Developing multimodal data-driven diagnostic systems has become a key clinical strategy for improving breast cancer outcomes. However, effectively modeling multimodal features remains challenging due to substantial semantic heterogeneity, scale discrepancies, and the inherent difficulty of cross-modal alignment. Although existing studies have proposed various multimodal fusion methods, most rely on direct feature concatenation or shallow integration, which fail to capture fine-grained intra-modality semantics as well as the complex interactions between histopathological and genomic modalities. In this study, we propose a multimodal diagnostic framework based on Feature Enhancement and Semantic Collaborative Alignment (FESCA). The method incorporates a semantic-guided modality feature enhancement mechanism that effectively extracts and strengthens diagnostic cues from both pathological images and genomic data. In addition, a contrastive-learning-based cross-modal alignment strategy is introduced to map heterogeneous modalities into a unified semantic space and achieve deep semantic collaboration through contrastive optimization. To ensure robust breast cancer classification under varying modality availability, a multimodal collaborative diagnostic strategy is employed to dynamically adapt the feature representations. We evaluate FESCA on the TCGA-BRCA dataset, and the experimental results demonstrate that it outperforms state-of-the-art methods in breast cancer classification while significantly improving both intra-modality representation quality and cross-modal semantic alignment. To enhance accessibility and practical application, we developed a web-based breast cancer pathological staging diagnosis system to visualize and deploy the FESCA model, demonstrating a step toward clinical application and providing a benchmark for other research methods.

Leaf Area Index is a crucial biophysical variable, and its accurate estimation is essential for understanding vegetation dynamics. However, cloud cover significantly restricts optical remote sensing, hindering the generation of spatially continuous Leaf Area Index products. Remote sensing foundation models offer novel solutions to this challenge. This study presents an end-to-end framework based on the fine-tuned Prithvi foundation model for direct LAI retrieval from cloud-contaminated 30 m Harmonized Landsat and Sentinel-2 imagery. By mapping inputs directly to Hi-GLASS reference labels, the proposed architecture processes cloud contamination and vegetation signals simultaneously and circumvents the error propagation inherent in cascaded retrieval pipelines. Results demonstrate that the end-to-end LAI retrieval model significantly outperforms cascaded variants, achieving a superior R2 (0.78) and lower RMSE (0.57). Furthermore, predictive accuracy exhibits a distinct U-shaped trajectory relative to the temporal mean cloud fraction, reaching an inflection point at 50–60% occlusion, which highlights the model’s implicit regularization capacity under severe atmospheric interference. This work establishes that direct feature learning with foundation models offers a more robust and streamlined pathway for generating continuous biophysical products from imperfect optical observations, prioritizing quantitative fidelity over artificial perceptual sharpness.

With the booming development of multimodal data (e.g., image, text) on internet platforms, multimodal sequential recommendation methods continue to emerge. Most existing methods incorporate item modal features as auxiliary information, typically concatenating them to learn unified user representations. However, these methods directly use modal features for representation learning, neglecting the impact of inherent modal noise. We argue that internal-modal noise and cross-modal noise hinder the acquisition of more accurate user representations. To address this problem, we propose SGP4SR - Separated-modality Guided user Preference learning for multimodal Sequential Recommendation. Globally, the user preference modeling is carried out from a separated-modality perspective to alleviate cross-modal noise. Locally, for each individual modality, we use item relationship graphs and user interest centers, aggregated with ID embeddings, to replace direct modal features, thereby mitigating internal-modal noise. Finally, user representations from both separated-modality and multimodal perspectives participate in prediction independently. In experiments conducted on four real-world datasets, our method outperforms state-of-the-art approaches, achieving an average performance improvement of up to 8.84% over the best baseline. The comprehensive experiments further validate the superior noise tolerance and robustness of our method.

