Fiber-optic distributed acoustic sensor (DAS) offers great potential for railway event monitoring due to its high sensitivity and robustness in complex environments. However, accurate recognition of acoustic events remains challenging, under real-time constraints where only short-duration signal segments with limited discriminative information are available. To overcome this, an efficient recognition framework was proposed by integrating multi-encoding image fusion and a lightweight transformer-convolutional network (CNN). Specifically, 0.032-second DAS signal segments were converted into complementary image representations, which were then fused into RGB images to enhance feature diversity. These multimodal images were processed by a compact backbone that combined vision transformer (ViT) modules with convolutional components, enabling effective extraction of multi-scale local and global features. The hybrid framework effectively mitigated the lack of temporal information in short segments, achieving 98.19% accuracy with only 1.33 million parameters and 3.3 ms latency per sample, substantially faster than existing methods. Compared to baseline models, the proposed architecture achieved superior accuracy and compactness, making it highly suitable for real-time and edge deployment in DAS-based monitoring systems.

Ship detection in remote sensing images plays an important role in various maritime activities. However, existing deep learning methods face challenges such as changes in ship target size, complex backgrounds, and noise interference in remote sensing images, which can lead to low detection accuracy and incomplete target detection. To address these issues, we proposed a synthetic aperture radar (SAR) image target detection framework called SDWPNet, aimed at improving target detection performance in complex scenes. Firstly, we proposed SDWavetpool (SDW), which optimizes feature downsampling through multiscale wavelet features, effectively reducing the dimensionality of the feature map while preserving the detailed information of small targets. It can more accurately identify medium and large targets in complex backgrounds, fully utilizing multi-level features. Then, the network structure was optimized using a feature extraction module that combines the PPA mechanism, making it more focused on the details of small targets. In addition, we further improved the detection accuracy by improving the loss function (ICMPIoU). The experiments on the SAR Ship Detection Dataset (SSDD) and High Resolution SAR Image Dataset (HRSID) show that this framework performs well in both accuracy and response speed of target detection, achieving 74.5% and 67.6% in mAP.50:.95, using only parameter 2.97M.

Emotion Recognition in Conversation (ERC) is valuable for diagnosing health conditions such as autism and depression (Maryenko, 2024), and for understanding the emotions of individuals who struggle to express their feelings. Current ERC methods primarily rely on semantic, audio and video data but face significant challenges in integrating physiological signals such as Electroencephalography (EEG), which has low signal-to-noise ratios, inter-subject variability, and temporal alignment issues. This research proposes Hypergraph Multi-Modal Learning (Hyper-MML), a novel framework for identifying emotions in conversation. Hyper-MML effectively integrates EEG with audio and video information to capture complex emotional dynamics. Firstly, we introduce an Adaptive Brain Encoder with Mutual-cross Attention (ABEMA) module for processing EEG signals. This module captures emotion-relevant features across different frequency bands and adapts to subject-specific variations through hierarchical mutual-cross attention mechanisms. Secondly, we propose an Adaptive Hypergraph Fusion Module (AHFM) to actively model the higher-order relationships among multi-modal signals in ERC. Experimental results on the EAV and AFFEC datasets demonstrate that our Hyper-MML model significantly outperforms current state-of-the-art methods. The proposed Hyper-MML can serve as an effective communication tool for healthcare professionals, enabling better engagement with patients who have difficulty expressing their emotions. The official implementation codes are available at https://github.com/NZWANG/Hyper-MML.

Lateralization is an effective model for exploring changes in brain activity and is widely used to assess brain function. Seafarers, as an occupation working in marine environments, are subjected to long-term specialized occupational demands and experiences, which inevitably impact brain function. By utilizing lateralization, the influence of occupational experience on brain activity can be further explored. A novel Effective Connectivity Lateralization Analysis (ECLA) framework is proposed, which incorporates a Transformer-based Granger causality model (Transformer-GC) to analyze the effects of seafaring on brain plasticity. The Transformer-GC model constructs effective connectivity (EC) matrices, and lateralization indices are derived to investigate occupational influences on brain activity. Two control groups of non-seafarers are included to identify seafarers' unique occupational plasticity brain networks. Results show that Transformer-GC achieves an accuracy improvement of nearly 16% and 19.4% over the GRU-based and MVGC model, respectively, and a 5% gain over Pearson-based functional connectivity, confirming its superior performance. Moreover, the results of the ECLA showed significant differences in VentralAttention, Somatomotor, DorsalAttention in the seafarer, demonstrating that these brain networks are affected by the long-term work of seafarers. The findings demonstrate the effectiveness of ECLA in revealing the impact of long-term maritime work on brain plasticity, particularly in identifying the brain network of seafarers' occupational plasticity. It is shown that occupational experience can reshape the lateralization of brain functional activity, offering new insights into neural plasticity across different professions.

