Frontiers in Human Neuroscience最新文献

Objective: This pilot study aimed to evaluate the feasibility and tolerability of motor imagery (MI)-based brain-computer interface (BCI) training with robotic hand assistance for upper limb rehabilitation, and to explore preliminary neural markers in ischemic stroke patients.

Methods: Three post-stroke participants performed MI tasks combined with exoskeleton-assisted movements to facilitate rehabilitation training. Electroencephalography (EEG) signals were recorded to assess the neural correlates of MI. Functional outcomes were evaluated using standard assessment tools.

Results: Our results demonstrated significant improvements in motor function across all participants. Additionally, EEG analysis revealed event-related desynchronization (ERD) in the high-alpha band power at motor cortex locations, with individual differences in both the frequency and power of neural activity. However, no significant trends in neural activity were observed across the training sessions.

Conclusion: These findings suggest that MI-based BCI training, combined with robotic assistance, offer a promising approach for enhancing upper limb function in ischemic stroke patients.

Introduction: Native and non-native listeners rely on different neural strategies when processing speech in their respective native and non-native languages, encoding distinct features of speech from acoustic to linguistic content in different ways. This study investigated differences in neural responses between native English and Korean speaker when they passively listened to speech in their native and non-native languages using electroencephalography.

Methods: The study employed two approaches to examine neural responses: Temporal Response Functions (TRFs) measure how the brain tracks continuous speech features (i.e., speech envelope, phoneme onset, phonemic surprisal, and semantic dissimilarity), and Phoneme-Related Potentials (PRPs) assess phonemic-level processes.

Results: Non-native speakers showed significantly stronger neural tracking of the speech envelope, but no group differences for higher-level linguistic features within analyses of TRFs. PRP analyses, however, revealed distinct response patterns across phoneme categories, with non-native speakers showing heightened peaks.

Conclusion: The results suggest that non-native speakers rely more on bottom-up acoustic cues during passive listening. TRFs and PRPs provide information on neural markers that indicate how speech is processed differently depending on the listener's native language and language experience.

Trauma is the fourth leading cause of death globally and the primary cause of mortality in the 15-45 age group, with traumatic brain injury (TBI) at the core of trauma care. Annually, over 50 million TBI patients are reported worldwide. The complex and heterogeneous pathophysiology of TBI presents substantial diagnostic and therapeutic challenges. In recent years, multimodal monitoring has emerged as a crucial tool to guide clinical management. The integration of multimodal monitoring with machine learning offers novel opportunities for TBI assessment and management, given the rapid development and widespread application of machine learning approaches. Therapeutic hypothermia has shown potential neuroprotective benefits in experimental and clinical contexts, though evidence remains mixed and its implementation in practice faces significant challenges. This review summarizes recent advancements in multimodal monitoring and explores how machine learning can optimize the application of therapeutic hypothermia in conjunction with multimodal data. For example, predictive models trained on multimodal signals (e.g., EEG, ICP, cerebral blood flow, and oxygenation) can help identify patient subgroups most likely to benefit from targeted temperature management. By enabling such stratification and adaptive treatment strategies, machine learning may support the development of more personalized and effective therapeutic approaches for TBI.

Background: Inner speech-the covert articulation of words in one's mind-is a fundamental phenomenon in human cognition with growing interest across BCI. This pilot study evaluates and compares deep learning models for inner-speech classification using non-invasive EEG derived from a bimodal EEG-fMRI dataset (4 participants, 8 words). The study assesses a compact CNN (EEGNet) and a spectro-temporal Transformer using leave-one-subject-out validation, reporting accuracy. Macro-F₁, precision, and recall.

Objective: This study aims to evaluate and compare deep learning models for inner speech classification using non-invasive electroencephalography (EEG) data, derived from a bimodal EEG-fMRI dataset. The goal is to assess the performance and generalizability of two architectures: the compact convolutional EEGNet and a novel spectro-temporal Transformer.

Methods: Data were obtained from four healthy participants who performed structured inner speech tasks involving eight target words. EEG signals were preprocessed and segmented into epochs for each imagined word. EEGNet and Transformer models were trained using a leave-one-subject-out (LOSO) cross-validation strategy. Performance metrics included accuracy, macro-averaged F₁ score, precision, and recall. An ablation study examined the contribution of Transformer components, including wavelet decomposition and self-attention mechanisms.

