Accurately predicting potential drug-drug interactions (DDIs) from multimodal data is critical for medication safety and adverse drug reaction prevention. Existing methods face challenges in modeling long-range dependencies and effectively integrating heterogeneous features from structured molecular data and unstructured text. To address these limitations, we propose C2M-Mamba, a cross-modal framework that integrates convolutional neural networks, Mamba, and cross-Mamba (CroMamba) to capture discriminative features from drug descriptions, SMILES sequences, and social media texts. The model efficiently handles long-range dependencies through state space models while enabling effective cross-modal fusion. Comprehensive evaluations on the DDIExtraction2013 dataset demonstrate that C2M-Mamba outperforms 10 state-of-the-art baselines, achieving 82.37% precision, 80.98% F1-score, and 88.73% AUC. The proposed approach also exhibits robust performance in handling class imbalance and provides interpretable predictions, offering a reliable solution for multimodal DDI prediction with potential applications in pharmacovigilance and personalized medicine.

Predicting potential drug-drug interactions (DDIs) from biomedical data plays a critical role in drug therapy, drug development, drug regulation, and public health. However, it remains challenging due to the large number of possible drug combinations, and multimodal biomedical data, which is disorder, imbalanced, more prone to linguistic errors, and difficult to label. A Semantic Cross-Attention Transformer (SCAT) model is constructed to address the above challenge. In the model, BioBERT, Doc2Vec and graph convolutional network are utilized to embed the multimodal biomedical data into vector representation, BiGRU is adopted to capture contextual dependencies in both forward and backward directions, Cross-Attention is employed to integrate the extracted features and explicitly model dependencies between them, and a feature-joint classifier is adopted to implement DDI predication (DDIP). The experiment results on the DDIExtraction-2013 dataset demonstrate that SCAT outperforms the state-of-the-art DDIP approaches. SCAT expands the application of multimodal deep learning in the field of multimodal DDIP, and can be applied to drug regulation systems to predict novel DDIs and DDI-related events.

Abstract Aircraft segmentation in remote sensing images (RSIs) is an important but challenging problem for both civil and military applications. U-Net and its variants are widely used in RSI detection, but they are not suitable for multi-scale aircraft segmentation in RSIs, due to the aircrafts in RSIs are relatively small with various orientations, different sizes, fuzzy illumination and shadow, obscure boundary and irregular background. To overcome this problem, a multi-scale residual U-Net with attention (MSRAU-Net) model is constructed for multi-scale aircraft segmentation in RSIs. A multi-scale convolutional module, two modified Respaths and two kinds of attention modules are designed and introduced into MSRAU-Net to extract the multi-scale feature and make the feature fusion between the contraction path and the expansion path more efficient. Different from U-Net, MSRAU-Net replaces the convolutional block of U-Net with the Inception residual block to help the U-Net architecture coordinate the features learned from aircrafts with different scales, and the residual module and attention module are introduced into the modified Respath to deepen the network layers and solve the gradient disappearing problem while extracting the more effective feature from RSIs. The experiments on the RSI dataset validate that MSRAU-Net outperforms the other networks, in particular for detecting the small aircrafts. Compared with attention U-Net and MultiMixUNet, the precision of MSRAU-Net is improved by 9.25 and 3.36, respectively.

For semantic segmentation of remote sensing images, convolutional neural networks (CNNs) have proven to be powerful tools. However, the existing CNN-based methods have the problems of feature information loss, serious interference by clutter information, and ignoring the correlation between different scale features. To solve these problems, this article proposes a novel hidden feature-guided semantic segmentation network (HFGNet) for remote sensing images, which achieves accurate semantic segmentation by hierarchically extracting and fusing valuable feature information. Specifically, the hidden feature extraction module (HFE-M) is introduced to suppress the salient feature representation to mine more valuable hidden features. Meanwhile, the multifeature interactive fusion module (MIF-M) establishes the correlation between different features to achieve hierarchical feature fusion. The multiscale feature calibration module (MSFC) is constructed to enhance the diversity and refinement representation of hierarchical fusion features. Besides, the local-channel attention mechanism (LCA-M) is designed to improve the feature perception capability of the object region and suppress background information interference. We conducted extensive experiments on the widely used ISPRS 2-D Semantic Labeling dataset and the 15-Class Gaofen Image dataset. Experimental results demonstrate that the proposed HFGNet has advantages over several state-of-the-art methods. The source code and models are available athttps://github.com/darkseid-arch/RS-HFGNet.

This paper aims to automatically diagnose thoracic diseases on chest X-ray (CXR) images using convolutional neural networks (CNN). Most existing approaches typically employ a global learning strategy and use CNN with small convolutional kernels for thoracic disease classification. However, irrelevant noisy regions may affect the global learning strategy; small convolutional kernels can only capture fewer discriminant features. To address the above problems, we construct a multi-feature fusion neural network (MFCNet), which can fully use the global and weighted local features. Specifically, the global features are first generated by the global branch. Weighted local features are generated by multiplying the global feature and the heart-lung region mask identified by the Lung-heart Region Generator (LHRG). At last, the fusion branch integrates the global and weighted local features to complement the lost discriminative feature of the global branch and the local branch, thus enabling a better feature presentation for thoracic disease classification. Extensive experiments on the NIH ChestX-ray 14 dataset demonstrate that the MFCNet model achieves superior performance (average AUC=0.844) compared to state-of-the-art methods. Source code is released in https://github.com/Warrior996/MFCNet.