The transformation of vehicles into software-defined platforms creates a fundamental trust gap between consumers and entities seeking detailed telemetry data. Traditional incentive models, including direct payments and feature personalization, fail to motivate sustained telemetric sharing because they cannot overcome consumer concerns about surveillance, fairness, and asymmetric value extraction. This article introduces the Automotive Data-Equity Loop, a socio-technical framework that reframes telemetry sharing from data extraction to asset protection by converting operational data into durable, consumer-owned verified credentials that enhance resale value. The framework operates through three interconnected phases: value creation during ordinary vehicle use, value crystallization through privacy-preserving conversion of raw telemetry into tamper-resistant cryptographic credentials, and value realization when verified history reduces buyer uncertainty in resale transactions. The hybrid asset conceptual model positions vehicles as composite assets integrating physical components with verifiable digital history layers, leveraging behavioral-economic principles, including loss aversion and endowment effects, to align long-term consumer interests with platform data requirements. The transparency slider interface operationalizes graduated consent through multi-level controls mapping sharing choices to certification outcomes, while purpose-bound certificates restrict secondary use by design. Comparative analysis establishes testable hypotheses predicting that asset value enhancement outperforms alternative incentive structures through increased perceived fairness and reduced exploitation concerns. The governance blueprint addresses multi-stakeholder trust through explicit role definitions for credential issuers, verifiers, and marketplace participants, alongside threat models addressing consumer gaming, forgery attacks, and marketplace pressure dynamics. Implementation requires empirical validation through controlled experiments measuring adoption rates, persistence duration, and depth of consent, complemented by observational market studies assessing effects on resale prices, transaction velocity, and buyer confidence indicators. The framework demonstrates how consumer-generated data ecosystems can align with autonomy and long-term value creation when systems treat individuals as asset owners rather than extractable data sources, with implications extending beyond automotive contexts to broader Internet of Things and smart asset domains.

USEM: A Unified Model for Simultaneous Estimation of Multiple Nutrient Concentrations in Coastal Waters using Landsat 5/7/8 and Sentinel-2 imagery.

Pseudo-label-free instance screening of non-tumor regions in whole slide images for improved classification and survival prediction.

When capturing star images in long-exposure mode, due to the relative motion between stars and space objects and the observation camera, strip tailings with different directions and lengths will be formed, resulting in a serious decline in image quality and inaccurate centroid positioning. Traditional methods for restoring star images are prone to ringing effects and cannot restore the non-uniformly blurred star images. Aiming at this problem, this paper proposes a star image restoration network based on a dynamic strip attention mechanism. Firstly, a Multi-scale Dynamic Strip Pooling Module is designed to adaptively extract blurred features of different lengths and directions by dynamically adjusting the strip convolution. After that, a Multi-scale Feature Fusion Module is designed to fuse multi-level features to reduce the loss of image details of stars and space objects in the image. Experimental results demonstrate that the proposed method achieves a PSNR of 84.08 and an SSIM of 0.9928 on the 16-bit simulated dataset, outperforming both traditional methods and other deep learning-based approaches. Specifically, the recognition accuracy of star points is increased by 174% in comparison with unprocessed images. Furthermore, this paper validates the network using the real-world dataset spotGEO, and the results indicate that the average number of successfully recognized star points is increased by 57% compared to direct processing of the original images.

Advances in global satellite-based remote sensing, big data, and machine learning are making large-scale geographic analysis increasingly accessible. Archaeology can benefit from multitemporal and multisensor satellite data even when direct archaeological site and feature detection is not possible by using these data to examine relationships between archaeological, ecological, and climactic variables across vast geographies. With this, archaeologists can ask questions about diverse landscape matrices and how these shaped the flow of past cultural processes. We demonstrate the value of thermal sensor data-common in satellite arrays-by developing an automated routine for analyzing Landsat 8 Thermal Infrared Sensor time series data. We fit a harmonic regression to extract amplitude and phase shift values for thousands of inland lakes in Michigan's lower peninsula and computed perimeter-to-area ratios. Treating these thermal data not as absolute equivalents to past conditions but as relative indices, we compared lakes with and without Late Precontact (ca. 1200-1600 CE) burial mounds. Results show lakes with mounds warmed later in the spring, cooled later in the fall, and were more regularly shaped, suggesting Late Precontact communities placed mounds on lakes with specific resource benefits, including extended growing seasons. Considering our findings alongside emerging evidence for widespread precontact Indigenous agriculture in the Great Lakes region, it is important to reconsider maize's subsistence and ceremonial significance. The analytic workflow-leveraging free, global Landsat data in Google Earth Engine-offers a replicable framework for archaeologists worldwide to integrate relative thermal patterns into cultural landscape studies.