Results: The spectro-temporal Transformer achieved the highest classification accuracy (82.4%) and macro-F₁ score (0.70), outperforming both the standard and improved EEGNet models. Discriminative power was also substantially improved by using wavelet-based time-frequency features and attention mechanisms. Results showed that confusion patterns of social word categories outperformed those of number concepts, corresponding to different mental processing strategies.

Conclusion: Deep learning models, in particular attention-based Transformers, demonstrate great promise in decoding internal speech from EEG. These findings lay the groundwork for non-invasive, real-time BCIs for communication rehabilitation in severely disabled patients. Future work will take into account vocabulary expansion, wider participant variety, and real-time validation in clinical settings.

Introduction: Postural control impairments are common in patients with multiple sclerosis (MS), resulting in postural instability and increased fall risk. Sensory inputs are crucial to maintain balance adequately. Additionally, fatigue is one of the common and most disabling symptoms of MS, possibly contributing to postural deficits. Previous studies have examined the effects of fatigue and altered sensory conditions on postural control in patients with MS. The present study aimed to extend this knowledge by jointly assessing these factors within the same experimental framework, providing additional insight into how fatigue modulates sensory contributions to balance.

Methods: A total of 21 patients with MS (age = 41.1 ± 10.1 years; EDSS = 1.9 ± 1.0; disease duration = 6.8 ± 4.9 years) completed balance assessments on firm and compliant surfaces with both eyes open and eyes closed, before and after a 6-min walk test used to induce fatigue. Postural sway was quantified using sway velocity and root mean square (RMS).

Results: There was a significant effect of surface on sway velocity (p < 0.001, η² = 0.60), with a greater sway on the compliant surface compared to the firm surface. Fatigue significantly increased sway RMS (p = 0.023, η² = 0.23) but did not affect sway velocity (p > 0.05). The absence of visual input (eyes closed) also significantly increased sway RMS (p = 0.001, η² = 0.46). There was a significant interaction between surface and vision for sway RMS (p < 0.001, η² = 0.54), with a larger effect of surface instability in the eyes-closed condition.

Discussion: Patients with MS face increased challenges in maintaining postural control under conditions of fatigue, surface instability, and lack of visual input. Sway RMS may be more sensitive to these effects than sway velocity.

India is a profoundly diverse nation-state where constitutional provisions for two official and 22 Scheduled languages overlay a vast substratum of numerous Non-Scheduled languages and over a thousand distinct mother tongues, creating a rich and layered linguistic hierarchy. The trajectory of linguistic growth of young Indians is invariably multilingual and multidialectal and involves at least one L1, followed most often by Hindi and English or other languages. We carried out a qualitative study with over over a thousand students entering university using a modified LEAP questionnaire for self-assessment of fluency, literacy, domain of use, and time course of language acquisition and loss. We explore the interaction between heritage language, multilingualism, and the formal (trilingual) education policy and show that they intersect to redraw the linguistic profiles of individuals with shifting language dominance, and impact linguistic ability, especially in L1. We also find that the understanding of "heritage language" needs to be more nuanced in this particular context of multilingualism and language acquisition.

Background: Recent studies suggest that amplitude-modulated transcranial alternating current stimulation (AM-tACS) may enhance cognitive functions, but its mechanisms and optimal application remain unclear.

Methods: Thirty-three healthy university students were randomly assigned to Sham, tACS (40 Hz, 1 mA, bilateral prefrontal cortex), or AM-tACS (200 Hz carrier frequency) groups, in AM-tACS, the baseband modulation frequency was individualized based on the pre-task phase-locking value (PLV) derived from occipitofrontal EEG. Working memory (WM) was assessed via a delayed-match-to-sample task (accuracy and sensitivity index d').

Results: Compared to Sham, the tACS group showed significant WM accuracy improvement (p < 0.05). AM-tACS exhibited a smaller but statistically significant enhancement in d' (p < 0.05). EEG analysis revealed no PLV increase between stimulated regions, but a trend toward heightened frontal-occipital functional connectivity.

Conclusion: Amplitude-modulated transcranial alternating current stimulation effectively enhances WM in college students, though physiological mechanisms require further investigation with multimodal approaches. The compatibility of AM-tACS with real-time EEG monitoring highlights its potential for closed-loop neuromodulation systems, where stimulation parameters could be dynamically adjusted based on neural feedback.

