Frontiers in Computational Neuroscience最新文献

Contactless Human Activity Recognition (HAR) has played a critical role in smart healthcare and elderly care homes to monitor patient behavior, detect falls or abnormal activities in real time. The effectiveness of non-invasive HAR is often hindered by location-centric variations in Channel State Information (CSI). These variations limit the ability of HAR models to generalize across new unseen cross-domain environments, for instance, a model trained in one location might not perform well in another physical location. To address this challenge, in this study, we present a novel federated learning (FL) algorithm designed to train a robust global model from local datasets in different localizations. The proposed Federated Weighted Averaging for HAR (Fed-WAHAR) algorithm mitigates location-induced disparities, including heterogeneity and non-Independent and Identically Distributed (non-IID) data distributions. Fed-WAHAR employs a dynamic weighting approach based on local models' accuracy to improve global model classification accuracy and reduce convergence time effectively. We evaluated the performance of Fed-WAHAR using various metrics, including accuracy, precision, recall, F1 score, confusion matrix, and convergence analysis. Experimental results demonstrate that Fed-WAHAR achieves an accuracy of 85% in recognizing human activities across different locations, enhancing the ability of model to infer across new unseen locations.

Understanding how language and linguistic constructions are processed in the brain is a fundamental question in cognitive computational neuroscience. This study builds directly on our previous work analyzing Argument Structure Constructions (ASCs) in the BERT language model, extending the investigation to a simpler, brain-constrained architecture: a recurrent neural language model. Specifically, we explore the representation and processing of four ASCs-transitive, ditransitive, caused-motion, and resultative-in a Long Short-Term Memory (LSTM) network. We trained the LSTM on a custom GPT-4-generated dataset of 2,000 syntactically balanced sentences. We then analyzed the internal hidden layer activations using Multidimensional Scaling (MDS) and t-Distributed Stochastic Neighbor Embedding (t-SNE) to visualize sentence representations. The Generalized Discrimination Value (GDV) was calculated to quantify cluster separation. Our results show distinct clusters for the four ASCs across all hidden layers, with the strongest separation observed in the final layer. These findings are consistent with our earlier study based on a large language model and demonstrate that even relatively simple RNNs can form abstract, construction-level representations. This supports the hypothesis that hierarchical linguistic structure can emerge through prediction-based learning. In future work, we plan to compare these model-derived representations with neuroimaging data from continuous speech perception, further bridging computational and biological perspectives on language processing.

EEG emotion recognition has important applications in human-computer interaction and mental health assessment, but existing models have limitations in capturing the complex spatial and temporal features of EEG signals. To overcome this problem, we propose an innovative model that combines CNN-BiLSTM and DC-IGN and fused both outputs for sentiment classification via a fully connected layer. In addition, we use a piecewise exponential decay strategy to optimize the training process. We conducted a comprehensive comparative experiment on the SEED and DEAP datasets, it includes traditional models, existing advanced models, and different combination models (such as CNN + LSTM, CNN + LSTM+DC-IGN). The results show that our model achieves 94.35% accuracy on SEED dataset, 89.84% on DEAP-valence, 90.31% on DEAP-arousal, which is significantly better than other models. In addition, we further verified the superiority of the model through subject independent experiment and learning rate scheduling strategy comparison experiment. These results not only improve the performance of EEG emotion recognition, but also provide new ideas and methods for research in related fields, and prove the significant advantages of our model in capturing complex features and improving classification accuracy.

Mathematical analysis of biological neural networks, specifically inhibitory networks with all-to-all connections, is challenging due to their complexity and non-linearity. In examining the dynamics of individual neurons, many fast currents are involved solely in spike generation, while slower currents play a significant role in shaping a neuron's behavior. We propose a discrete map approach to analyze the behavior of inhibitory neurons that exhibit bursting modulated by slow calcium currents, leveraging the time-scale differences among neural currents. This discrete map tracks the number of spikes per burst for individual neurons. We compared the map's predictions for the number of spikes per burst and the long-term system behavior to data obtained from the continuous system. Our findings demonstrate that the discrete map can accurately predict the canonical behavioral signatures of bursting performance observed in the continuous system. Specifically, we show that the proposed map a) accounts for the dependence of the number of spikes per burst on initial calcium levels, b) explains the roles of individual currents in shaping the system's behavior, and c) can be explicitly analyzed to determine fixed points and assess their stability.

