Front Comput Neurosci [JOURNAL]
Sun
200 papers
A neural network model combining the successor representation and actor-critic methods reveals effective biological use of the representation.
Tsurumi T, Morita K
Front Comput Neurosci · 2025 · PMID 41383549 · Full text
In learning goal-directed behavior, state representation is important for adapting to the environment and achieving goals. A predictive state representation called successive representation (SR) has recently attracted at...
In learning goal-directed behavior, state representation is important for adapting to the environment and achieving goals. A predictive state representation called successive representation (SR) has recently attracted attention as a candidate for state representation in animal brains, especially in the hippocampus. The relationship between the SR and the animal brain has been studied, and several neural network models for computing the SR have been proposed based on the findings. However, studies on implementation of the SR involving action selection have not yet advanced significantly. Therefore, we explore possible mechanisms by which the SR is utilized biologically for action selection and learning optimal action policies. The actor-critic architecture is a promising model of animal behavioral learning in terms of its correspondence to the anatomy and function of the basal ganglia, so it is suitable for our purpose. In this study, we construct neural network models for behavioral learning using the SR. By using them to perform reinforcement learning, we investigate their properties. Specifically, we investigated the effect of using different state representations for the actor and critic in the actor-critic method, and also compared the actor-critic method with Q-learning and SARSA. We found the difference between the effect of using the SR for the actor and the effect of using the SR for the critic in the actor-critic method, and observed that using the SR in conjunction with one-hot encoding makes it possible to learn with the benefits of both representations. These results suggest the possibility that the striatum can learn using multiple state representations complementarily.
From generative AI to the brain: five takeaways.
Gros C
Front Comput Neurosci · 2025 · PMID 41368649 · Full text
The big strides seen in generative AI are not based on somewhat obscure algorithms, but due to clearly defined generative principles. The resulting concrete implementations have proven themselves in large numbers of appl...
The big strides seen in generative AI are not based on somewhat obscure algorithms, but due to clearly defined generative principles. The resulting concrete implementations have proven themselves in large numbers of applications. We suggest that it is imperative to thoroughly investigate which of these generative principles may be operative also in the brain, and hence relevant for cognitive neuroscience. In addition, ML research led to a range of interesting characterizations of neural information processing systems. We discuss five examples, the shortcomings of world modeling, the generation of thought processes, attention, neural scaling laws, and quantization, that illustrate how much neuroscience could potentially learn from ML research.
Exploring internal representations of self-supervised networks: few-shot learning abilities and comparison with human semantics and recognition of objects.
Kataoka A, Nagano Y, Oizumi M
Front Comput Neurosci · 2025 · PMID 41357072 · Full text
Recent advances in self-supervised learning have attracted significant attention from both machine learning and neuroscience. This is primarily because self-supervised methods do not require annotated supervisory informa...
Recent advances in self-supervised learning have attracted significant attention from both machine learning and neuroscience. This is primarily because self-supervised methods do not require annotated supervisory information, making them applicable to training artificial networks without relying on large amounts of curated data, and potentially offering insights into how the brain adapts to its environment in an unsupervised manner. Although several previous studies have elucidated the correspondence between neural representations in deep convolutional neural networks (DCNNs) and biological systems, the extent to which unsupervised or self-supervised learning can explain the human-like acquisition of categorically structured information remains less explored. In this study, we investigate the correspondence between the internal representations of DCNNs trained using a self-supervised contrastive learning algorithm and human semantics and recognition. To this end, we employ a few-shot learning evaluation procedure, which measures the ability of DCNNs to recognize novel concepts from limited exposure, to examine the inter-categorical structure of the learned representations. Two comparative approaches are used to relate the few-shot learning outcomes to human semantics and recognition, with results suggesting that the representations acquired through contrastive learning are well aligned with human cognition. These findings underscore the potential of self-supervised contrastive learning frameworks to model learning mechanisms similar to those of the human brain, particularly in scenarios where explicit supervision is unavailable, such as in human infants prior to language acquisition.
