工程技术最新文献

Fatigue-induced incidents in transportation, aerospace, military, and other areas have been on the rise, posing a threat to human life and safety. The determination of fatigue states holds significant importance, especially through reliable and conveniently available physiological indicators. Here, a portable custom-built fNIRS system was used to monitor the fatigue state caused by nap deprivation. fNIRS signals in ten channels at the prefrontal cortex were collected, changes in blood oxygen concentration were analyzed, followed by a deep learning model to classify fatigue states. For the high-dimensionality and multi-channel characteristics of the fNIRS signal data, a novel 1D revised CNN-ResNet network was proposed based on the double-layer channel attenuation residual block. The results showed a 97.78% accuracy in fatigue state classification, significantly superior than several conventional methods. Furthermore, a fatigue-arousal experiment was designed to explore the feasibility of forced arousal of fatigued subjects through exercise stimulation. The fNIRS results showed a significant increase in brain activity with the conduction of exercise. The proposed method serves as a reliable tool for the evaluation of fatigue states, potentially reducing fatigue-induced harms and risks.

Enhancing the accuracy of emotion recognition models through multimodal learning is a common approach. However, challenges such as insufficient modal feature learning in multimodal inference and scarcity of sample data continue to pose obstacles that need to be overcome. Therefore, we propose a novel adaptive lightweight multimodal efficient feature inference network (ALME-FIN). We introduce a time-domain lightweight adaptive network (TDLAN) and a two-dimensional dynamic focusing network (TDDFN) for multimodal feature learning. The TDLAN incorporates the denoising process as an integral part of network training, achieving adaptive denoising for each sample through the continuous optimization of the trainable filtering threshold. Simultaneously, it incorporates an interactive convolutional sampling module, enabling lightweight multi-scale feature extraction in the time domain. TDDFN effectively extracts core image features while filtering out redundancies. During the training process, the Multi-network dynamic gradient adjustment framework (MDGAF) dynamically monitors the feature learning efficacy across different modalities. It timely adjusts the training gradients of networks to allocate additional optimization time for under-optimized modalities, thereby maximizing the utilization of multimodal feature information. Moreover, the introduction of a Multi-class relationship interaction module prior to the classifier aids the model in clearly understanding the relationships among different category samples. This approach enables the model to achieve relatively accurate emotion recognition even in scenarios of limited sample availability. Compared to existing multimodal learning techniques, ALME-FIN exhibits a more efficient multimodal feature inference method that can achieve satisfactory emotional recognition performance even with a limited number of samples.

To deploy Electroencephalogram (EEG) based Mental Workload Recognition (MWR) systems in the real world, it is crucial to develop general models that can be applied across subjects. Previous studies have utilized domain adaptation to mitigate inter-subject discrepancies in EEG data distributions. However, they have focused on reducing global domain discrepancy, while neglecting local workload-categorical domain divergence. This degrades the workload-discriminating ability of subject-invariant features. To deal with this problem, we propose a novel joint category-wise and domain-wise alignment Domain Adaptation (cdaDA) algorithm, using bi-classifier learning and domain discriminative adversarial learning. The bi-classifier learning approach is adopted to address the similarities and differences between categories, helping to align EEG data within the same mental workload categories. Additionally, the domain discriminative adversarial learning technique is adopted to consider global domain information by minimizing global domain discrepancy. By integrating both local category information and global domain information, the cdaDA model performs a coarse-to-fine alignment and achieves promising cross-subject MWR results.