Major depressive disorder (MDD) is one of the most common mental disorders, with significant impacts on many daily activities and quality of life. It stands as one of the most common mental disorders globally and ranks as the second leading cause of disability. The current diagnostic approach for MDD primarily relies on clinical observations and patient-reported symptoms, overlooking the diverse underlying causes and pathophysiological factors contributing to depression. Therefore, scientific researchers and clinicians must gain a deeper understanding of the pathophysiological mechanisms involved in MDD. There is growing evidence in neuroscience that depression is a brain network disorder, and the use of neuroimaging, such as magnetic resonance imaging (MRI), plays a significant role in identifying and treating MDD. Rest-state functional MRI (rs-fMRI) is among the most popular neuroimaging techniques used to study MDD. Deep learning techniques have been widely applied to neuroimaging data to help with early mental health disorder detection. Recent years have seen a rise in interest in graph neural networks (GNNs), which are deep neural architectures specifically designed to handle graph-structured data like rs-fMRI. This research aimed to develop an ensemble-based GNN model capable of detecting discriminative features from rs-fMRI images for the purpose of diagnosing MDD. Specifically, we constructed an ensemble model by combining functional connectivity features from multiple brain region segmentation atlases to capture brain complexity and detect distinct features more accurately than single atlas-based models. Further, the effectiveness of our model is demonstrated by assessing its performance on a large multi-site MDD dataset. We applied the synthetic minority over-sampling technique (SMOTE) to handle class imbalance across sites. Using stratified 10-fold cross-validation, the best performing model achieved an accuracy of 75.80%, a sensitivity of 88.89%, a specificity of 61.84%, a precision of 71.29%, and an F1-score of 79.12%. The results indicate that the proposed multi-atlas ensemble GNN model provides a reliable and generalizable solution for accurately detecting MDD.

Introduction: Preserving privacy is a critical concern in medical imaging, especially in resource limited settings like smart devices connected to the IoT. To address this, a novel encryption method for medical images that operates at the bit plane level, tailored for IoT environments, is developed.

Methods: The approach initializes by processing the original image through the Secure Hash Algorithm (SHA) to derive the initial conditions for the Chen chaotic map. Using the Chen chaotic system, three random number vectors are generated. The first two vectors are employed to shuffle each bit plane of the plaintext image, rearranging rows and columns. The third vector is used to create a random matrix, which further diffuses the permuted bit planes. Finally, the bit planes are combined to produce the ciphertext image. For further security enhancement, this ciphertext is embedded into a carrier image, resulting in a visually secured output.

Results: To evaluate the effectiveness of our algorithm, various tests are conducted, including correlation coefficient analysis (C.C < or negative), histogram analysis, key space [(10⁹⁰)⁸] and sensitivity assessments, entropy evaluation [E(S) > 7.98], and occlusion analysis.

Conclusion: Extensive evaluations have proven that the designed scheme exhibits a high degree of resilience to attacks, making it particularly suitable for small IoT devices with limited processing power and memory.

Introduction: Emerging evidence suggests that different metabolic characteristics, particularly bioenergetic differences, between the synaptic terminal and soma may contribute to the selective vulnerability of dopaminergic neurons in patients with Parkinson's disease (PD).

Method: To investigate the metabolic differences, we generated four thermodynamically flux-consistent metabolic models representing the synaptic and non-synaptic (somatic) components under both control and PD conditions. Differences in bioenergetic features and metabolite exchanges were analyzed between these models to explore potential mechanisms underlying the selective vulnerability of dopaminergic neurons. Bioenergetic rescue analyses were performed to identify potential therapeutic targets for mitigating observed energy failure and metabolic dysfunction in PD models.