A hierarchical Bayesian inference model for volatile multivariate exponentially distributed signals.
Zhu C, Zhou K, Tang F … +3 more , Tang Y, Li X, Si B
Front Comput Neurosci · 2025 · PMID 41312357 · Full text
Brain activities often follow an exponential family of distributions. The exponential distribution is the maximum entropy distribution of continuous random variables in the presence of a mean. The memoryless and peakless...
Brain activities often follow an exponential family of distributions. The exponential distribution is the maximum entropy distribution of continuous random variables in the presence of a mean. The memoryless and peakless properties of an exponential distribution impose difficulties for data analysis methods. To estimate the rate parameter of multivariate exponential distribution from a time series of sensory inputs (i.e., observations), we constructed a hierarchical Bayesian inference model based on a variant of general hierarchical Brownian filter (GHBF). To account for the complex interactions among multivariate exponential random variables, the model estimates the second-order interaction of the rate intensity parameter in logarithmic space. Using variational Bayesian scheme, a family of closed-form and analytical update equations are introduced. These update equations also constitute a complete predictive coding framework. The simulation study shows that our model has the ability to evaluate the time-varying rate parameters and the underlying correlation structure of volatile multivariate exponentially distributed signals. The proposed hierarchical Bayesian inference model is of practical utility in analyzing high-dimensional neural activities.
Common characteristics of variants linked to autism spectrum disorder in the WAVE regulatory complex.
Xie S, Zuo K, De Rubeis S … +4 more , Bonollo G, Colombo G, Ruggerone P, Carloni P
Front Comput Neurosci · 2025 · PMID 41312356 · Full text
Six variants associated with autism spectrum disorder (ASD) abnormally activate the WASP-family Verprolin-homologous protein (WAVE) regulatory complex (WRC), a critical regulator of actin dynamics. This abnormal activati...
Six variants associated with autism spectrum disorder (ASD) abnormally activate the WASP-family Verprolin-homologous protein (WAVE) regulatory complex (WRC), a critical regulator of actin dynamics. This abnormal activation may contribute to the pathogenesis of this disorder. Using molecular dynamics (MD) simulations, we recently investigated the structural dynamics of wild-type (WT) WRC and R87C, A455P, and Q725R WRC disease-linked variants. Here, by extending MD simulations to I664M, E665K, and D724H WRC, we suggest that of the mutations weaken the interactions and affect intra-complex allosteric communication between the WAVE1 active C-terminal region (ACR) and the rest of the complex. This might contribute to an abnormal complex activation, a hallmark of WRC-linked ASD. In addition, all mutants but I664M destabilize the ACR V-helix and increase the participation of ACR in large-scale movements. All these features may also abnormally influence the inactive WRC toward a dysfunctional state. We hypothesize that small-molecule ligands counteracting these effects may help restore normal WRC regulation in ASD-related variants.
Time delays in computational models of neuronal and synaptic dynamics.
Madadi Asl M
Front Comput Neurosci · 2025 · PMID 41293077 · Full text
Abstract loading — click title to view on PubMed.
Triboelectric nanogenerators for neural data interpretation: bridging multi-sensing interfaces with neuromorphic and deep learning paradigms.
Gan L, Yuan S, Guo M … +3 more , Wang Q, Deng Z, Jia B
Front Comput Neurosci · 2025 · PMID 41281720 · Full text
The rapid growth of computational neuroscience and brain-computer interface (BCI) technologies require efficient, scalable, and biologically compatible approaches for neural data acquisition and interpretation. Tradition...