Introduction: Aerobic exercise mitigates symptoms of Parkinson's disease (PD) and may slow disease progression; however, the neural mechanisms underlying these improvements are not well understood. In this study, we discuss the methodology for simultaneously recording local field potentials (LFP) from the subthalamic nucleus (STN), cortical activity using scalp electroencephalography (EEG), and exercise performance metrics during a 40-min aerobic cycling session. Data from a single patient with PD are presented to illustrate the utility, feasibility, and data integrity of the experimental set up.

Methods: The Medtronic Percept™ DBS system was used to record and stream bilateral STN-LFP in the OFF-therapy condition (OFF-DBS and OFF-antiparkinson medications) during a 40-min aerobic exercise session. A 64-channel mobile EEG system recorded cortical data. The neural data streams were synchronized using a TENS device that injected a specified electrical signal into the EEG and LFP recordings. Exercise performance metrics, heart rate, cadence, and power were synchronized with neural data and collected during the exercise session. The study is registered on ClinicalTrials.gov, trial identifying numbers NCT05905302 and NCT05972759.

Results: STN-LFP, EEG, and exercise performance data can be synchronized, recorded for more than 40 min, and analyzed to evaluate how aerobic exercise impacts patterns of cortical and subcortical neural activity.

Conclusion: While exercise positively affects symptoms of PD, the precise effects of exercise on network activity remain unclear. The methods utilized for collecting and analyzing neural (cortical and subcortical) and exercise-related data during a typical bout of aerobic exercise suggest that this approach can be adopted for larger, long-term exercise studies in patients with PD and deep brain stimulation (DBS). The described protocol provides a roadmap for future projects aiming to combine STN-LFP and cortical data to better understand how exercise may alter cortico-basal-ganglia-thalamic dynamics in PD.

Reading is a crucial human skill and learning novel written word forms is a life-long process. Here, we tracked the emergence of novel word lexical and neural representations after a training procedure, contrasting two learning methods, in 32 monolingual adults. Half of the novel words were provided with orthographic and phonological information (OP), and half with additional explicit semantic information (OPS). At the neural level, we demonstrate for the first time the sensitivity of EEG recordings with fast periodic visual stimulation (FPVS) to track novel visual word learning. We used an oddball paradigm, with base stimuli (pseudowords) displayed at 10 Hz with deviant stimuli (words) every fifth item (at 2 Hz), in which word-selective responses at 2 Hz demonstrate lexical discrimination. While at pre-test, novel words were not discriminated, results show clear word-selective responses over the left occipital-temporal cortex (VOTC) post-learning with both methods. This finding suggests the creation of orthographic representations for novel words and fits with current views that this region is specialized for the rapid recognition and fast learning of novel word forms. Moreover, the behavioral lexical decision data reveal significant increases in reaction times after learning, for novel words' lexical neighbors, which suggests lexical engagement through competition arising from newly formed representation. Contrary to our expectations, no advantage was found for the OPS method. Instead, results show stronger behavioral and neural changes with the OP method. In the discussion, we highlight possible reasons for this unexpected finding. First, the current implementation of the OPS method displaying simultaneous images and words during learning could have dragged the participant's attention away from the orthographic form. Second, the speed of presentation of stimuli might have been too fast to allow fast semantic retrieval. Finally, semantic learning might have a different timeframe than word form learning, and the current findings would reflect only the latter. Our results nevertheless highlight the rapid emergence of new word-form representations, captured by the EEG-FPVS approach.

Introduction: Spatial memory supports orientation and navigation by integrating multiple spatial reference frames. Neuroimaging and lesion studies implicate the hippocampus (HIP) and retrosplenial cortex (RSC), but causal evidence from non-invasive brain stimulation is limited.

Methods: Eighteen participants performed a spatial localization task in a virtual room under three stimulation conditions: anodal transcranial direct current stimulation (tDCS) over the left RSC, anodal tDCS over the left HIP, and sham. Task conditions varied in reference frame (viewer-, object-, room-centered) and perspective shift (0°, 45°, 135°). Accuracy was analyzed with non-parametric statistics.

Results: Performance declined with increasing viewpoint rotation, especially in room-centered trials. RSC stimulation selectively reduced accuracy in room-centered trials with large perspective shifts (135°), whereas HIP stimulation did not significantly modulate performance.

Discussion: Findings provide causal evidence for the involvement of the RSC in viewpoint-invariant spatial updating, supporting its role in integrating stable environmental cues. HIP stimulation yielded no reliable behavioral effects, suggesting functional specificity of the RSC and highlighting the challenges of modulating deep cortical structures with tDCS.