Brain-X最新文献

Sleep disorders are a major public health concern. Electroencephalogram (EEG) analysis offers an effective approach for evaluating neurological status using K-complexes and sleep spindles as key biomarkers for sleep stage classification. This study proposes a hybrid architecture based on Temporal Convolutional Network (TCN) and Kolmogorov-Arnold Network (KAN). The model was evaluated on the DREAMS dataset and achieved a detection accuracy of 93.18% for K-complexes and 86.26% for spindles, outperforming baseline TCN models by 1%–2% across metrics. These results validate KAN's efficacy in time-domain signal classification. Notably, multi-layered KAN configurations failed to yield additional performance gains. Furthermore, we developed a sleep stage identification framework leveraging EEG biomarkers to consolidate physiologically similar stages into three macro-categories, attaining classification accuracies of 79.7% (K-complex subset) and 68.4% (spindle subset) on the DREAMs dataset. Extending this approach, we implemented a five-stage recognition system based on sleep phase duration ratios, which achieved 81.6% accuracy on the Haaglanden Medisch Centrum dataset. This work supports the feasibility of automated sleep staging using characteristic EEG waveforms.

Anxiety during speech is often assumed to differ in type between native (L1) and non-native (L2) language use. We combined continuous self-reports with high-resolution cardiac and electrodermal recordings in 40 Dutch–English bilingual adults who delivered 4-min monologues in both languages, followed by second-by-second replay ratings of felt anxiety and self-perceived proficiency. Across conditions, anxiety emerged as brief episodes (typically 1–4 s) superimposed on slower physiological adaptation. L2 speech elicited higher and more sustained autonomic arousal and a delayed return to baseline relative to L1. However, time-varying and multilevel vector autoregressive models showed that the core psychophysiological network linking felt anxiety, perceived proficiency and physiology was statistically indistinguishable across languages. What is commonly labelled ‘foreign language anxiety’ therefore reflects a context-intensified expression of a shared anxiety mechanism rather than a distinct construct. By mapping second-scale trajectories of subjective and physiological states during naturalistic speech, our findings highlight anxiety as a fluid dynamic system that adapts over tens of seconds yet is built from much shorter micro-episodes. This dynamic, multimodal approach offers a template for probing how contextual demands shape common affective mechanisms in communication across domains.

The substantial symptomatic overlap between unipolar depression, primarily major depressive disorder (MDD), and bipolar disorder complicates clinical diagnosis, underscoring the critical need for objective biomarkers. Resting-state electroencephalography (rsEEG) holds promise for differentiating these two conditions due to its non-invasive nature, high temporal resolution, and clinical feasibility. In this study, we reviewed rsEEG studies that explored spectral power, functional connectivity, nonlinear dynamics, machine learning (ML) applications, and animal models. The key discriminative features identified included specific spatial patterns in the high alpha band, distinctions between periodic and aperiodic components, genetically influenced alpha coherence, and reduced central–parietal alpha phase variability in patients with MDD. While ML models integrating these diverse features achieved high classification accuracy, findings for frequency power and network metrics remained inconsistent across studies. Current limitations, however, include substantial methodological heterogeneity, small sample sizes, a lack of longitudinal data, poor model generalizability, and minimal supporting evidence from animal models. We conclude by outlining these key potential markers, identifying current research gaps, and proposing future directions, specifically emphasizing the adoption of standardized pipelines, large multicenter longitudinal studies, advanced ML techniques, and the critical integration of genetic and animal findings for the development of reliable rsEEG-based diagnostic tools.

In human cerebral hemodynamics, low-frequency oscillations (LFOs) occur due to the sympathetic nervous system and blood pressure regulation. LFOs have recently been reported to be associated with cognitive-behavioral performance, though their functional relevance to cognitive load remains unclear. Here, we employed functional near-infrared spectroscopy to record changes in oxygenated (Δ[oxy-Hb]) and deoxygenated (Δ[deoxy-Hb]) hemoglobin concentrations in the prefrontal cortex (PFC) of 47 healthy young adults (aged 18–23) during N-back working memory tasks and extracted LFOs from the cerebral hemodynamic data to study their relationship with working memory load. Increasing the task load led to a marked decrease in both LFO power and LFO peak amplitude of Δ[oxy-Hb], alongside an increase in LFO peak frequency and PFC activation, revealing a load-dependent feature of cognitive engagement. Correlations between LFO power and behavioral performance, including accuracy and response time, were observed. LFOs and their characteristic parameters exhibited a strong effect on working memory load, indicating the potential of LFOs in cerebral hemodynamics as a sensitive marker for quantifying the cognitive load effect of brain activity.