The rapid growth of computational neuroscience and brain-computer interface (BCI) technologies require efficient, scalable, and biologically compatible approaches for neural data acquisition and interpretation. Traditional sensors and signal processing pipelines often struggle with the high dimensionality, temporal variability, and noise inherent in neural signals, particularly in elderly populations where continuous monitoring is essential. Triboelectric nanogenerators (TENGs), as self-powered and flexible multi-sensing devices, offer a promising avenue for capturing neural-related biophysical signals such as electroencephalography (EEG), electromyography (EMG), and cardiorespiratory dynamics. Their low-power and wearable characteristics make them suitable for long-term health and neurocognitive monitoring. When combined with deep learning models-including convolutional neural networks (CNNs), recurrent neural networks (RNNs), and spiking neural networks (SNNs)-TENG-generated signals can be efficiently decoded, enabling insights into neural states, cognitive functions, and disease progression. Furthermore, neuromorphic computing paradigms provide an energy-efficient and biologically inspired framework that naturally aligns with the event-driven characteristics of TENG outputs. This mini review highlights the convergence of TENG-based sensing, deep learning algorithms, and neuromorphic systems for neural data interpretation. We discuss recent progress, challenges, and future perspectives, with an emphasis on applications in computational neuroscience, neurorehabilitation, and elderly health care.
Neural heterogeneity as a unifying mechanism for efficient learning in spiking neural networks.
Zhang F, Cui J
Front Comput Neurosci · 2025 · PMID 41281719 · Full text
The brain is a highly diverse and heterogeneous network, yet the functional role of this neural heterogeneity remains largely unclear. Despite growing interest in neural heterogeneity, a comprehensive understanding of ho...
The brain is a highly diverse and heterogeneous network, yet the functional role of this neural heterogeneity remains largely unclear. Despite growing interest in neural heterogeneity, a comprehensive understanding of how it influences computation across different neural levels and learning methods is still lacking. In this work, we systematically examine the neural computation of spiking neural networks (SNNs) in three key sources of neural heterogeneity: external, network, and intrinsic heterogeneity. We evaluate their impact using three distinct learning methods, which can carry out tasks ranging from simple curve fitting to complex network reconstruction and real-world applications. Our results show that while different types of neural heterogeneity contribute in distinct ways, they consistently improve learning accuracy and robustness. These findings suggest that neural heterogeneity across multiple levels improves learning capacity and robustness of neural computation, and should be considered a core design principle in the optimization of SNNs.
Interleaving cortex-analog mixing improves deep non-negative matrix factorization networks.
Nouri M, Rotermund D, Garcia-Ortiz A … +1 more , Pawelzik KR
Front Comput Neurosci · 2025 · PMID 41267934 · Full text
Considering biological constraints in artificial neural networks has led to dramatic improvements in performance. Nevertheless, to date, the positivity of long-range signals in the cortex has not been shown to yield impr...
Considering biological constraints in artificial neural networks has led to dramatic improvements in performance. Nevertheless, to date, the positivity of long-range signals in the cortex has not been shown to yield improvements. While Non-negative matrix factorization (NMF) captures biological constraints of positive long-range interactions, deep convolutional neural networks with NMF modules do not match the performance of conventional neural networks (CNNs) of a similar size. This work shows that introducing intermediate modules that combine the NMF's positive activities, analogous to the processing in cortical columns, leads to improved performance on benchmark data that exceeds that of vanilla deep convolutional networks. This demonstrates that including positive long-range signaling together with local interactions of both signs in analogy to cortical hyper-columns has the potential to enhance the performance of deep networks.
Universal differential equations as a unifying modeling language for neuroscience.
El-Gazzar A, van Gerven M
Front Comput Neurosci · 2025 · PMID 41244995 · Full text
The rapid growth of large-scale neuroscience datasets has spurred diverse modeling strategies, ranging from mechanistic models grounded in biophysics, to phenomenological descriptions of neural dynamics, to data-driven d...
