Retrieving visual content from electroencephalography (EEG) signals represents a challenging frontier in implicit multimedia analysis, yet it suffers from extreme modal heterogeneity. The high-dimensional, non-stationary noise in neural signals limits conventional point-to-point similarity measures in capturing complex semantic manifolds. To bridge this gap, we propose the Neuro-Image Geometric Contrastive Learning (NIGCL) framework. Departing from reliance solely on simple first-order similarity metrics, NIGCL employs a geometry-aware alignment mechanism rooted in manifold learning. Specifically, we incorporate a Geometric Area Contrastive Loss based on the Gram matrix determinant, which constrains the geometric area of cross-modal feature pairs to enforce intra-class compactness and mitigate the impact of orthogonal perturbations. This global constraint is complemented by a local dot-product objective for fine-grained consistency. Additionally, we propose Geometric Area Ranking (GaR) to replace standard ranking protocols, identifying semantically consistent images via the geometric area in high-dimensional spaces. Experiments on the THINGS-EEG dataset show that NIGCL achieves superior performance, attaining 31.2% Top-1 and 61.6% Top-5 accuracy in 200-way retrieval. Reconstruction evaluations further confirm that our geometrically-aligned representations significantly improve semantic fidelity over traditional methods. This framework offers a novel perspective on aligning highly heterogeneous multimedia data through explicit geometric constraints.

Emotion recognition from multimodal data remains challenging due to the semantic gap and temporal-spatial misalignment between EEG signals and facial expressions. To address this, we propose a cross-modal framework that integrates EEG and facial features via modality-guided semantic representation learning. Temporal features are extracted by stacked MAMBA-based blocks capturing long-range dependencies. A Cross-Modal Scaling and Shifting (CMSS) mechanism uses EEG features to refine and align facial representations, reducing modality discrepancies. The fused features pass through a GRUcontrolled Mixture-of-Experts (MoE-GRU) module, where a learnable gating network dynamically selects specialized Transformer experts. This combination effectively handles misalignment and enables dynamic feature fusion, enhancing recognition accuracy and robustness. Experiments on DEAP and MAHNOBHCI demonstrate state-of-the-art results, validating its real-world applicability in affective computing.

To tackle the challenge of effectively representing time-varying information in cross-subject EEG emotion recognition, we introduce Swin Transformer-Based Temporal-Channel Network (Swin-TCNet), a novel multi-scale neural network with parallel temporal pathways to enhance the extraction of dynamic temporal features. EEG signals are simultaneously processed through a Temporal Swin Transformer for 3D feature representation and a dynamic spatiotemporal convolutional layer with multi-head attention for extracting differential entropy-based channel features, which enhances channel learning while preserving temporal information. Swin-TCNet attains state-of-the-art performance, achieving 93.47% and 86.80% accuracy in cross-subject experiments on SEED and SEED-IV datasets, respectively, as substantiated by ablation studies. By leveraging temporal dynamics, this framework enhances the extraction of temporal variations and spatial information, leading to more robust and generalizable cross-subject emotion recognition.

To decode Motor Imagery EEG signals (MI-EEG), most studies have increasingly complicated network models and parameters without fully considering EEG characteristics, thereby limiting advancements in Brain-Computer Interface (BCI) systems and classification performance. To address these issues, we propose a lightweight end-to-end tri-domain feature fusion network named LTDFNet. Firstly, we introduce an Attention-based Spatio-temporal Convolution module (ABST) to extract low-dimensional spatio-temporal features from EEG. This module employs a lightweight Squeeze-and-Excitation (SE) attention mechanism to enhance the model's perception of crucial information. Secondly, Temporal Domain Convolutional (TDC) and Frequency Domain Convolution (SDC) modules utilize Temporal Convolutional Networks (TCN) and Fast Fourier Transform (FFT) to respectively learn high-dimensional temporal and frequency domain information. Finally, the Feature Fusion (FF) module integrates low-dimensional spatio-temporal features and high-dimensional temporal-frequency features effectively through learnable parameters. LTDFNet is trained with joint constraints of Softmax loss and Center loss functions to achieve optimal inter-class separation and intra-class compactness, thereby enhancing overall model performance. This study conducts extensive experimental validation on BCI Competition IV datasets 2a (BCI 2a) and 2b (BCI 2b). LTDFNet achieves classification accuracies of 76.89% (kappa score: 0.692) and 85.22% (kappa score: 0.704) on the BCI 2a and BCI 2b datasets, respectively. Compared to other high-performance decoding methods, LTDFNet utilizes only 20,576 parameters, balancing network scale and decoding performance requirements.

