Detecting dense breast cancer in dense breast tissue remains a significant challenge in mammographic screening, where conventional methods demonstrate limited sensitivity due to subtle imaging features and the masking effect of dense tissue. We propose EAGANet, a novel deep learning architecture integrating an innovative Learning Weight Matrix Mamba (LWM-Mamba) module for enhanced feature extraction and dynamic weighting in mammographic image analysis. The LWM-Mamba module implements selective state-space modeling with learnable weight matrices, enabling adaptive feature selection and transformation across different imaging domains while maintaining computational efficiency through shared state transition matrices. The architecture incorporates EdgeNet for structural feature enhancement and moment channel attention for refined channel-wise feature weighting. To address model vulnerability to perturbations, we develop a multi-level defense strategy including Robustness of Edge Features (REF), Cumulative Effects of Multi-Step Perturbations (CEMSP), and Collaborative Nature of the Fusion Module (CNFM), significantly enhancing model robustness against adversarial attacks. Furthermore, our approach provides hierarchical interpretability through visualization techniques such as Grad-CAM and SHAP value analysis, allowing clinically relevant visualization of model decision processes. The dual-stream architecture, designed to emulate the clinical practice where radiologists assess mammographic views in conjunction with breast density information, processes both MLO and CC mammographic projections alongside breast density maps, facilitating robust feature extraction and seamless fusion from multiple imaging modalities. Experiments on multiple datasets demonstrate the proposed method's effectiveness, particularly in challenging dense breast tissue cases where conventional models often struggle with feature disambiguation.
A new quassinoid, named vilmorinine G (1), along with two known congeners (2 and 3), has been isolated from the root bark of Ailanthus altissima (Mill.) Swingle. The structure of 1 was elucidated through spectroscopic evidence in NMR, HR-ESI-MS, and ECD. The absolute configuration of 1 was further confirmed by X-ray crystallography.
INTRODUCTION:Pancreatic cancer (PC) is a highly aggressive malignancy with limited treatment options and poor prognosis. Dahuang Zhechong Pill, a traditional Chinese medicine, has shown promise in inhibiting inflammation. This study investigates the effects of combining Dahuang Zhechong Pill with TNS4 silencing on PC, focusing on their combined role in suppressing PC progression and exploring potential therapeutic applications. Intellectual property implications related to this combination are also explored. METHODS:Bioinformatic analysis was used to identify the key role of TNS4 in PC. CFPAC-1 PC cells were cultured and genetically modified using lentiviral transfection to stably knock down TNS4 expression. Cell proliferation was assessed using the CCK-8 assay, while cell migration and invasion capabilities were evaluated through Transwell assays. Colony formation and flow cytometry were performed to analyze clonogenic potential and cell cycle distribution, respectively. Apoptosis was assessed using tunel staining. Subcutaneous and orthotopic tumor models were established in nude mice to investigate the in vivo effects. Mice were treated with Dahuang Zhechong Pill by oral gavage. Immunohistochemistry and immunofluorescence were employed to detect the expression of key proteins involved in the NF-κB/VEGF pathway, including E-cadherin and Vimentin. ELISA was used to measure circulating IL-17 and amylase levels in mouse serum to test inflammation response. RESULTS:TNS4 was upregulated in PC and positively associated with PC progression. TNS4 silencing significantly reduced CFPAC-1 cell proliferation, migration, and invasion in vitro. Flow cytometry demonstrated an increase in G0/G1 phase arrest and apoptosis in the TNS4 silencing group. Subcutaneous models showed the anti-tumor effect of TNS4 silencing. Furthermore, Dahuang Zhechong Pill treatment, when combined with TNS4 knockdown, resulted in a marked decrease in tumor size in orthotopic models. Immunohistochemical analysis revealed reduced expression of NF-κB and VEGF in tumor tissues from the combination treatment group. ELISA results indicated lower levels of serum IL-17, and amylase and higher levels of insulin in combination-treated mice. DISCUSSION:We proposed an innovative therapeutic patent combining traditional Chinese medicine with targeted gene silencing to inhibit PC progression. By investigating the synergistic effects of TNS4 silencing and Dahuang Zhechong Pill in suppressing the NF-κB/VEGF signaling pathway, our findings highlighted a promising strategy that targets tumor proliferation and modulates the inflammatory microenvironment in PC. CONCLUSION:Dahuang Zhechong Pill, in combination with TNS4 silencing, effectively inhibits the NF-κB/VEGF pathway and inflammation response, leading to reduced PC progression. These findings suggest a potential therapeutic approach for targeting PC through the combined use of traditional Chinese medicine and gene editing.
Plant disease identification and management are vital for crop protection, productivity, and sustainable agriculture. This article presents RepAgrViT, a lightweight vision transformer architecture for agricultural disease recognition in Internet of Things edge environments. The proposed model integrates CNN with vision transformer principles through a novel dual-stream design comprising token mixer and channel mixer arranged in series. A Bilinear Attention Transformation module performs extensive local-global attention via direct feature position connections. This enables comprehensive capture of long-range dependencies between healthy and diseased leaf regions. We introduce an efficient parameter fusion methodology that optimizes depthwise separable convolutions while integrating normalized weights with biases, reducing model complexity. Furthermore, we propose Lite-AVPSO, a hyperparameter optimization algorithm for precise model configuration. It incorporates adaptive weighted delayed velocity optimization and neighborhood-based local search strategies. Visualization techniques using multichannel color heatmaps and 3-D feature spaces provide interpretable diagnostic results. Experiments across diverse plant disease datasets demonstrate RepAgrViT efficacy in capturing discriminative disease features while maintaining computational efficiency suitable for resource-constrained agricultural monitoring systems, establishing a foundation for efficient vision-based monitoring in real-time agricultural applications.
Ligand-receptor recognition-based nanodrug delivery systems incorporate stimulus-responsive functionality to proactively transport chemotherapeutics to lesional tissues and intelligently release drugs in response to stimulation, thereby augmenting drug accumulation at target sites, boosting drug bioavailability, and avoiding systemic toxicity, which reveals a promising potential in malignant tumor therapy. Herein, we designed and prepared a dual-targeted nanomedicine, FA-HA-SS-PPT, utilizing microwave-assisted quantitative conjugating of podophyllotoxin (PPT) to folic acid (FA) and hyaluronic acid (HA) carriers via short-chain PEG bearing the tumor microenvironment-sensitive disulfide bond. The FA-HA-SS-PPT nanoparticles developed by self-assembly exploited the peculiarity of specifically binding to the folate receptor (FR) and CD44 receptor, initiated targeting to translocate the drug into the tumor cells and released the PPT by self-cyclization in high intracellular GSH concentration. Furthermore, this nanomedicine demonstrated superior tumor cytotoxicity and proliferation inhibition in vitro by triggering the mitochondrial apoptotic pathway and suppressing the microtubule protein formation. Notably, quantitative disulfide bond conjugation and nanomedicine formation endowed FA-HA-SS-PPT nanoparticles with astonishing in vivo circulatory stability and tumor microenvironmental responsiveness, further reinforcing drug security and anti-tumor capability. Therefore, quantitatively conjugating drugs provide a promising safe and efficient means to treat multi-receptor overexpressing tumors with multi-targeted nanomedicines.