The rapid growth of large-scale neuroscience datasets has spurred diverse modeling strategies, ranging from mechanistic models grounded in biophysics, to phenomenological descriptions of neural dynamics, to data-driven deep neural networks (DNNs). Each approach offers distinct strengths as mechanistic models provide interpretability, phenomenological models capture emergent dynamics, and DNNs excel at predictive accuracy but this also comes with limitations when applied in isolation. Universal differential equations (UDEs) offer a unifying modeling framework that integrates these complementary approaches. By treating differential equations as parameterizable, differentiable objects that can be combined with modern deep learning techniques, UDEs enable hybrid models that balance interpretability with predictive power. We provide a systematic overview of the UDE framework, covering its mathematical foundations, training methodologies, and recent innovations. We argue that UDEs fill a critical gap between mechanistic, phenomenological, and data-driven models in neuroscience, with potential to advance applications in neural computation, neural control, neural decoding, and normative modeling in neuroscience.
Multiscale intracranial EEG dynamics across sleep-wake states: toward memory-related processing.
Tenti JM, Pallares Di Nunzio M, Bab MA … +3 more , Rosso OA, Montani F, Arlego MJF
Front Comput Neurosci · 2025 · PMID 41210195 · Full text
Sleep is known to support memory consolidation through a complex interplay of neural dynamics across multiple timescales. Using intracranial EEG (iEEG) recordings from patients undergoing clinical monitoring, we characte...
Sleep is known to support memory consolidation through a complex interplay of neural dynamics across multiple timescales. Using intracranial EEG (iEEG) recordings from patients undergoing clinical monitoring, we characterize spectral activity, neuronal avalanche dynamics, and temporal correlations across sleep-wake states, with a focus on their spatial distribution and potential functional relevance. We observe increased low-frequency power, larger avalanches, and enhanced long-range temporal correlations-quantified via Detrended Fluctuation Analysis-during N2 and N3 sleep. In contrast, REM sleep and wakefulness show reduced temporal persistence and fewer large-scale cascades, suggesting a shift toward more fragmented and flexible dynamics. These signatures vary across cortical regions, with distinctive patterns emerging in medial temporal and frontal areas-regions implicated in memory processing. Rather than providing direct evidence of consolidation, our results point to a functional neural landscape that may favor both stabilization and reconfiguration of internal representations during sleep. Overall, our findings highlight the utility of iEEG in revealing the multiscale spatio-temporal structure of sleep-related brain dynamics, offering insights into the physiological conditions that support memory-related processing.
Sudden restructuring of memory representations in recurrent neural networks with repeated stimulus presentations.
Howlett JR
Front Comput Neurosci · 2025 · PMID 41200719 · Full text
While acquisition curves in human learning averaged at the group level display smooth, gradual changes in performance, individual learning curves across cognitive domains reveal sudden, discontinuous jumps in performance...
While acquisition curves in human learning averaged at the group level display smooth, gradual changes in performance, individual learning curves across cognitive domains reveal sudden, discontinuous jumps in performance. Similar thresholding effects are a hallmark of a range of nonlinear systems which can be explored using simple, abstract models. Here, I investigate discontinuous changes in learning performance using Amari-Hopfield networks with Hebbian learning rules which are repeatedly exposed to a single stimulus. Simulations reveal that the attractor basin size for a target stimulus increases in discrete jumps rather than gradual changes with repeated stimulus exposure. The distribution of the size of these positive jumps in basin size is best approximated by a lognormal distribution, suggesting that the distribution is heavy-tailed. Examination of the transition graph structure for networks before and after basin size changes reveals that newly acquired states are often organized into hierarchically branching tree structures, and that the distribution of branch sizes is best approximated by a power law distribution. The findings suggest that even simple nonlinear network models of associative learning exhibit discontinuous changes in performance with repeated learning which mirror behavioral results observed in humans. Future work can investigate similar mechanisms in more biologically detailed network models, potentially offering insight into the network mechanisms of learning with repeated exposure or practice.
An AI methodology to reduce training intensity, error rates, and size of neural networks.