Decoding neural visual representations from electroencephalogram (EEG)-based brain activity is crucial for advancing brain-machine interfaces (BMI) and has transformative potential for neural sensory rehabilitation. While multimodal contrastive representation learning (MCRL) has shown promise in neural decoding, existing methods often overlook semantic consistency and completeness within modalities and lack effective semantic alignment across modalities. This limits their ability to capture the complex representations of visual neural responses. We propose Neural-MCRL, a novel framework that achieves multimodal alignment through semantic bridging and cross-attention mechanisms, while ensuring completeness within modalities and consistency across modalities. Our framework also features the Neural Encoder with Spectral-Temporal Adaptation (NESTA), a EEG encoder that adaptively captures spectral patterns and learns subject-specific transformations. Experimental results demonstrate significant improvements in visual decoding accuracy and model generalization compared to state-of-the-art methods, advancing the field of EEG-based neural visual representation decoding in BMI. Code will be available at: https://github.com/NZWANG/Neural-MCRL.

With the rise of autonomous shipping and intelligent maritime surveillance, accurately detecting ships under multiscale dense occlusions has become a critical challenge. Traditional detectors struggle with feature degradation caused by overlapping targets and scale variation. To address this, we propose DSONet, a lightweight framework tailored for robust ship detection in complex maritime environments. Specifically, we design a task-specific feature extraction structure, DualFeatureDetection (DF-Det), which enhances spatial detail preservation while reducing redundant computation. Additionally, we introduce the AdaptiveUpsample (AU) module to improve multiscale feature fusion and spatial reconstruction, especially under occlusion. Integrated with a four-branch oriented bounding box (OBB) detection head, DSONet achieves precise localization of elongated and overlapping ships. Extensive experiments on the MID, SeaShip and SSDD datasets demonstrate that DSONet outperforms existing detectors in both accuracy and efficiency, offering a practical solution for maritime occlusion detection.

Graph deep learning (GDL) has demonstrated impressive performance in predicting population-based brain disorders (BDs) through the integration of both imaging and non-imaging data. However, the effectiveness of GDL-based methods heavily depends on the quality of modeling multi-modal population graphs and tends to degrade as the graph scale increases. Moreover, these methods often limit interactions between imaging and non-imaging data to node-edge interactions within the graph, overlooking complex inter-modal correlations and resulting in suboptimal outcomes. To address these challenges, we propose MM-GTUNets, an end-to-end Graph Transformer-based multi-modal graph deep learning (MMGDL) framework designed for large-scale brain disorders prediction. To effectively utilize rich multi-modal disease-related information, we introduce Modality Reward Representation Learning (MRRL), which dynamically constructs population graphs using an Affinity Metric Reward System (AMRS). We also employ a variational autoencoder to reconstruct latent representations of non-imaging features aligned with imaging features. Based on this, we introduce Adaptive Cross-Modal Graph Learning (ACMGL), which captures critical modality-specific and modality-shared features through a unified GTUNet encoder, taking advantages of Graph UNet and Graph Transformer, along with a feature fusion module. We validated our method on two public multi-modal datasets ABIDE and ADHD-200, demonstrating its superior performance in diagnosing BDs. Our code is available at https://github.com/NZWANG/MM-GTUNetshttps://github.com/NZWANG/MM-GTUNets.

Spiking neural networks (SNNs) have attracted much attention due to their low energy consumption and fast inference on neuromorphic hardware. Currently, the most effective way to implement deep SNNs is through ANN-SNN conversion, which combines the maturity of ANNs with the fast inference of SNNs, achieving accuracy comparable to ANNs on large-scale datasets. However, to achieve this accuracy, a large number of time steps are usually required, which undermines the low energy consumption and fast inference advantages of SNNs. When the number of time steps is very limited, the converted SNNs face a severe drop in accuracy, especially in the case of a single time step, which severely restricts the practical application of SNNs. In this paper, we analyze the main reasons why ANN- SNN cannot achieve lossless conversion with very limited number of time steps, and then proposes a group IF units with shared membrane potential to achieve lossless conversion of ANN-SNN at extremely low latency. We validate the effectiveness of our method on CIFAR-10, CIFAR-100, and ImageNet. Experiments show that our proposed method outperforms the existing state- of-the-art methods at the same number of time steps. Moreover, we have achieved an nearly lossless conversion of ANN-SNN in a single time step to get a high-accuracy SNN for the first time. For example, for VGG-16, our method only loses 0.8% and 3.15% of the accuracy compared to ANNs on CIFAR-10/100 in a single time step.