Serum uric acid (SUA) demonstrates dual roles as both an antioxidant and a pro-oxidant; however, its long-term effects on glymphatic function and cognitive performance remain unclear. This study included 944 participants in a longitudinal cohort to investigate the associations between the temporal progression of SUA levels and diffusion tensor imaging (DTI) analysis along the perivascular space (DTI-ALPS) with cognition. The results revealed that there were negative linear correlations between early average SUA levels and DTI-ALPS index for the left hemisphere (beta = −0.02, 95% confidence interval [CI] −0.04 to 0.00, p = 0.03), and cumulative SUA (cumSUA) levels and DTI-ALPS index for the left hemisphere (beta = −0.02, 95% CI −0.04 to −0.01, p = 0.01) and the whole brain (beta = −0.02, 95% CI −0.04 to 0.00, p = 0.03). A U-shaped relationship was observed between coefficient of variation of early SUA (CVSUA) levels and DTI-ALPS index in the left hemisphere (p = 0.01) and whole brain (p = 0.03). Higher early CVSUA levels, lower early cumSUA levels and reduced DTI-ALPS index were associated with lower MoCA scores. These findings suggest that moderately elevated SUA levels within the normal ranges, as well as stable SUA concentrations, may help mitigate future cognitive decline.

With the establishment of interdisciplinary platforms in medical engineering, the twenty-first century has witnessed rapid advancements in brain‒computer interface (BCI) technology, demonstrating significant potential in clinical applications. The fundamental framework of BCIs encompasses signal acquisition, preprocessing, feature extraction, signal translation, and control devices. This article explores the applications of BCIs across various disorders, including post-stroke motor recovery, consciousness disorders, mental health, and neurological diseases. Through technological innovations, BCI technology has assisted patients in overcoming communication and motor impairments resulting from neurological damage. By bypassing damaged neural pathways and enabling bidirectional interactions with brain signals, BCIs facilitate the gradual rehabilitation of motor functions in post-stroke patients and show promise in treating psychiatric disorders such as depression. Compared with traditional treatment methods, BCI technology has several unique advantages. Despite challenges in signal processing accuracy, hardware stability, and real-time data processing technology, technological innovation and interdisciplinary collaboration facilitate greater breakthroughs in the near future for BCIs. In summary, the clinical application of BCI technology presents unprecedented opportunities and challenges in modern healthcare, underscoring the need for continued research and development in this area.

Once understood in binary terms, gender identity is increasingly recognized as a multidimensional and continuous construct shaped by both sociocultural and neurobiological factors. Although prior studies have reported associations between gender identity and brain structure, few have adopted an integrative approach to examine how gender identity emerges. Drawing on a large, non-clinical sample of young adults from the Amsterdam Open Magnetic Resonance Imaging Collection (n = 544), this study integrated psychological assessments, socioeconomic indicators, and structural MRI to investigate the relationship between gender identity and brain morphology. For participants assigned female at birth, a feminine identity was linked to reduced cortical thickness in several brain regions, including the parahippocampal, fusiform, lingual, and pericalcarine cortices. Among these regions, two distinct pathways related to the fusiform cortex were identified: a self-referential pathway (through the parahippocampal cortex) and a visual-perceptual pathway (through the pericalcarine and lingual cortices). Besides, an additional pathway related to the fusiform cortex was also identified, which connected higher socioeconomic status to crystallized intelligence. For participants assigned male at birth, a feminine identity was associated with increased anxiety and reduced cortical thickness in visual-emotional regions. In contrast, masculine identity was linked to a larger cortical area in the supramarginal gyrus and insula. Altogether, these findings suggest that gender identity is embedded in distributed neural systems that support self-representation, and that its structural correlates emerge through distinct psychological and cognitive-contextual mechanisms. By moving beyond binary classification, this study may offer a more nuanced neurobiological model of gendered self-concept in the general population.

The perception of hazardous information is a crucial factor in ensuring safety in production. In recent years, multi-mode sensing has been proven to be an effective approach for developing efficient perception systems. However, these systems still rely on various combinations of single-function sensors within traditional von Neumann architecture, which increases the system's overall complexity. In this study, carbon-doped black phosphorus (C–BP)-based multi-perception memristors were successfully developed for hazardous information perception. The C–BP multi-perception memristor exhibits remarkable stability and high surface activity due to the coupling and synergistic effects of C doping. Its high surface activity enables the reliable perception of hazardous visual (ultraviolet light) and olfactory (ethanol, acetone, and human expirations) information in an open environment. Consequently, a hazardous detection system based on the C–BP multi-perception memristor was simulated. The results indicate that the developed system outperforms traditional detection systems with an enhanced performance rate (97.6% vs. 90.5%) in perceiving hazardous information. This work may provide new insights into developing enhanced-performance hazardous information perception systems.

Brain-computer interface (BCI) technology aims to create a connection pathway for exchanging information between the brain and devices with computing capabilities. This technology has become a global research focus, and many countries and regions are working to establish a BCI industry. BCIs have many potential applications, especially in the medical field. However, the complexities of non-invasive BCIs and the implantation risks associated with invasive BCIs have limited these technologies to laboratory settings. The main challenges for the practical implementation of BCIs include the lack of foundational technologies for non-invasive and invasive BCIs, the signal processing challenges associated with BCIs, the key components of BCIs, and the compatibility of BCI software and hardware. These shortcomings should be addressed to enhance the competitiveness of BCI products and promote the application of BCIs in medicine. In the future, if novel methods for acquiring or decoding neural signals are developed that enable non-invasive BCIs to achieve signal quality comparable to that of invasive techniques, it will propel BCI technology to leapfrog in development. Technological breakthroughs will enable BCIs to enhance medical technology and improve people's quality of life.