Biology subjects

Nozari, E.

Publications and source records attributed to Nozari, E..

12 recordsLinked to original sources

bioRxiv2026

Neural dynamics become significantly faster and more nonlinear at the peak of information processing

The temporal scale at which neural populations preserve stimulus information is often treated as fixed, even though population activity changes rapidly during sensory processing. Whether the information-optimal scale changes within a single stimulus-evoked episode, and what determines its evolution, remains unclear. We developed Integration-Split, a decoding-based framework that estimates the time-varying temporal resolution of population spike trains. Across four visual paradigms and throughout the mouse visual hierarchy, the optimal scale contracted after stimulus onset and subsequently re-expanded. This pattern persisted across stimulus temporal frequencies and was largely absent in a hippocampal control. The optimal scale varied inversely with stimulus decodability, reaching its finest resolution when stimulus identity was most accurately decoded. Beyond this association, independently estimating the signal-noise autocorrelation balance across candidate window widths identified the window that maximized the autocorrelation-based signal-to-noise ratio, and this optimal width closely followed the Integration-Split trajectory. Consistent with these findings, stronger nonlinear predictive structure coincided with finer optimal temporal resolution, consistent with the idea that shorter integration windows are needed to preserve informative nonlinear population structure that would otherwise be attenuated by temporal averaging. To test the functional relevance of this organization, we trained a Transformer solely to decode stimulus identity, without providing the optimal temporal-scale pattern or Integration-Split outputs. Its learned attention exhibited a similar contraction-and-recovery profile, and disrupting heads with stronger correspondence produced larger decoding deficits. Together, these findings reveal a dynamically changing, information-optimal temporal scale in population coding, shaped by the evolving balance of signal and noise autocorrelation and by nonlinear population structure, and independently recapitulated in the learned attention of a Transformer trained without temporal-scale supervision.

neuroscience↗

bioRxiv2026

Lateral Propagation from Sensory Cortex Provides Differential Gain for Sensory Selection

Selectively responding to target stimuli while ignoring distractors is a fundamental cognitive ability. In neocortex, such selection is associated with enhanced responses to target stimuli relative to distractor stimuli, yet the gain mechanisms underlying this differential responsivity are poorly understood. Most proposed gain mechanisms emphasize modulation within the center response field of a stimulus. In contrast, here we investigate modulation of lateral propagation from the center response field as a potential gain mechanism during a whisker-based selective detection task in mice. Both target (Go) and distractor (NoGo) stimuli evoked activity that propagated beyond their center response fields, with substantially greater lateral propagation for target stimuli. Target-evoked propagation increased with task learning and, in expert mice, exhibited the spatiotemporal characteristics of a traveling cortical wave preferentially involving infragranular layers. Furthermore, we demonstrate that frontal cortex actively suppresses distractor-evoked sensory responses, including both center region activity and lateral cortical recruitment. Computational modeling showed that modulation of lateral excitation could reproduce key features of these experimental findings. Together, our results support the regulation of lateral cortical propagation as a mechanism for differentially amplifying behaviorally relevant sensory representations during sensory selection.

neuroscience↗

bioRxiv2026

Bracket Coding: The Optimal Balance Between Temporal Integration and Segregation in Early Visual Processing

Despite over a century of research into the neural code, the fundamental principles by which the brain encodes sensory information remain debated. In this study we provide converging evidence for the presence of a dynamic, fast-switching integration of rate and temporal coding in the thalamus, primary visual cortex, and higher-order visual cortical areas of mice viewing an array of visual stimuli. This scheme is primarily characterized by the presence of distinct, temporally coordinated "bracket"s that tile the duration of each trial, are rate-coded within, and are separated by boundaries that are precisely-timed and synchronized across the population. Using large-scale Neuropixels recordings from the Allen Institute Visual Coding dataset, we provide evidence for the robustness and generality of bracket coding across several visual tasks and brain regions, as well as its optimality for information decoding, functional relevance for information representation, pronounced hierarchical organization, long-range bottom-up synchrony across visual regions, and coherence with low-frequency local field oscillations. These findings were all subsequently validated in a second, independent dataset provided by the International Brain Laboratory consortium. Finally, we provide a computational model that can serve as a potential mechanism for the generation of bracket-coded population spiking activity. Together, our results demonstrate the presence of a novel form of sensory information encoding in the brain, with broad implications for neuroscience and neuroengineering.