ABSTRACT Marine two-stroke low-pressure direct injection natural gas dual-fuel engines feature low compression ratios (CR) to avoid knock, which restricts thermal efficiency improvement. This study optimises CR, pilot injection timing (PIT) and exhaust gas recirculation (EGR) via multi-objective methods to boost efficiency without knock. A validated 1D simulation model was established, followed by single-factor analysis and response surface modelling based on Latin Hypercube Design simulation results. NSGA-II and TOPSIS were then used to derive the optimal scheme. Single-factor results show higher CR cuts brake-specific fuel consumption (BSFC) but causes knock; delayed PIT mitigates knock but worsens BSFC; elevated EGR reduces knock risk with BSFC first dropping then rising and NOx emissions varying consistently with knock induction time integral (KITI). Optimisation finds optimal parameters (CR=14.6, PIT=-8.08 °CA, EGR=40%) reduce BSFC by 8.62% and NOx by 85.8% vs the prototype, with KITI and peak cylinder pressure within safe thresholds. 1. Highlights Performance optimization of marine two-stroke DF engines considering knock suppression Increasing compression ratio reduces BSFC but can lead to knock Delaying pilot fuel injection timing reduces knock risk but increases BSFC Appropriate EGR rates improve both BSFC and knock performance simultaneously Cooperative optimization cuts BSFC by 8.62% and NOX emissions by 85.8%

The aim. To carry out an analysis of the state of the modern types of substrates for growing cucumbers in a protected soil, their compliance with the requirements of plants, in order to highlight the most suitable quality cucumber harvest for obtaining a high yield under various technological solutions. Results. An analysis of modern domestic and foreign scientific and patent literature indicates that the basic advantage of using substrates is based on the factors like cost, availability, functionality, easy utilization, physical properties (appropriate fixation and support of the plant), aeration (air permeability), water holding capacity, drainage, ion exchange capacity, absence of phytopathogens and weeds. It is being used in the combination with bioactive compounds that makes plants able to implement its productivity potential. The study found a correlation between the biological and physiological characteristics of cucumber plants and the choice of substrate. It was established that since the surface root system of plants requires stable moisture in the upper layer, this can be provided by coconut fibre, peat and vermiculite; Intensive fruiting of plants requires high buffering capacity of nutrients, which can be provided by coconut fibre and mineral wool. The sensitivity of the root system to hypoxia requires high air permeability, which can be provided by perlite and mineral wool. the need to regulate acidity (pH) requires high buffer capacity of the substrate, which can be provided by coconut substrate and limed peat substrate; high sensitivity of plants to salinity requires low salt accumulation capacity, which can be provided by inert substrates. Conclusion: It is established that bio-physiological features of cucumber and direction of use of received products dependent on clear selection requirements for substrates. Thus, for intensive production with predominant mineral nutrition, it is advisable to use mineral wool or coconut fiber. For organic growth or repeated use - wood fiber, peat compost mixtures, sometimes with biocharcoal. For plants in the initial growth season, it is important to provide conditions that are as comfortable as possible and that have stable humidity and pH. For such purpose, the most suitable soil mixture is coconut fiber with perlite.

Electroencephalogram (EEG)-based emotion recognition holds great potential in affective computing, mental health assessment, and human-computer interaction. However, EEG signals are non-stationary, noisy, and composed of multiple frequency bands, making direct feature learning from raw data particularly challenging. While end-to-end models alleviate the need for manual feature engineering, advancing the performance frontier of lightweight architectures remains a crucial and complex challenge for practical deployment. To address these issues, we propose LMSA-Net (Lightweight Multi-Scale Attention Network), a lightweight, interpretable, and end-to-end model that directly learns spatio-temporal features from raw EEG signals. The architecture integrates learnable channel weighting for adaptive spatial encoding, multi-scale temporal separable convolution for rhythm-specific feature extraction, and Sim Attention Module for parameter-free saliency enhancement. Our proposed LMSA-Net is evaluated on three benchmark datasets, SEED, SEED-IV, and DEAP, under subject-dependent protocols. It achieves top performance on SEED (65.53% accuracy), competitive results on SEED-IV (48.52% accuracy), and strong performance in arousal classification on DEAP, demonstrating good generalization. Ablation studies confirm the critical role of each proposed module. Frequency analysis reveals that our multi-scale temporal kernels inherently specialize in distinct EEG rhythms, validating their neurophysiological alignment. Lightweight design is evidenced by minimal parameters (7.64K) and low latency, ideal for edge deployment. Interpretability analysis further shows the model's focus on emotion-related brain regions. LMSA-Net thus delivers an efficient, interpretable, and high-performing solution. The code is available athttps://github.com/rhr0411/LMSA-Net.git.