Kobylarz TJA
Front Comput Neurosci · 2025 · PMID 41195189 · Full text
Massive computing systems are required to train neural networks. The prodigious amount of consumed energy makes the creation of AI applications significant polluters. Despite the enormous training effort, neural network...
Massive computing systems are required to train neural networks. The prodigious amount of consumed energy makes the creation of AI applications significant polluters. Despite the enormous training effort, neural network error rates limit its use for medical applications, because errors can lead to intolerable morbidity and mortality. Two reasons contribute to the excessive training requirements and high error rates; an iterative reinforcement process (tuning) that does not guarantee convergence and the deployment of neuron models only capable of realizing linearly separable switching functions. tuning procedures require tens of thousands of training iterations. In addition, linearly separable neuron models have severely limited capability; which leads to large neural nets. For seven inputs, the ratio of total possible switching functions to linearly separable switching functions is 41 octillion. Addressed here is the creation of neuron models for the application of disease diagnosis. Algorithms are described that perform direct neuron creation. This results in far fewer training steps than that of current AI systems. The design algorithms result in neurons that do not manufacture errors (hallucinations). The algorithms utilize a template to create neuron models that are capable of performing any type of switching function. The algorithms show that a neuron model capable of performing both linearly and nonlinearly separable switching functions is vastly superior to the neuron models currently being used. Included examples illustrate use of the template for determining disease diagnoses (outputs) from symptoms (inputs). The examples show convergence with a single training iteration.
Neuron synchronization analyzed through spatial-temporal attention.
Yang H, Kc P, Chen P … +4 more , Lei H, Sponberg S, Tarokh V, Riffell JA
Front Comput Neurosci · 2025 · PMID 41180119 · Full text
Neuronal synchronization refers to the temporal coordination of activity across populations of neurons, a process that underlies coherent information processing, supports the encoding of diverse sensory stimuli, and faci...
Neuronal synchronization refers to the temporal coordination of activity across populations of neurons, a process that underlies coherent information processing, supports the encoding of diverse sensory stimuli, and facilitates adaptive behavior in dynamic environments. Previous studies of synchronization have predominantly emphasized rate coding and pairwise interactions between neurons, which have provided valuable insights into emergent network phenomena but remain insufficient for capturing the full complexity of temporal dynamics in spike trains, particularly the interspike interval. To address this limitation, we performed neural ensemble recording in the primary olfactory center-the antennal lobe (AL) of the hawk moth -by stimulating with floral odor blends and systematically varying the concentration of an individual odorant within one of the mixtures. We then applied machine learning methods integrating modern attention mechanisms and generative normalizing flows, enabling the extraction of semi-interpretable attention weights that characterize dynamic neuronal interactions. These learned weights not only recapitulated the established principles of neuronal synchronization but also facilitated the functional classification of two major cell types in the antennal lobe (AL) [local interneurons (LNs) and projection neurons (PNs)]. Furthermore, by experimentally manipulating the excitation/inhibition balance within the circuit, our approach revealed the relationships between synchronization strength and odorant composition, providing new insight into the principles by which olfactory networks encode and integrate complex sensory inputs.
Using noise to distinguish between system and observer effects in multimodal neuroimaging.
Fagerholm ED, Tanaka H, Scott G … +5 more , Leech R, Turkheimer FE, Zeidman P, Friston KJ, Brázdil M
Front Comput Neurosci · 2025 · PMID 41180118 · Full text
INTRODUCTION: It has become increasingly common to record brain activity simultaneously at more than one spatiotemporal scale. Here, we address a central question raised by such cross-scale datasets: do they reflect the...