neuroscience↗

bioRxiv2026

On the Optimal Temporal Resolution for Information Representation in Neural Activity: A Theoretical Analysis

IntroductionAlthough neural activity is organized across multiple temporal and spatial scales, the principles determining information representation across scales remain unclear. In particular, while recent empirical results have reported mesoscale optimality in neural decoding, no theoretical accounts exist that can explain when and why such intermediate scales emerge as optimal. Here, we develop an analytical framework to determine optimal temporal scales of neural information representation and their dependence on signal and noise dynamics. Materials and MethodsWe formulate a multiscale model where neural population activity is represented by temporally encoded trial vectors at micro-, coarse meso-, fine meso- and macroscale resolutions. Neural responses are modeled as stimulus-dependent mean activations corrupted by temporally correlated noise, with signal and noise autocorrelation decay rates varied parametrically. Representational quality is quantified using the sensitivity index (d-prime), measuring the ability of an optimal decoder to distinguish stimulus conditions. ResultsWe derive closed-form expressions for the sensitivity index at each temporal scale and identify signal and noise autocorrelations as key determinants of decodability. We then validate our theoretical predictions against empirical decodability estimates from synthetic neural data. Comparing these expressions under various combinations of signal and noise autocorrelations across time reveals two main regimes. First, when signal and noise correlations are absent or persistent over time, the optimal resolution falls at one of the two extremes: macroscale (resp. microscale) if signal autocorrelations are significantly stronger (resp. weaker) than noise autocorrelations. When both signal and noise autocorrelations decay, temporal integration creates a trade-off: moderate integration improves decodability by suppressing noise while preserving coherent signal, whereas excessive integration degrades signal and decodability. Therefore, only in the latter regime, mesoscale representations emerge as the optimal regime across a broad range of biologically plausible parameters. DiscussionThis work provides a theoretical explanation for how optimal temporal scales depend on the interplay between signal and noise autocorrelations. The framework establishes temporal integration as a principled mechanism linking multiscale neural dynamics to information representation, explains when preprocessing operations such as binning and smoothing enhance or degrade decodability, and provides testable predictions across recording modalities and neural systems.

neuroscience↗

bioRxiv2026

Passive neuromodulation: an energy-driven mechanism for closed-loop suppression of epileptic seizure

For decades, electrical neuromodulation has been used as a therapeutic mechanism to disrupt and desynchronize pathological neural activity in various neurological disorders. Despite notable progress, however, patient outcomes remain highly variable, particularly in medically intractable epilepsy where surgery still provides the greatest chance of seizure freedom. Here we propose passive neuromodulation (PNM) as a radical alternative to conventional neurostimulation, whereby analogue feedback is used to drain energy from an epileptic circuit and thus suppress the initiation or spread of electrographic seizures. We provide pilot evidence on the efficacy and robustness of PNM using two computational models of epileptic dynamics: a detailed biophysical network model of dentate gyrus, and the Epileptor neural mass model of seizure dynamics. Despite the vast differences between these models, our results show the robust ability of PNM to suppress seizures in both models. We further demonstrate the efficacy and robustness of responsive PNM, whereby brief (50ms) windows of PNM are triggered by a simultaneously-running seizure detection algorithm, as well as the safe and tunable nature of PNM, where more robust seizure suppression can be achieved by parametrically titrating the amount of power drained from the tissue, without inducing any seizures even if applied interictally. Overall, our results provide strong evidence on the promise of PNM for the closed-loop control of epileptic seizures and other neurological disorders where damping pathological network activity can restore healthy dynamics.

neuroscience↗

bioRxiv2025

Multiscale predictive modeling robustly improves the accuracy of pseudo-prospective seizure forecasting in drug-resistant epilepsy