Results: All models predicted that oxidative phosphorylation plays a significant role under lower energy demand, while glycolysis predominates when energy demand exceeds mitochondrial constraints. The synaptic PD model predicted a lower mitochondrial energy contribution and higher sensitivity to Complex I inhibition compared to the non-synaptic PD model. Both PD models predicted reduced uptake of lysine and lactate, indicating coordinated metabolic processes between these components. In contrast, decreased methionine and urea uptake was exclusively predicted in the synaptic PD model, while decreased histidine and glyceric acid uptake was exclusive to the non-synaptic PD model. Furthermore, increased flux of the mitochondrial ornithine transaminase reaction (ORNTArm), which converts oxoglutaric acid and ornithine into glutamate-5-semialdehyde and glutamate, was predicted to rescue bioenergetic failure and improve metabolite exchanges for both the synaptic and non-synaptic PD models.

Discussion: The predicted differences in ATP contribution between models highlight the bioenergetic differences between these neuronal components, thereby contributing to the selective vulnerability observed in PD. The observed differences in metabolite exchanges reflect distinct metabolic patterns between these neuronal components. Additionally, mitochondrial ornithine transaminase was predicted to be the potential bioenergetic rescue target for both the synaptic and non-synaptic PD models. Further research is needed to validate these dysfunction mechanisms across different components of dopaminergic neurons and to explore targeted therapeutic strategies for PD patients.

Neural oscillation, particularly gamma oscillation, are fundamental to cognitive processes such as attention, perception, and decision-making. Experimental studies have shown that the phase of gamma oscillation modulates neuronal response selectivity, suggesting a direct link between oscillatory dynamics and cognition. However, there remains a lack of computational models that can systematically simulate and investigate this effect. To address this, we construct a low-rank spiking neural network (low-rank SNN) based on the voltage-dependent theta model to explore how structured connectivity shapes oscillatory dynamics and cognitive function. Using macroscopic model analysis, we identify different network states, ranging from stationary firing to gamma oscillation. Our model successfully reproduces phase-dependent response modulation in a Go-Nogo task, consistent with in vivo findings, providing an explanation for how neural oscillation influences task performance. Besides phase dependency, our findings suggest that gamma oscillation can enhance and prolong signal response. Compared to prior studies that applied low-rank connectivity to SNNs but remained limited to stationary or weak oscillatory regimes, our work extends to population-level synchronous activity while maintaining biological plausibility under Dale's principle. Our study offers a theoretical framework for understanding how neural oscillations emerge in structured spiking networks and provides a foundation for future experimental and computational investigations into oscillatory modulation of cognition.

Background: In response to the shortcomings of the current Alzheimer's disease (AD) early populations assessment, which is based on neuropsychological scales with high subjectivity, low accuracy of repeated measurements, tedious process and dependence on physicians, it was found that digital biomarkers based on the writing process can effectively characterize the cognitive deficits of patients with mild cognitive impairment (MCI) due to AD.

Methods: This study designed a digital writing assessment paradigm, extracted dynamic handwriting and image data during the paradigm assessment process, and analyzed digital biomarkers of the writing process to assess subjects' cognitive functions. A total of 72 subjects, including 34 health controls (HC) and 38 MCI due to AD, were enrolled in this study.

Results: Their combined screening efficacy of digital biomarkers based on the MCI writing process due to AD populations having an area under curve (AUC) of 0.918, and a confidence interval (CI) of 0.854-0.982, was higher than the Montreal Cognitive Assessment Scale (AUC = 0.859, CI = 0.772-0.947) and the Mini-mental State Examination Scale (AUC = 0.783, CI = 0.678-0.888).

Conclusion: Therefore, digital biomarkers based on the writing process can characterize and quantify the cognitive function of MCI due to AD populations at a fine-grained level, which is expected to be a new method for intelligent screening and early warning of early AD populations in a community-based physician-free setting.

[This corrects the article DOI: 10.3389/fncom.2024.1418280.].