INTRODUCTION: It has become increasingly common to record brain activity simultaneously at more than one spatiotemporal scale. Here, we address a central question raised by such cross-scale datasets: do they reflect the same underlying dynamics observed in different ways, or different dynamics observed in the same way? In other words, to what extent can variation between modalities be attributed to system-level versus observer-level effects? System-level effects reflect genuine differences in neural dynamics at the resolution sampled by each device. Observer-level effects, by contrast, reflect artefactual differences introduced by the nonlinear transformations each device imposes on the signal. We demonstrate that noise, when incorporated into generative models, can help disentangle these two sources of variation. METHODS: We apply this noise-based approach to simultaneously recorded high-frequency broadband signals from macroelectrodes and microwires in the human hippocampus. RESULTS: Most subjects show a complex mixture of system- and observer-level contributions to their time series. However, in one subject, the cross-scale difference is statistically attributable to an observer-level effect-i.e., consistent with the same dynamics at both microwire and macroelectrode scales. DISCUSSION: This study shows that noise can be used in empirical datasets to determine whether cross-scale variation arises from differences in neural dynamics or differences in observer functions.
Modeling cognition through adaptive neural synchronization: a multimodal framework using EEG, fMRI, and reinforcement learning.
Hall R, Jackson M, Maleki M … +1 more , Crogman HT
Front Comput Neurosci · 2025 · PMID 41180117 · Full text
INTRODUCTION: Understanding the cognitive process of thinking as a neural phenomenon remains a central challenge in neuroscience and computational modeling. This study addresses this challenge by presenting a biologicall...
INTRODUCTION: Understanding the cognitive process of thinking as a neural phenomenon remains a central challenge in neuroscience and computational modeling. This study addresses this challenge by presenting a biologically grounded framework that simulates adaptive decision making across cognitive states. METHODS: The model integrates neuronal synchronization, metabolic energy consumption, and reinforcement learning. Neural synchronization is simulated using Kuramoto oscillators, while energy dynamics are constrained by multimodal activity profiles. Reinforcement learning agents-Q-learning and Deep Q-Network (DQN)-modulate external inputs to maintain optimal synchrony with minimal energy cost. The model is validated using real EEG and fMRI data, comparing simulated and empirical outputs across spectral power, phase synchrony, and BOLD activity. RESULTS: The DQN agent achieved rapid convergence, stabilizing cumulative rewards within 200 episodes and reducing mean synchronization error by over 40%, outperforming Q-learning in speed and generalization. The model successfully reproduced canonical brain states-focused attention, multitasking, and rest. Simulated EEG showed dominant alpha-band power (3.2 × 10 a.u.), while real EEG exhibited beta-dominance (3.2 × 10 a.u.), indicating accurate modeling of resting states and tunability for active tasks. Phase Locking Value (PLV) ranged from 0.9806 to 0.9926, with the focused condition yielding the lowest circular variance (0.0456) and a near significant phase shift compared to rest ( = -2.15, = 0.075). Cross-modal validation revealed moderate correlation between simulated and real BOLD signals ( = 0.30, resting condition), with delayed inputs improving temporal alignment. General Linear Model (GLM) analysis of simulated BOLD data showed high region-specific prediction accuracy ( = 0.973-0.993, < 0.001), particularly in prefrontal, parietal, and anterior cingulate cortices. Voxel-wise correlation and ICA decomposition confirmed structured network dynamics. DISCUSSION: These findings demonstrate that the framework captures both electrophysiological and spatial aspects of brain activity, respects neuroenergetic constraints, and adaptively regulates brain-like states through reinforcement learning. The model offers a scalable platform for simulating cognition and developing biologically inspired neuroadaptive systems. CONCLUSION: This work provides a novel and testable approach to modeling thinking as a biologically constrained control problem and lays the groundwork for future applications in cognitive modeling and brain-computer interfaces.
Advancing epileptic seizure recognition through bidirectional LSTM networks.
Al-Marzouki S
Front Comput Neurosci · 2025 · PMID 41180116 · Full text
Seizure detection in a timely and accurate manner remains a primary challenge in clinical neurology, affecting diagnosis planning and patient management. Most of the traditional methods rely on feature extraction and tra...