Extensive research over the past two decades has focused on identifying a preictal period in scalp as well as intracranial EEG (iEEG). This has led to a plethora of seizure prediction and forecasting algorithms which have reached only moderate success on curated and pre-segmented EEG datasets (accuracy/AUC {gtrsim} 0.8). Furthermore, when tested on their ability to pseudo-prospectively predict seizures from continuous EEG recordings, all existing algorithms suffer from low sensitivity (large false negatives), high time in warning (large false positives), or both. In this study we provide pilot evidence that predictive modeling of the dynamics of iEEG features (biomarkers), seizure risk, or both at the scale of tens of minutes can significantly improve the pseudo-prospective accuracy of almost any state-of-the-art seizure forecasting model. In contrast to the bulk of prior research that has focused on designing better features and classifiers, we start from off-the-shelf features and classifiers and shift the focus to learning how iEEG features (classifier input) and seizure risk (classifier output) evolve over time. Using iEEG from n = 5 patients undergoing presurgical evaluation at the Hospital of the University of Pennsylvania and six state-of-the-art baseline models, we first demonstrate that a wide array of iEEG features are highly predictable over time, with over 99% and 35% of studied features, respectively, having R2 > 0 for 10-second- and 10-minute-ahead prediction (mean R2 of 0.85 and 0.2). Furthermore, in almost all patients and baseline models, we observe a strong correlation between feature predictability (with some features remaining predictable up to 30 minutes) and classification-based feature importance. As a result, we subsequently demonstrate that adding an autoregressive model that predicts iEEG features on 12 {+/-} 4 minutes into the future is almost universally beneficial, with a mean improvement of 28% in terms of area under pseudo-prospective sensitivity-time in warning curve (PP-AUC). Addition of the second autoregressive predictive model at the level of seizure risk further improved accuracy, with a total mean improvement of 51% in PP-AUC. Our results provide pioneering evidence for the long-term predictability of seizure-relevant iEEG features and the vast utility of time series predictive modeling for improving seizure forecasting using continuous intracranial EEG.

neuroscience↗

bioRxiv2024

Whole-brain causal connectivity during decoded neurofeedback: a meta study

Decoded Neurofeedback (DecNef) enables modulation of targeted brain activity patterns without subjective awareness through multivariate pattern analysis, reinforcement learning, and real-time fMRI feedback. Despite its proven effectiveness, the causal mechanisms underlying DecNef and the neural dynamics that distinguish successful learners from those who struggle remain poorly understood. We conducted a meta-study using fMRI data from five DecNef experiments and extracted causal network dynamics as well as their associations with performance differences. Across studies, we found that connectivity within a posterior control hub-consisting of posterior cingulate, precuneus, and lateral posterior parietal cortices-is stronger during DecNef and positively correlates with neurofeedback success. Comparisons across cognition- and perception-targeted DecNef revealed separation in connections to somatomotor network, where connections between somatomotor and control-default-attention networks are larger during cognitive neurofeedback while connections between somatomotor and subcortical-visual-limbic networks are larger during perceptive DecNef. Whole-brain causal connectivity during DecNef further exhibited distinct network reorganizations, with greater subject-to-subject variability, increased engagement of control, limbic and visual, and decreased engagement of ventral and dorsal attention networks. Our results distill the complex and distributed network mechanisms underlying DecNef into dissociable roles for well-known functional subnetworks, thus advancing the research and clinical applications of decoded neurofeedback.

neuroscience↗

bioRxiv2023

Optimal Decoding of Neural Dynamics Occurs at Mesoscale Spatial and Temporal Resolutions