Accurate identification of weed species is the premise for controlling weeds in field. But it is a challenging task due to the complexity and high-dimensional nonlinearity of the weed images in natural field. Convolutional neural networks (CNNs) model has been widely applied to image identification, but most of the CNNs models have the problems of large parameters, low identification accuracy, and single feature scale. This paper presents a novel deep neural network structure, named as MPF-Net for weed species identification. In MPF-Net, firstly, the weed images is sent into two different scales of depthwise separable convolution layers; secondly, the parallel output feature information is cross-fused, and uses the residual learning structure to increase the network model depth and feature extraction ability; finally the lightweight model PL-Model and the scale reduction module SR-Model are stacked together to construct the lightweight network. We have performed extensive experiments on real weed datasets, and compared the proposed MPF-Net against several variations of lightweight networks. The experimental results on the weed image dataset show that the proposed method is effective and feasible for weed species identification.

Despite recent works that have achieved remarkable progress on salient object detection for natural scene images, to detect various types and scales of objects, complex backgrounds in remote sensing images are still challenging. In this study, a novel global perception network (GPNet) is constructed for the salient object detection of remote sensing images. The proposed GPNet includes a global perception module (GPM), an axial attention block (AAB), and a feature distillation structure (FDS). The GPM is used to preserve the relationships of the entire dataset, the AAB is designed to capture the dependencies between the space and channel, the FDS is introduced to enable the helpful multilevel information flow into deep layers to enhance feature generation, and the global and the local attention information are mutually fused to enhance the network mode. Extensive experiments on three public datasets demonstrate that the proposed method outperforms other compared state-of-the-art methods both qualitatively and quantitatively (https://github.com/liuyu1002/GPnet).

Building extraction is an essential task due to its relevance to urban planning and automatic surveying mapping activities. Despite the existing convolutional neural network-based methods that have achieved remarkable progress on building extraction from remote sensing images, the accurate extraction of buildings with extremely large variations of scales and layouts is still challenging. In this study, a novel multiregion scale-aware network is proposed to address these issues. The network consists of two key components. First, a multiregion attention module is proposed to capture long-range context dependencies and exploit different regions’ attention information, alleviating the interference of cluttered backgrounds and variations in building layouts. With the multiscale features generated by a backbone network as input, a graph-based scale-aware structure is designed to model and reason the interactions between different scale features to enable a better understanding of multiscale features. Extensive experiments conducted on three datasets demonstrate that the proposed method achieves superior performance compared with the other state-of-the-art methods.

Side-scan sonar (SSS) is a vital sensor for marine survey, which is widely used in military and civilian fields. The accurate segmentation of SSS images is critical in sonar image intelligent interpretation. Existing SSS image segmentation methods have several limitations, such as insufficient feature extraction, relatively worse segmentation results for tiny target categories, and serious interference by seabed reverberation noise and bright shadow region. To overcome these issues, we propose a novel encoder-decoder architecture SSS image segmentation method based on convolution neural network (CNN). First, we extract the multi-scale feature information contained in target region using the dynamic multi-scale dilated convolution (DMDC_Conv). Second, to further obtain the global and detail feature information, we construct the adaptive receptive field mechanism block (ARFM_Block). Third, we design a feature fusion attention mechanism block (FFAM_Block) to fuse high-level and low-level feature information with different scales and suppress background information interference. Final, we construct a tree structure optimization module (TSOM) to solve the problem of pixel misclassification and obtain refine SSS image segmentation results. Extensive experiments are carried out on the constructed real scene SSS image dataset. The experimental results show that the proposed method achieves 93.24% and 90.82% of MPA and MIoU, respectively, which outperforms other state-of-the-art methods and has a substantial advantage in inference speed and calculation parameters.

In industry 4.0, the most popular way to identify and track objects is to add tags. Because the cost of smart tag is still high, most companies still use cheap barcodes or QR codes. In order to solve the real-time positioning problem of upgrading traditional tags into smart tags, this paper proposes a QR tag location method based on Binarized Normed Gradients (BING) and AdaBoost-SVM. BING algorithm is the fastest general object detection algorithm at present, but its disadvantage is that the recall rate decreases sharply with the increase of Intersection-over-Union (IoU) threshold. In the proposed algorithm, Adaboost-SVM method is introduced to make up for the shortcomings of BING algorithm. More specifically, before training and prediction process of Adaboost-SVM, Contrast Limited Adaptive Histogram Equalization (CLAHE) mechanism is used for image enhancement, and thus can greatly shorten the training time and improve the precision of prediction. As a result, the precision of low-quality image prediction is significantly improved. Compared with the existing methods based on Neural Network (NN), the proposed algorithm does not depend on the parallel acceleration hardware such as GPU, hence, it can significantly reduce the hardware cost, and expand QR code automatic positioning algorithm utilization in the field of lower hardware requirements.