Epilepsy is a neurological disorder in which complex electrophysiological processes are closely linked to inherent nonlinear kinetic properties. This study investigates the effects of sinusoidal sensory stimulation bias and time-delay on the dynamics of epileptic seizures within a corticothalamic neural network model. The results indicate that an increase in sensory stimulation bias can prematurely terminate seizures, and high-frequency stimulation can induce a phenomenon of frequency resonance. Meanwhile, discharge states transitions are associated with the emergence of bifurcation points. Time-delay exerts a significant regulatory influence on pathways with delay embedding (I2-PY), whereas its impact on pathways without delay embedding (I1-I1 and thalamic relay nucleus (TC)-I2) is negligible. Under sinusoidal sensory stimulation, the responses of three pathways (I1-I1, I1-PY, and I2-PY) associated with inhibitory interneurons reveal that the inhibitory properties of interneurons can suppress seizures; however, an excessively strong inhibitory effect may also precipitate seizures and facilitate state transitions. These findings contribute to a deeper understanding of seizure dynamics and may guide future research in the transmission and evolution of seizures.

Chitosan nanoparticles have been extensively utilised as polymeric drug carriers in nanoparticles formulations due to their potential to enhance drug delivery, efficacy, and safety. Numerous toxicity studies have been previously conducted to assess the safety profile of chitosan-based nanoparticles. These toxicity studies employed various methodologies, including test animals, interventions, and different routes of administration. This review aims to summarise research on the safety profile of chitosan-based nanoparticles in drug delivery, with a focus on general toxicity tests to determine LD50 and NOAEL values. It can serve as a repository and reference for chitosan-based nanoparticles, facilitating future research and further development of drugs delivery system using chitosan nanoparticles. Publications from 2014 to 2024 were obtained from PubMed, Scopus, Google Scholar, and ScienceDirect, in accordance with the inclusion and exclusion criteria.The ARRIVE 2.0 guidelines were employed to evaluate the quality and risk-of-bias in the in vivo toxicity studies. The results demonstrated favourable toxicity profiles, often exhibiting reduced toxicity compared to free drugs or substances. Acute toxicity studies consistently reported high LD50 values, frequently exceeding 5000 mg/kg body weight, while subacute studies typically revealed no significant adverse effects. Various routes of administration varied, including oral, intravenous, intraperitoneal, inhalation, and topical, each demonstrating promising safety profiles.

Rumination thinking is a type of negative repetitive thinking, a tendency to constantly focus on the causes, consequences and other aspects of negative events, which has implications for a variety of psychiatric disorders. Previous studies have confirmed a strong association between rumination thinking and poor sleep or insomnia, but the direction of causality between the two is not entirely clear. This study examined the relationship between rumination thinking and sleep quality using a longitudinal approach and resting-state functional MRI data. Participants were 373 university students (males: n = 84, 18.67 ± 0.76 years old) who completed questionnaires at two time points (T1 and T2) and had resting-state MRI data collected. The results of the cross-lagged model analysis revealed a bidirectional causal relationship between rumination thinking and sleep quality. Additionally, the functional connectivity (FC) of the precuneus and lingual gyrus was found to be negatively correlated with rumination thinking and sleep quality. Furthermore, mediation analysis showed that rumination thinking at T1 fully mediated the relationship between FC of the precuneus-lingual and sleep quality at T2. These findings suggest that rumination thinking and sleep quality are causally related in a bidirectional manner and that the FC of the precuneus and lingual gyrus may serve as the neural basis for rumination thinking to predict sleep quality. Overall, this study provides new insights for enhancing sleep quality and promoting overall health.

Supplementary information: The online version contains supplementary material available at 10.1007/s11571-025-10223-3.