IntroductionUnderstanding the neural code has been one of the central aims of neuroscience research for decades. Spikes are commonly referred to as the units of information transfer, but multi-unit activity (MUA) recordings are routinely analyzed in aggregate forms such as binned spike counts, peri-stimulus time histograms, firing rates, or population codes. Various forms of averaging also occur in the brain, from the spatial averaging of spikes within dendritic trees to their temporal averaging through synaptic dynamics. However, how these forms of averaging are related to each other or to the spatial and temporal units of information representation within the neural code has remained poorly understood. Materials and MethodsIn this work we developed NeuroPixelHD, a symbolic hyperdimensional model of MUA, and used it to decode the spatial location and identity of static images shown to n = 9 mice in the Allen Institute Visual Coding - NeuroPixels dataset from large-scale MUA recordings. We parametrically varied the spatial and temporal resolutions of the MUA data provided to the model, and compared its resulting decoding accuracy. ResultsFor almost all subjects, we found 125ms temporal resolution to maximize decoding accuracy for both the spatial location of Gabor patches (81 classes for patches presented over a 9x9 grid) as well as the identity of natural images (118 classes corresponding to 118 images). The optimal spatial resolution was more heterogeneous among subjects, but was still found at either of two mesoscale levels in nearly all cases: the area level, where the spiking activity of neurons within each brain area are combined, and the population level, where the former are aggregated into two variables corresponding to fast spiking (putatively inhibitory) and regular spiking (putatively excitatory) neurons, respectively. DiscussionOur findings corroborate existing empirical practices of spatiotemporal binning and averaging in MUA data analysis, and provide a rigorous computational framework for optimizing the level of such aggregations. Our findings can also synthesize these empirical practices with existing knowledge of the various sources of biological averaging in the brain into a new theory of neural information processing in which the unit of information varies dynamically based on neuronal signal and noise correlations across space and time.

neuroscience↗

bioRxiv2023

Whole-Brain Causal Discovery Using fMRI

Despite significant research, discovering causal relationships from fMRI remains a challenge. Popular methods such as Granger Causality and Dynamic Causal Modeling fall short in handling contemporaneous effects and latent common causes. Methods from causal structure learning literature can address these limitations but often scale poorly with network size and need acyclicity. In this study, we first provide a taxonomy of existing methods and compare their accuracy and efficiency on simulated fMRI from simple topologies. This analysis demonstrates a pressing need for more accurate and scalable methods, motivating the design of Causal discovery for Large-scale Low-resolution Time-series with Feedback (CaLLTiF). CaLLTiF is a constraint-based method that uses conditional independence between contemporaneous and lagged variables to extract causal relationships. On simulated fMRI from the macaque connectome, CaLLTiF achieves significantly higher accuracy and scalability than all tested alternatives. From resting-state human fMRI, CaLLTiF learns causal connectomes that are highly consistent across individuals, show clear top-down flow of causal effect from attention and default mode to sensorimotor networks, exhibit Euclidean distance-dependence in causal interactions, and are highly dominated by contemporaneous effects. Overall, this work takes a major step in enhancing causal discovery from whole-brain fMRI and defines a new standard for future investigations. AUTHOR SUMMARYDiscovering causal relationships from fMRI data is challenging due to contemporaneous effects and latent causes. Popular methods like Granger Causality and Dynamic Causal Modeling struggle with these issues, especially in large networks. To address this, we introduce CaLLTiF, a scalable method that uses both lagged and contemporaneous variables to identify causal relationships. CaLLTiF outperforms various existing techniques in accuracy and scalability on simulated fMRI data. When applied to human resting-state fMRI, it reveals consistent and biologically-plausible patterns across individuals, with a clear top-down causal flow from attention and default mode networks to sensorimotor areas. Overall, this work advances the field of causal discovery in large-scale fMRI studies.

neuroscience↗

bioRxiv2023

Predictive Modeling of Evoked Intracranial EEG Response to Medial Temporal Lobe Stimulation in Patients with Epilepsy

Despite promising advancements, closed-loop neurostimulation for drug-resistant epilepsy (DRE) still relies on manual tuning and produces variable outcomes, while automated predictable algorithms remain an aspiration. As a fundamental step towards addressing this gap, here we study predictive dynamical models of human intracranial EEG (iEEG) response under parametrically rich neurostimulation. Using data from n = 13 DRE patients, we find that stimulation-triggered switched-linear models with [~]300ms of causal historical dependence best explain evoked iEEG dynamics. These models are highly consistent across different stimulation amplitudes and frequencies, allowing for learning a generalizable model from abundant STIM OFF and limited STIM ON data. Further, evoked iEEG in nearly all subjects exhibited a distance-dependent pattern, whereby stimulation directly impacts the actuation site and nearby regions ([lsim] 20mm), affects medium-distance regions (20 [~] 100mm) through network interactions, and hardly reaches more distal areas ({gtrsim} 100mm). Peak network interaction occurs at 60[~]80mm from the stimulation site. Due to their predictive accuracy and mechanistic interpretability, these models hold significant potential for model-based seizure forecasting and closed-loop neurostimulation design.