Research into the use of Instagram for learning writing has mostly focused on how Instagram was used by teachers to improve students’ writing skills rather than how it is used by students in self-directed learning. To fill this gap, this research aimed at uncovering students’ perceptions towards Instagram for learning writing skills in their natural settings without any intervention to any variables. The data were derived from Likert Scale questionnaires distributed to 53 students and interviews with selected students. The data revealed that the students tended to have positive perceptions toward: 1) caption and photo feature (M=4.06), 2) Instagram social media interaction (M=3.81), and 3) Instagram direct message (M=3.89) to learn English writing. The students believed that they could explore ideas, learn vocabulary, and grammar through the caption and photo feature. They added that interactions through comment and direct message feature on Instagram allowed them to exchange ideas and promote their writing skills. The research also revealed that the students could receive feedback to correct the mistakes in grammar and vocabulary from other online users although it happened occasionally.

Ultrasonography, especially Transvaginal sonography (TVS) is an effective, non-invasive and reliable investigation for the diagnosis of adenomyosis. The Morphological Uterus Sonographic Assessment group consensus terminology provides a standardized lexicon for the description of myometrial lesions and has been recently revised to include direct and indirect features of adenomyosis on sonography. In this article, we aim to provide a simplified framework for the practical application of the MUSA group consensus terminology in the ultrasonographic evaluation of adenomyosis, aiding in accurate diagnosis and informed decision-making.

Content Marketing Strategy on Social Media of Mayoutfit Fashion Brand

Electronic waste (E-waste) is an escalating global challenge, with discarded laptops forming a major share of this growing environmental burden. To support sustainable consumption and informed consumer decision making, this study proposes an unsupervised deep learning framework that ranks refurbished and new laptop brands based on consumer sentiment extracted from online reviews. The framework identifies not only direct product features called aspects (such as battery, display, or customer support) but also experiential dimensions (such as reliability, performance, or overall satisfaction), providing a holistic view of consumer perception. By leveraging a transformer-based multi-headed attention mechanism and part-of-speech tagging, the model extracts rich five-part sentiment structures: Aspect/Dimension, Category, Opinion, Irrealis (hypotheticals), and Sentiment, collectively represented as ACOIS and DCOIS quintuples. These insights feed into a folksonomy-based consumer brand ranking (CBR) algorithm, which aggregates sentiment scores to rank laptop brands effectively. Unlike traditional models, this framework requires no labeled training data, increasing its adaptability across domains. Comparative evaluations against state-of-the-art supervised and self-supervised models, including Large Language Models (LLMs), demonstrate superior performance with F1 score improvements of 9%, 6%, and 4% in extracting product aspects, dimensions, and opinions, respectively. The model is applied to a curated dataset comprising new and refurbished laptops within the same price segment. Results show that 40% of refurbished brands appear in the top 25% of recommendations. We ensured the framework's robustness check, including McNemar's statistical testing on 6 subtasks (5/6 above 0.05 threshold), ablation studies with 2 alternative attention mechanisms, and validation against 14 benchmark methods, confirming framework's stability.

We present FISN, a generalizable novel view synthesis algorithm that enables feedforward inference of Neural Radiance Fields (NeRF) or 3D Gaussian Splatting (3DGS) from reference images. Unlike existing work that either separately model the 3D feature space on each view or process multiview reference features by 3D-point-based view aggregation, FISN integrates multi-reference 3D cost volumes into a unified high-dimensional entity. Specifically, we reconceptualize the generalizable novel view synthesis task as a feedforward process of FInding Spatial Neighborhoods across this unified 4D feature space, comprising both view and spatial dimensions, and introduce View-Spatial Convolutions for direct 4D feature aggregation. This enhances the correlation among multiview neighboring points in a window-to-window manner and incorporates 3D spatial awareness. However, this approach poses two intertwined challenges: high computational expense for high-dimensional features and degraded rendering performance with low-resolution features. To address these challenges, FISN constructs a new efficient convolution paradigm, Decomposable View-Spatial Convolution, which includes a Spatial Cross Decomposition strategy as well as a Feature Compression and Upscaling module. This paradigm maintains multiview geometric consistency better than existing decomposition methods and achieves a balance between efficiency and fine-grained spatial features. Furthermore, by integrating Depth Refinement modules based on this paradigm, FISN further improves global depth understanding. Comprehensive evaluations on mainstream datasets and benchmarks demonstrate that FISN achieves state-of-the-art performance for both NeRF and 3DGS, and remains robust in challenging scenarios where existing 3DGS-based methods struggle, such as those with noisy poses or dense references. The code will be released soon.