Major depressive disorder (MDD) which is a widespread disorder worldwide cause disruption in some brain functions and thus leads to brain network changes. An increasing number of clinical and cognitive neuroscience studies have used broadband EEG microstate method to assess the electrical activity of large-scale cortical networks; however, the topographic frequency patterns of the different EEG microstate categories in MDD patients are not clear. In this study, EEG microstate frequency spectra were analyzed using 5-min resting-state electroencephalography (EEG) data from 55 subjects (including 27 MDD patients and 28 healthy controls) with variational empirical mode decomposition in Hilbert-Huang transform. The results showed that microstate D and the other microstates (A, C, and E) display opposite patterns. Most specifically, in the beta band, the marginal spectral energies of microstates A, C, and E were increased in MDD patients compared to HCs, while the energy of microstate D was decreased. Meanwhile, we observed that the marginal spectral energies of microstates A and C in the beta band were positively correlated with the severity of depressive symptoms, suggesting that alterations in the beta band energy can serve as an important reference for disease progression. These results confirm the abnormalities of beta-band energy in MDD patient and provide a new perspective for exploring the abnormalities of psychomotor function in patients with MDD.

Background: Despite potential neural mechanism of depression being the object of a thriving field, the research of the alterations in brain functional connections is not entirely clear. In this context, we used source-level effective connectivity and microstate analysis to study resting-state brain activity in depression.Methods: Resting-state electroencephalogram (EEG) data from 17 depressive subjects and 19 controls were included. We applied multivariate autoregressive models combined independent component analysis (MVARICA) and generalized partial directed coherence (GPDC) to analyze the brain functional system (BFS) alterations induced by depression. Further, microstate characterized the spatial organization and temporal dynamics of large-scale cortical activities was used for depression disease to understand brain network dynamics. Results: Compared with controls, depression had enhanced information flow from frontal to parietal in alpha band. Especially, the frontal and parietal lobes respectively was correspond to dominant hub in abnormally weaker and stronger causal pathways in patients with depression. In addition, microstate analysis revealed that mean duration, occurrence rate, time coverage of microstate class D were significantly lower in patients compared to controls. Meanwhile, patients preferred bilateral transitions between C and D compared with that in controls. Microstate D may be associated with the frontoparietal dorsal attention network, the findings reflects switching and reorientation of attention to relevant information occur more frequently for depressed patients.Conclusions: Patients exhibited clear effective network alterations compared to controls. Notably, EEG microstate analysis might provide useful biomarkers to understand the deviant functions of large-scale cortical activities in clinical researches of patients with depression.

Emotional conflict control is impaired in major depression disorders (MDDs) and affects decision-making with further consequent social interactions dysfunction. However, neural correlates of conflict monitoring processes being modulated by different affective distractor stimuli are not clear in MDDs. In this article, we investigated abnormal neural basis of conflict monitoring processes in MDD patients by applying dynamic causal modeling (DCM) technique on electroencephalography (EEG). The results indicated that MDD patients showed lower N2 amplitudes regardless of stimulus conditions, and reduced activation within ACC region for incongruent stimuli, relative to healthy controls. Especially, MDDs had more negative N2 amplitudes to happy incongruent trials than happy congruent trials. Source localization analyses revealed that MDD patients had significantly enhanced left inferior temporal gyrus (ITG) activation, which is involved in written words processing. Further DCM analysis provided abnormal neural correlates through greater backward connections (fusiform→ITG, amygdala→ITG) on happy incongruent trials than happy congruent trials in MDD group. These findings indicate that only sad words induce significantly greater interference effects to positive target faces in MDD patients, which may be associated with ITG activity dysfunction. The findings may share new insights into the neural mechanisms of emotional conflict processing in MDDs.