neuroscience↗

bioRxiv2022

Network controllability measures of subnetworks: implications for neurosciences

Recent progress in network sciences has made it possible to apply key findings from control theory to the study of networks. Referred to as network control theory, this framework describes how the interactions between interconnected system elements and external energy sources, potentially constrained by different optimality criteria, result in complex network behavior. A typical example is the quantification of the functional role certain brain regions or symptoms play in shaping the temporal dynamics of brain activity or the clinical course of a disease, a property that is quantified in terms of the so-called controllability metrics. Critically though, contrary to the engineering context in which control theory was originally developed, a mathematical understanding of the network nodes and connections in neurosciences cannot be assumed. For instance, in the case of psychological systems such as those studied to understand the psychiatric disorders, a potentially large set variables are unknown. As such, while the measures offered by network control theory would be mathematically correct, in that they can be calculated with high precision, they could have little translational values with respect to their putative role suggested by controllability metrics. It is therefore critical to understand if and how the controllability metrics computer over subnetworks would deviate, if access to the complete set of variables, as in neurosciences, cannot be taken for granted. In this paper, we use a host of simulations based on synthetic as well as structural MRI data to study the potential deviation of controllability metrics in sub-compared to the full networks. Specifically, we estimate average- and modal-controllability, two of the most widely used controllability measures in neurosciences, in a large number of settings where we systematically vary network type, network size, and edge density. We find out, across all network types we test, that average and modal controllability are systematically, either over- or underestimated depending on the number of nodes in the sub- and full network and the edge density. Finally, we provide a formal theoretical proof that our observations generalize to any network type and discuss the ramifications of this systematic bias and potential solutions to alleviate the problem.

neuroscience↗

bioRxiv2020

Is the brain macroscopically linear? A system identification of resting state dynamics

A central challenge in the computational modeling of neural dynamics is the trade-off between accuracy and simplicity. At the level of individual neurons, nonlinear dynamics are both experimentally established and essential for neuronal functioning. One may therefore expect the collective dynamics of massive networks of such neurons to exhibit an even larger repertoire of nonlinear behaviors. An implicit assumption has thus formed that an "accurate" computational model of whole-brain dynamics must inevitably be non-linear whereas linear models may provide a first-order approximation. To what extent this assumption holds, however, has remained an open question. Here, we provide new evidence that challenges this assumption at the level of whole-brain blood-oxygen-level-dependent (BOLD) and macroscopic field potential dynamics by leveraging the theory of system identification. Using functional magnetic resonance imaging (fMRI) and intracranial electroencephalography (iEEG), we model the spontaneous, resting state activity of 700 subjects in the Human Connectome Project (HCP) and 122 subjects from the Restoring Active Memory (RAM) project using state-of-the-art linear and nonlinear model families. We assess relative model fit using predictive power, computational complexity, and the extent of residual dynamics unexplained by the model. Contrary to our expectations, linear auto-regressive models achieve the best measures across all three metrics. To understand and explain this linearity, we highlight four properties of macroscopic neurodynamics which can counteract or mask microscopic nonlinear dynamics: averaging over space, averaging over time, observation noise, and limited data samples. Whereas the latter two are technological limitations and can improve in the future, the former two are inherent to aggregated macroscopic brain activity. Our results demonstrate the discounted potential of linear models in accurately capturing macroscopic brain dynamics. This, together with the unparalleled interpretability of linear models, can greatly facilitate our understanding of macroscopic neural dynamics, which in turn may facilitate the principled design of model-based interventions for the treatment of neuropsychiatric disorders.

neuroscience↗