Droplet microfluidics, which generates and manipulates water-in-oil microdroplets within continuous phases, has emerged as a compelling platform in modern science. The core advantage of this technology lies in the fact that each picoliter to nanoliter droplet functions as an independent microreactor, ensuring no cross-contamination. This enables ultra-high-throughput experiments while dramatically reducing the consumption of expensive reagents and rare samples. However, the efficient extraction of solid precipitates (such as crystals and particles) formed within droplets remains a fundamental challenge for subsequent analysis and utilization. This study proposes a novel microfluidic device and operational method to address these challenges: (1) the difficulty in extracting solids that cannot be recovered through simple fluid flow and (2) sample loss during long-distance transport. The key innovation combines (1) a passive trap structure for in situ solid formation processes within droplets and (2) a physically accessible harvesting chamber positioned nearby. This design eliminates the need for long-distance sample transport, enabling the gentle transfer of droplets containing precipitated solids to an adjacent extraction chamber with an open top, allowing for physical solid recovery. We demonstrated the system functionality using fluorescent microbeads as model particles, followed by the successful generation and recovery of protein (lysozyme) crystals as a practical application.

While no-reference point cloud quality assessment (NR-PCQA) approaches have achieved significant progress over the past decade, their performance often degrades substantially when a distribution gap exists between the training (source domain) and testing (target domain) data. However, to date, limited attention has been paid to transferring NR-PCQA models across domains. To address this challenge, we propose the first unsupervised progressive domain adaptation (UPDA) framework for NR-PCQA, which introduces a two-stage coarse-to-fine alignment paradigm to address domain shifts. At the coarse-grained stage, a discrepancy-aware coarse-grained alignment method is designed to capture relative quality relationships between cross-domain samples through a novel quality-discrepancy-aware hybrid loss, circumventing the challenges of direct absolute feature alignment. At the fine-grained stage, a perception fusion fine-grained alignment approach with symmetric feature fusion is developed to identify domain-invariant features, while a conditional discriminator selectively enhances the transfer of quality-relevant features. Extensive experiments demonstrate that the proposed UPDA effectively enhances the performance of NR-PCQA methods in cross-domain scenarios, validating its practical applicability. The code is available at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/yokeno1/UPDA-main</uri>

The extensive deployment of end-to-end encryption renders traditional network traffic analysis methods ineffective, necessitating the development of new advanced deep models with high classification performance. Combination models of CNNs and Transformers have shown efficacy by acquiring both local and global feature extraction capabilities. Existing hybrid architectures often adopt heuristic fusion operations (e.g., direct feature concatenation or hand-crafted cross-branch feature exchange), without explicitly encouraging complementarity between heterogeneous representations, which may result in redundant features and suboptimal performance. This paper introduces DAIR-MTC, a new model for Decoupled but Aligned Information Representation in Multi-Task Classification. DAIR-MTC innovates by splitting the feature space of each branch (1D-CNN and Transformer) into two disjoint subspaces: a shared space and a private space. We introduce a dual information-theoretic goal to structure this representation. First, a contrastive learning loss is employed to enforce alignment of shared features, compelling the two branches to agree on the intrinsic, modality-invariant characteristics of the traffic. Second, a mutual information minimization loss is employed to enforce decoupling of private features, compelling each branch to extract distinctive, complementary information. This ‘‘consensus and specificity’’ framework creates a highly efficient and robust feature representation. Final classification is done on an aggregation of the aligned common features and the diversified private features. Experiments on large scales on benchmark ISCX VPN-nonVPN and CICIDS2017 datasets indicate that DAIR-MTC performs better than the baseline MTC model and numerous other state-of-the-art approaches by a significant margin on a range of classification tasks.

The persistent operation of rotating detonation engines (RDEs) is severely constrained by intense combustion chamber vibrations and excessive noise emissions. Existing research on engine noise has been predominantly confined to experimental investigations of noise generation mechanisms and sound pressure levels, leaving critical gaps in predictive modeling. To address this, this study employs a numerical methodology for analyzing transient flow fields within the combustion chamber, enabling systematic characterization of vibration-coupled noise phenomena. Leveraging this computational approach, we elucidate the spectral and directivity features of vibration-induced noise. Subsequently, parametric studies identify key factors governing vibrational and acoustic responses. The results demonstrate strong agreement between simulated noise profiles and prior experimental data, confirming the model's predictive capability. Notably, the acoustic radiation exhibits pronounced directionality, with dominant emissions concentrated in specific azimuthal orientations. Furthermore, the parametric analysis yields actionable insights for structural optimization, including recommendations for chamber design parameters. These findings provide a theoretical foundation for mitigating vibration and its resultant acoustic emissions, thereby supporting future advancements in the performance and durability of RDEs.