Event cameras, serving as innovative vision sensors, provide novel insights and approaches for image classification tasks. These cameras generate a sparse and discrete event stream, with each individual event carrying minimal information. Consequently, it becomes crucial to convert this event stream into a suitable event representation that aids in feature extraction and object recognition. In this research, we present a computationally efficient spatio-temporal event representation that not only preserves the spatio-temporal information of events in its entirety but also simplifies the computation of temporal information through linearly weighted timestamps. Furthermore, we propose an adaptive segmentation method for event streams. This method generates time bins that exhibit high robustness to motion speed by integrating both global and local distribution information of event counts. To verify the efficacy of our proposed method, we conducted experiments on three publicly available datasets. The results demonstrate that our method surpasses other methods on both N-Caltech101 and CIFAR10-DVS, with enhancements of 1.1% and 4.1% respectively, and produces competitive results on N-CARS.

Ship detection in remote sensing imagery is crucial for various maritime applications such as surveillance and navigation. Convolutional neural networks (CNNs) and transformers have shown significant potential in object detection within the field of image processing. However, existing models applied directly to ship detection in synthetic aperture radar (SAR) imagery encounter challenges due to the varying sizes of ship targets. This often leads to issues such as low detection accuracy, missed detections, and false alarms. In this letter, we propose a new detection network, HMA-Net, to further address these issues. Initially, we introduce the Cwin module, which enhances interference resistance at a relatively low cost, enabling the model to more accurately capture target information. Subsequently, we design a multiscale ship feature extraction module, which uses a parallel multibranch structure to extract features of ships of various sizes and shapes. Finally, we introduce an adaptive fusion loss function that flexibly allocates loss calculation methods to detected targets, thereby enhancing the robustness of the model and achieving high-quality detection boxes. The proposed HMA-Net achieved improvements of 2.0% and 0.9% in mAP.50:.95 over the baseline models on the SAR Ship Detection dataset and the High-Resolution SAR Images dataset, using only 3.52 M parameters.

Detecting dense small objects (DSOs) of varying scales still remains a challenging research problem in remote sensing imagery (RSI). Due to their weak feature extraction capabilities for small objects, most existing detection approaches struggle to handle the high proportion of DSO in RSI, thereby increasing the likelihood of missed detections. In addition, the close proximity and overlap of multiscale objects further complicate detection due to occlusion between bounding boxes. In this study, we systematically propose a novel loss function, remote sensing dense small target detection (RSDS), for detecting DSO in RSI, which contains three main components. The first is Gaussian reassignment loss (GRL), which adaptively redistributes sample weights to prevent any single sample (such as positive, negative, easy, and hard samples) from dominating the overall loss. To solve the zero-loss issue in traditional intersection over union (IoU) and intersection over ground truth (IoG) metrics when object boxes do not intersect, we design the Gaussian Wasserstein distance (WD) penalty loss, which models the eligible 2-D detection boxes as Gaussian distributions and calculates the similarity between them. The final one, which we called the occlusion box interaction loss, explains the attraction between DSO and the repulsion from their surroundings. Deploying RSDS not only significantly reduces the probability of missing DSO in RSI, but also enhances the detection accuracy in other similar computer vision tasks. Experiments on the HRSID, NWPU-10, and SSDD datasets show that some general models incorporating RSDS achieve precision improvements of 6.68%, 11.52%, and 5.26%, and 8.71%, 5.17%, and 9.34% in$\text {mAP}_{0.5}$and$\text {mAP}_{0.5\text {:}0.95}$, respectively, compared with other baselines. The code will be found athttps://github.com/CCC0090/RDSD-Loss.