Seizure detection in a timely and accurate manner remains a primary challenge in clinical neurology, affecting diagnosis planning and patient management. Most of the traditional methods rely on feature extraction and traditional machine learning techniques, which are not efficient in capturing the dynamic characteristics of neural signals. It is the aim of this study to address such limitations by designing a deep learning model from bidirectional Long Short-Term Memory (BiLSTM) networks in a bid to enhance epileptic seizure identification reliability and accuracy. The dataset used, drawn from Kaggle's Epileptic Seizure Recognition challenge, consists of 11,500 samples with 179 features per sample corresponding to different electroencephalogram (EEG) readings. Data preprocessing was utilized to normalize and structure the input to the deep learning model. The proposed BiLSTM model employs sophisticated architecture to leverage temporal dependency and bidirectional data flows. It incorporates multiple dense and dropout layers alongside batch normalization to enhance the capability of the model in learning from the EEG data in an efficient manner. It supports end-to-end feature learning from the raw EEG signals without the need for intensive preprocessing and feature engineering. BiLSTM model performed better than others with 98.70% accuracy on the validation set and surpassed traditional techniques. The F1-score and other statistical metrics also validated the performance of the model as the confusion matrix achieved high values for recall and precision. The results confirm the capability of bidirectional LSTM networks to better identify seizures with significant improvements over conventional practices. Apart from facilitating seizure detection in a reliable fashion, the method improves the overall field of biomedical signal processing and can also be used in real-time observation and intervention protocols.
CRISP: a correlation-filtered recursive feature elimination and integration of SMOTE pipeline for gait-based Parkinson's disease screening.
Afzal N, Iqbal J, Waris A … +5 more , Khan MJ, Hazzazi F, Ali H, Ijaz MA, Gilani SO
Front Comput Neurosci · 2025 · PMID 41140842 · Full text
INTRODUCTION: Parkinson's disease (PD) is the fastest-growing neurodegenerative disorder, with subtle gait changes such as reduced vertical ground-reaction forces (VGRF) often preceding motor symptoms. These gait abnorma...
INTRODUCTION: Parkinson's disease (PD) is the fastest-growing neurodegenerative disorder, with subtle gait changes such as reduced vertical ground-reaction forces (VGRF) often preceding motor symptoms. These gait abnormalities, measurable via wearable VGRF sensors, offer a non-invasive means for early PD detection. However, current computational approaches often suffer from redundant features and class imbalance, limiting both accuracy and generalizability. METHODS: We propose CRISP (Correlation-filtered Recursive Feature Elimination and Integration of SMOTE Pipeline for Gait-Based Parkinson's Disease Screening), a lightweight multistage framework that sequentially applies correlation-based feature pruning, recursive feature elimination (RFE), and Synthetic Minority Oversampling Technique (SMOTE) based class balancing. To ensure clinically meaningful evaluation, a novel subject-wise protocol was also introduced that assigns one prediction per individual enhancing patient-level variability capture and better aligning with diagnostic workflows. Using 306 VGRF recordings (93 PD, 76 controls), five classifiers, i.e., k-Nearest Neighbours (KNN), Decision Tree (DT), Random Forest (RF), Gradient boosting (GB), and Extreme Gradient Boosting (XGBoost) were evaluated for both binary PD detection and multiclass severity grading. RESULTS: CRISP consistently improved performance across all models under 5-fold cross-validation. XGBoost achieved the highest performance, increasing subject-wise PD detection accuracy from 96.1 ± 0.8% to 98.3 ± 0.8%, and severity grading accuracy from 96.2 ± 0.7% to 99.3 ± 0.5%. CONCLUSION: CRISP is the first VGRF-based pipeline to combine correlation-filtered feature pruning, recursive feature elimination, and SMOTE to enhance PD detection performance, while also introducing a subject-wise evaluation protocol that captures patient-level variability for truly personalized diagnostics. These twin novelties deliver clinically significant gains and lay the foundation for real-time, on-device PD detection and severity monitoring.