With the rapid and bursting development of communication engineering and some related techniques, spread spectrum communication and sparse analysis have been a hot research topic in the research community. A novel anti-jamming driven sparse analysis-based spread spectrum communication methodology is proposed in this paper, which mainly increases the spread spectrum modulation and the spread spectrum demodulation in the receiving end. The process of spread spectrum communication according to the working methods of different methodologies including direct-sequence spread spectrum. In this paper, the sparse presentation, dictionary learning, anti-jamming analysis and the basic communication theories are integrated altogether to enhance the traditional spread spectrum communication analysis framework. The experimental result proves the robustness of the proposed method.

Protein-protein interactions (PPIs) perform a very important function in a number of cellular processes, including signal transduction, post-translational modifications, apoptosis, and cell growth. Deregulation of PPIs will lead to many diseases, including pernicious anemia or cancers. Although a large number of high-throughput techniques are designed to generate PPIs data, they are generally expensive, inefficient, and labor-intensive. Hence, there is an urgent need for developing a computational method to accurately and rapidly detect PPIs. In this article, we proposed a highly efficient method to detect PPIs by integrating a new protein sequence sub-stitution matrix feature representation and ensemble weighted sparse representation model classifier. The proposed method is demonstrated on Saccharomyces cerevisiae dataset and achieved 99.26 percent prediction accuracy with 98.53 percent sensitivity at precision of 100 percent, which is shown to have much higher predictive accuracy than the state-of-the-art methods. Extensive contrast experiments are performed with the benchmark data set from Human and Helicobacter pylori that our proposed method can achieve outstanding better success rates than other existing approaches in this problem. Experiment results illustrate that our proposed method presents an economical approach for computational building of PPI networks, which can be a helpful supplementary method for future proteomics researches.

A novel supervised dimensionality reduction method called orthogonal maximum margin discriminant projection (OMMDP) is proposed to cope with the high dimensionality, complex, various, irregular-shape plant leaf image data. OMMDP aims at learning a linear transformation. After projecting the original data into a low dimensional subspace by OMMDP, the data points of the same class get as near as possible while the data points of the different classes become as far as possible, thus the classification ability is enhanced. The main differences from linear discriminant analysis (LDA), discriminant locality preserving projections (DLPP) and other supervised manifold learning-based methods are as follows: (1) In OMMDP, Warshall algorithm is first applied to constructing both of the must-link and class-class scatter matrices, whose process is easily and quickly implemented without judging whether any pairwise points belong to the same class. (2) The neighborhood density is defined to construct the objective function of OMMDP, which makes OMMDP be robust to noise and outliers. Experimental results on two public plant leaf databases clearly demonstrate the effectiveness of the proposed method for classifying leaf images.

As for gait recognition, we propose a new discriminant dimensionality reduction method, named Bimodal Discriminant Projection Analysis (BDPA) algorithm. In BDPA, a weight path-based similarity measure is designed, the intra-class scatter matrix is constructed by the weight, while the inter-class scatter matrix is constructed by the heat kernel function. Compared with the classical methods, such as Multimodal Preserving Embedding (MPE) and Minimax Risk Criterion methods, the proposed method can preserve within-class neighborhood geometry and extract between-class relevant structures for recognition by minimizing the intra-class scatter and maximizing the inter-class scatter. The experimental results on real-world gait data show that BDPA is effective and feasible for gait recognition.

In this paper, an efficient feature extraction algorithm called orthogonal linear local spline discriminant embedding (O-LLSDE) is proposed for face recognition. Derived from local spline embedding (LSE), O-LLSDE not only inherits the advantages of LSE which uses local tangent space as a representation of the local geometry so as to preserve the local structure, but also makes full use of class information and orthogonal subspace to improve discriminant power. Extensive experiments on standard face databases demonstrate the effectiveness of the proposed method.

Classical linear dimensional reduction algorithms, such as Linear Discriminant Analysis (LDA) and Locality Preserving Projections (LPP) have been widely used in computer vision and pattern recognition. However, when dealing with the multidimensional dataset, they usually first transform the original data to vectors, and then analyze the data in such a high dimensional space. This process inevitably results in some obvious disadvantages. This paper proposes a novel two-dimensional dimensionality reduction algorithm called 2D Neighborhood Discriminant Projection (2D-NDP), which is based directly on 2D image matrices rather than 1D vectors. 2D-NDP detects the intrinsic class-relationships between the images by incorporating both class label information and neighborhood information. It can optimally preserve not only the local class information but discriminant information as well. Under the orthogonal constrain, 2D-NDP is developed as orthogonal 2D-NDP for classification. Experiments on the face database and the plant leaf database demonstrate that orthogonal 2D-NDP is effective and feasible for classification.