Major depressive disorder (MDD) may be driven by dysfunction in intrinsic dynamic properties of the brain, and EEG microstate is a promising method for analyzing brain dynamics. However, the alterations in EEG microstate is still not entirely clear, and its ability for MDDs detection is worth probing. Moreover, the mechanism behind the neural networks contributing to microstates remains poorly understood in MDDs. Therefore, we applied microstate analysis and Topographic Electrophysiological State Source-imaging (TESS) on EEG data of 27 MDDs and 28 healthy controls (HCs). Compared to HCs, MDDs had apparent increase in microstate C and decrease in microstate D. Furthermore, TESS results showed that the underlying network of microstate C in MDDs overlapped with the anterior cingulate cortex and left insula gyrus, whereas main source of microstate D was in the orbital part of inferior frontal gyrus. The reduced transition probability from C to D in MDDs may reveal an imbalance between the networks of microstates. The microstate parameters as features reached good performance in identifying MDD (89.09% accuracy, 92.86% sensitivity, 85.19% specificity), indicating their potential as biomarkers of depression pathology. Collectively, these results highlight alteration of brain activity patterns and provide new insights into abnormal EEG dynamics in MDDs.

Drug addicts are characterized by difficulty neglecting monetary reward, but its underlying neural mechanisms remain unclear. The current study aimed to investigate the behavioral and electrophysiological signatures of abnormal attentional bias based on different amounts of reward in abstinent heroin addicts (AHAs). We used a modified attentional capture task while recording EEG in 18 AHAs and 18 age-, gander-, and education-matched healthy controls (HCs). We analyzed the attentional distribution of the relative positional changes in space of the target and reward-related stimulus. When targets integrated reward-related colors, participants were more responsive and deployed more attention to targets, especially those with high-value colors. When targets and reward-related distractors were spatially separated, high-value distractors captured the AHA's attention and slowed their responses. Moreover, AHAs had weaker attentional control than HCs, exhibiting an inability to suppress the attentional bias driven by high-value stimuli. Overall, these results demonstrated that AHAs was hypersensitive to task-irrelevant and previous reward-related stimuli, possibly due to damage to brain reward circuits caused by chronic heroin abuse. Our work provides novel behavioral and neurophysiological evidence that are closely associated with the maintenance and relapse of addiction.

In the attentional bias and cognitive processing relating to the abstinent heroin addicts (AHAs), there were considerable studies about event related potentials (ERP) and behavioral data. However, the large amount of data lead to longer data processing time, and few studies were done on single channel data about AHA. This study investigated whether single channel's data can be used to identify AHAs from healthy controls (HCs) accurately. Two groups of age-, education-, and gender-matched adults (22 AHAs, 21 HCs) performed on the gambling task. ERP features and behavior features were used to classify. For discriminating AHAs and HCs, ReliefF and SVM-RFE were applied for feature selection, and Particle Swarm Optimization (PSO) and Genetic Algorithm (GA) were used to search the optimal classification model of Support Vector Machine (SVM). We analyzed the statistical significance of all the features and obtained the classification result of different stimulation conditions. In statistics, we found that AHAs were significantly different from HCs in the amplitude of P300, ERP's mean value and ERP's variance under the monetary stimulation. For large money stimulation, P300 power in delta band and N100 power in delta band had significant difference between AHAs and HCs. Combining feature sorting algorithms and optimization algorithms, the results indicated that optimal performance was achieved by using ReliefF and GA. Use the above method, the best accuracy is 86.04% in four kind (+99, +9, -9, -99) of stimulation. This is the first study that used single channel's ERP data to identify AHAs with HCs, our study provided a new insight and objective method for the rapid diagnosis of AHAs.