Gastric cancer remains one of the deadliest cancers globally due to delayed detection and limited treatment options, underscoring the critical need for innovative prognostic methods. Disulfidptosis, a recently discovered programmed cell death triggered by disulphide stress, presents a fresh avenue for therapeutic exploration. This research examines disulfidptosis-related long noncoding RNAs (DRLs) in gastric cancer, with the goal of leveraging these lncRNAs as potential markers to enhance patient outcomes and treatment approaches. Comprehensive genomic and clinical data from stomach adenocarcinoma (STAD) were obtained from The Cancer Genome Atlas (TCGA). Employing least absolute shrinkage and selection operator (LASSO) regression analysis, a prognostic model was devised incorporating five key DRLs to forecast survival rates. The effectiveness of this model was validated using Kaplan-Meier survival plots, receiver operating characteristic (ROC) curves, and extensive functional enrichment studies. The importance of select lncRNAs and the expression variability of genes tied to disulfidptosis were validated via quantitative real-time PCR (qRT-PCR) and Western blot tests, establishing a solid foundation for their prognostic utility. Analyses of functional enrichment and tumour mutation burden highlighted the biological importance of these DRLs, connecting them to critical cancer pathways and immune responses. These discoveries broaden our comprehension of the molecular framework of gastric cancer and bolster the development of tailored treatment plans, highlighting the substantial role of DRLs in clinical prognosis and therapeutic intervention.

The correlations of several fundamental properties of human brain connections are investigated in a consensus connectome, constructed from 1064 braingraphs, each on 1015 vertices, corresponding to 1015 anatomical brain areas. The properties examined include the edge length, the fiber count, or edge width, meaning the number of discovered axon bundles forming the edge and the occurrence number of the edge, meaning the number of individual braingraphs where the edge exists. By using our previously published robust braingraphs at https://braingraph.org, we have prepared a single consensus graph from the data and compared the statistical similarity of the edge occurrence numbers, edge lengths, and fiber counts of the edges. We have found a strong positive Spearman correlation between the edge occurrence numbers and the fiber count numbers, showing that statistically, the most frequent cerebral connections have the largest widths, i.e., the fiber count. We have found a negative Spearman correlation between the fiber lengths and fiber counts, showing that, typically, the shortest edges are the widest or strongest by their fiber counts. We have also found a negative Spearman correlation between the occurrence numbers and the edge lengths: it shows that typically, the long edges are infrequent, and the frequent edges are short.

Brainprint recognition technology, regarded as a promising biometric technology, encounters challenges stemming from the time-varied, low signal-to-noise ratio of brain signals, such as electroencephalogram (EEG). Steady-state visual evoked potentials (SSVEP) exhibit high signal-to-noise ratio and frequency locking, making them a promising paradigm for brainprint recognition. Consequently, the extraction of time-invariant identity information from SSVEP EEG signals is essential. In this paper, we propose an Attentive Multi-sub-band Depth Identity Embedding Learning Network for stable cross-session SSVEP brainprint recognition. To address the issue of low recognition accuracy across sessions, we introduce the Sub-band Attentive Frequency mechanism, which integrates the frequency-domain relevant characteristics of the SSVEP paradigm and focuses on exploring depth-frequency identity embedding information. Also, we employ Attentive Statistic Pooling to enhance the stability of frequency domain feature distributions across sessions. Extensive experimentation and validation were conducted on two multi-session SSVEP benchmark datasets. The experimental results show that our approach outperforms other state-of-art models on 2-second samples across sessions and has the potential to serve as a benchmark in multi-subject biometric recognition systems.

Hippocampus in the mammalian brain supports navigation by building a cognitive map of the environment. However, only a few studies have investigated cognitive maps in large-scale arenas. To reveal the computational mechanisms underlying the formation of cognitive maps in large-scale environments, we propose a neural network model of the entorhinal-hippocampal neural circuit that integrates both spatial and non-spatial information. Spatial information is relayed from the grid units in medial entorhinal cortex (MEC) by integrating multimodal sensory-motor signals. Non-spatial, such as object, information is imparted from the visual units in lateral entorhinal cortex (LEC) by encoding visual scenes through a deep neural network. The synaptic weights from the grid units and the visual units to the place units in the hippocampus are learned by a competitive learning rule. We simulated the model in a large box maze. The place units in the model form irregularly-spaced multiple fields across the environment. When the strength of visual inputs is dominant, the responses of place units become conjunctive and egocentric. These results point to the key role of the hippocampus in balancing spatial and non-spatial information relayed via LEC and MEC.