (TEPs). TEPs are spatiotemporal waveform patterns with characteristic inflections at [~]50ms, [~]100ms, and [~]150-200ms following a single TMS pulse that disperse from, and later reconverge to, the primary stimulated regions. What parts of the TEP are due to recurrent activity? And what light might this shed on more general principles of brain organization? We studied this using source-localized TMS-EEG analyses and whole-brain connectome-based computational modelling. Results indicated that recurrent network feedback begins to drive TEP responses from [~]100ms post-stimulation, with earlier TEP components being attributable to local reverberatory activity within the stimulated region. Subject-specific estimation of neurophysiological parameters additionally indicated an important role for inhibitory GABAergic neural populations in scaling cortical excitability levels, as reflected in TEP waveform characteristics.

neuroscience↗

bioRxiv2022

Deep Learning-Based Parameter Estimation for Neurophysiological Models of Neuroimaging Data

AO_SCPLOWBSTRACTC_SCPLOWConnectome-based neural mass modelling is the emerging computational neuroscience paradigm for simulating large-scale network dynamics observed in whole-brain activity measurements such as fMRI, M/EEG, and related techniques. Estimating physiological parameters by fitting these models to empirical data is challenging however, due to large network sizes, often physiologically detailed fast-timescale system equations, and the need for long (e.g. tens of minutes) simulation runs. Here we introduce a novel approach to connectome-based neural mass model parameter estimation by employing optimization tools developed for deep learning. We cast the system of differential equations representing both neural and haemodynamic activity dynamics as a deep neural network, implemented within a widely used machine learning programming environment (PyTorch). This allows us to use robust industry-standard optimization algorithms, automatic differentiation for computation of gradients, and other useful functionality. The approach is demonstrated using a connectome-based network with nodal dynamics specified by the two-state RWW mean-field neural mass model equations, which we use here as a model of fMRI-measured activity and correlation fluctuations. Additional optimization constraints are explored and prove fruitful, including restricting the model to domains of parameter space near a bifurcation point that yield metastable dynamics. Using these techniques, we first show robust recovery of physiological model parameters in synthetic data and then, as a proof-of-principle, apply the framework to modelling of empirical resting-state fMRI data from the Human Connectome Project database. For resting state activity, the system can be understood as a deep net that receives uncorrelated noise on its input layer, which is transformed into network-wide modelled functional connectivity on its output layer. This is consistent with the prevailing conception in theoretical neuroscience of resting-state functional connectivity patterns as an emergent phenomenon that is driven by (effectively) random activity fluctuations, which are then in turn spatiotemporally filtered by anatomical connectivity and local neural dynamics.

neuroscience↗

bioRxiv2021

Mapping inter-individual functional connectivity variability in TMS targets for major depressive disorder

Transcranial magnetic stimulation (TMS) is an emerging alternative to existing treatments for major depressive disorder (MDD). The effects of TMS on both brain physiology and therapeutic outcomes are known to be highly variable from subject to subject, however. Proposed reasons for this variability include individual differences in neurophysiology, in cortical geometry, and in brain connectivity. Standard approaches to TMS target site definition tend to focus on coordinates or landmarks within the individual brain regions implicated in MDD, such as the dorsolateral prefrontal cortex (dlPFC) and orbitofrontal cortex (OFC). Additionally considering the network connectivity of these sites has the potential to improve subject-specificity of TMS targeting and, in turn, improve treatment outcomes. We looked at the functional connectivity (FC) of dlPFC and OFC TMS targets, based on induced electrical field (E-field) maps, estimated using the SimNIBS library. We generated individualized E-field maps on the cortical surface for 121 subjects from the Human Connectome Project database using tetrahedral head models generated from T1-weighted MR images. We analyzed inter-subject variability in the shape and location of these TMS target E-field patterns, their FC, and the major functional networks to which they belong. Our results revealed the key differences in TMS target FC between the dlPFC and OFC, and also how this connectivity varies across subjects. Three major functional networks were targeted across the dlPFC and OFC: the ventral attention, fronto-parietal and default-mode networks in the dlPFC, and the fronto-parietal and default mode networks in the OFC. Inter-subject variability in cortical geometry and in FC was high. Our results characterize the FC patterns of canonical therapeutic TMS targets, and the key dimensions of their variability across subjects. The high inter-individual variability in cortical geometry and FC, leading to high variability in distributions of targeted brain networks, may account for the high levels of variability in physiological and therapeutic TMS outcomes. These insights should, we hope, prove useful as part of the broader effort by the psychiatry, neurology, and neuroimaging communities to help improve and refine TMS therapy, through a better understanding of the technology and its neurophysiological effects. HighlightsO_LIE-field modelling and functional connectivity used to study TMS targets (dlPFC,OFC) C_LIO_LIConsiderable variability in TMS target E-field patterns seen across subjects C_LIO_LILarge inter-subject differences in target connectivity observed and characterized C_LIO_LIMajor functional networks targeted by dlPFC, OFC TMS were the VAN, FPN and DMN C_LIO_LIInsights can contribute to improved and more personalized TMS therapies in the future C_LI

neuroscience↗