Judging and identifying biological activities and biomarkers inside tissues from imaging features of diseases is challenging, so correlating pathological image data with genes inside organisms is of great significance for clinical diagnosis. This paper proposes a high-dimensional kernel non-negative matrix factorization (NMF) method based on muti-modal information fusion. This algorithm can project RNA gene expression data and pathological images (WSI) into a common feature space, where the heterogeneous variables with the largest coefficient in the same projection direction form a co-module. In addition, the miRNA-mRNA and miRNA-lncRNA interaction networks in the ceRNA network are added to the algorithm as a priori information to explore the relationship between the images and the internal activities of the gene. Furthermore, the radial basis kernel function is used to calculate the feature proportion between different kinds of genes mapped in the high-dimensional feature space and projected into the common feature space to explore the gene interaction in the high-dimensional situation. The original feature matrix is regularized to improve biological correlation, and the feature factors are sparse by orthogonal constraints to reduce redundancy. Experimental results show that the proposed NMF method is better than the traditional NMF method in stability, decomposition accuracy, and robustness. Through data analysis applied to lung cancer, genes related to tissue morphology are found, such as COL7A1, CENPF and BIRC5. In addition, gene pairs with a correlation degree exceeding 0.8 are found, and potential biomarkers of significant correlation with survival are obtained such as CAPN8. It has potential application value for the clinical diagnosis of lung cancer.

In recent years, transformer-based networks, known for their ability to model long-range dependencies, have been widely used in downstream computer vision tasks, surpassing certain neural network architectures. However, transformer-based networks suffer from issues such as large parameter size, high computational complexity, and difficulties in extending spatial and channel features to the third or even higher orders, resulting in convergence challenges for small to medium-sized datasets and limited effectiveness in extracting high-order detailed features. In this paper, we propose a high-order spatial and channel controllable convolution module, named CSn, which can replace standard convolutions in any convolutional network. In the context of remote sensing small object detection, CSndemonstrates superior performance compared to Self-Attention structures embedded in neural networks. Moreover, it introduces long-range dependency relationships among pixels, similar to Self-Attention, and adopts a cascaded recursive approach to extend spatial and channel features to arbitrary higher orders without introducing significant additional computation. This extension captures crucial information from high-order spatial and channel dimensions, resulting in improved accuracy for small object detection. Additionally, we construct a novel, versatile CSn-(FPN+PAN) structure for object detection networks, referred to as CSnNet. Finally, our proposed model exhibits significant advantages in remote sensing detection when compared to state-of-the-art methods on publicly available SAR datasets (SSDD, HRSID) and optical remote sensing dataset (NWPU-10), achieving respective improvements of 3.4%, 4.5%, and 0.4% in mAP50compared to baseline models.

Convolutional Neural Networks (CNNs) have recently been introduced to Alzheimer's Disease (AD) diagnosis. Despite their encouraging prospects, most of the existing models only process AD-related brain atrophy on a single spatial scale, and have high computational complexity. Here, we propose a novel Attention-based 3D Multi-scale CNN model (AMSNet), which can better capture and integrate multiple spatial-scale features of AD, with a concise structure. For the binary classification between 384 AD patients and 389 Cognitively Normal (CN) controls using sMRI scannings, AMSNet achieves remarkable overall performance (91.3% accuracy, 88.3% sensitivity, and 94.2% specificity) with fewer parameters and lower computational load, generally surpassing seven comparative models. Furthermore, AMSNet generalizes well in other AD-related classification tasks, such as the three-way classification (AD-MCI-CN). Our results manifest the feasibility and efficiency of the proposed multi-scale spatial feature integration and attention mechanism used in AMSNet for AD classification, and provide potential biomarkers to explore the neuropathological causes of AD.

Independent component analysis aiming at detecting the functional connectivity among discrete cortical brain regions has been extensively used to explore the functional magnetic resonance imaging data. Although the independent components (ICs) were with relatively high quality, the noise embedding in ICs has a great impact on the true active/inactive region inference and the reproducibility, in postprocessing stage, e.g., the extraction of statistical parametrical maps (SPMs). In this paper, a novel brain network enhancement model (BNEM) is proposed, which mainly consists of two key techniques: 1) 3-D wavelet noise filter (3DWNF) for the meaningful ICs, which greatly suppresses noise and enforces the real activation inference of SPMs; and 2) a spatial reproducibility enhancement algorithm (SREA), aiming to improve the reproducibility of SPMs. The simulated experiment demonstrated that the postfiltering signals by 3DWNF were with higher correlation and less normalized mean square error to the ground truths than the prefiltering ones; SREA could further enhance the quality of most postfiltering ones, preserving the consistency with 3DWNF. The real data experiments also revealed that 1) 3DWNF could lead to more accurate preservation of the true positive voxels by correctly identifying the high proportionally misclassified voxels of the nonenhanced SPMs; 2) SREA could further improve the classification accuracy of the active/inactive voxels of SPMs corresponding to the 3DWNF denoised ICs; and 3) both 3DWNF and SREA contribute to the reproducibility enhancement of the reproduced SPMs by BNEM. Thus, BNEM is expected to have wide applicability in the neuroscience and clinical domain.