Circuit-level modeling of prediction error computation of multi-dimensional features in voluntary actions.
Huang Y, Li Y
Front Comput Neurosci · 2025 · PMID 41089074 · Full text
INTRODUCTION: Predictive processing posits that the brain minimizes discrepancies between internal predictions and sensory inputs, offering a unifying account of perception, cognition, and action. In voluntary actions, i...
INTRODUCTION: Predictive processing posits that the brain minimizes discrepancies between internal predictions and sensory inputs, offering a unifying account of perception, cognition, and action. In voluntary actions, it is thought to suppress self-generated sensory outcomes. Although sensory mismatch signals have been extensively investigated and modeled, mechanistic insights into the neural computation of predictive processing in voluntary actions remain limited. METHODS: We developed a computational model comprising two-compartment excitatory pyramidal cells (PCs) and three major types of inhibitory interneurons with biologically realistic connectivity. The model incorporates experience-dependent inhibitory plasticity and feature selectivity to shape excitation-inhibition (E/I) balance. We then extended it to a two-dimensional prediction-error (PE) circuit in which each PC has two segregated, top-down modulated dendrites-each bell-tuned to a distinct feature-enabling combination selectivity. RESULTS: The model reveals that top-down predictions can selectively suppress PCs with matching feature selectivity via experience-dependent inhibitory plasticity. This suppression depends on the response selectivity of inhibitory interneurons and on balanced excitation and inhibition across multiple pathways. The framework also accommodates predictions involving two independent features. DISCUSSION: By combining biological connectivity data with computational modeling, this study provides insights into the neural circuits and computations underlying the active suppression of sensory responses in voluntary actions. These findings contribute to understanding how the brain generates and processes predictions to guide behavior.
Effects of AC induced electric fields on neuronal firing sensitivity and activity patterns.
Yuan C, Chen R, Li X … +1 more , Zhao Y
Front Comput Neurosci · 2025 · PMID 41049356 · Full text
INTRODUCTION: Understanding how neurons respond to time-varying electric fields is essential for both basic neuroscience and the development of neuromodulation strategies. However, the mechanisms by which alternating-cur...
INTRODUCTION: Understanding how neurons respond to time-varying electric fields is essential for both basic neuroscience and the development of neuromodulation strategies. However, the mechanisms by which alternating-current induced electric fields (AC-IEF) influence neuronal sensitivity and firing remain unclear. METHODS: We developed a modified two-compartment Pinsky-Rinzel (PR) neuron model incorporating AC-IEF stimulation. Using systematic simulations, we examined firing responses across a wide range of field frequencies, amplitudes, and intrinsic membrane parameters, including inter-compartmental conductance and potassium reversal potential. RESULTS: Neurons exhibited no firing or sensitivity when the field amplitude was less than twice the baseline membrane potential, regardless of conductance or reversal potential. Sensitivity increased markedly with amplitude: for example, when the amplitude exceeded 0.5 mV/cm, maximum firing rates rose by up to 45% and the sensitivity frequency range extended to 10-50 Hz. Phase-locking phenomena (1:1 and 2:1) were observed, with bandwidths widening as amplitude increased. For amplitudes below 30 mV, firing pattern transitions depended strongly on inter-compartmental conductance, whereas amplitudes ≥30 mV produced a consistent progression ending in subthreshold oscillations. Similar parameter-dependent transitions occurred for different potassium reversal potentials, converging at high amplitudes. DISCUSSION: These results reveal a parameter-dependent mechanism by which AC-IEF modulate neuronal excitability. The findings provide qualitative rather than strictly quantitative insights into how external electromagnetic environments can shape neural activity, offering new directions for targeted neuromodulation in both health and disease.
About
- Frequency
- Sun
- Papers found
- 200
- RSS feed
- Subscribe