Dalian Neusoft University of Information

Rumor propagation significantly impacts both individual well‐being and societal stability, with various factors influencing its spread. While existing research has primarily focused on the role of interpersonal familiarity in rumor diffusion, this study introduces a different perspective: an individual’s familiarity with the subject matter itself plays a crucial role in the dissemination process. To explore this, we develop a novel rumor‐spreading model that explicitly accounts for subject‐matter familiarity. A rigorous theoretical analysis is conducted to assess the existence and stability of equilibrium points, and we derive the system’s basic reproduction number. Through numerical simulations, we systematically investigate how key parameters affect rumor dynamics. Our results confirm that subject‐matter familiarity is a significant driver of rumor spread, supporting our initial assumption. Additionally, we identify three key control variables and demonstrate that optimizing these factors enhances the effectiveness of rumor suppression. Based on these findings, we propose practical strategies for managing and mitigating rumor dissemination.

BackgroundKnee cartilage injury (KCI) poses significant challenges in the early clinical diagnosis process, primarily due to its high incidence, the complexity of healing, and the limited sensitivity of initial imaging modalities.ObjectiveThis study aims to employ magnetic resonance imaging and machine learning methods to enhance the classification accuracy of the classifier for KCI, improve the existing network structure, and demonstrate important clinical application value.MethodsThe proposed methodology is a multidimensional feature cross-level fusion classification network driven by the large separable kernel attention, which enables high-precision hierarchical diagnosis of KCI through deep learning. The network first fuses shallow high-resolution features with deep semantic features via the cross-level fusion module. Then, the large separable kernel attention module is embedded in the YOLOv8 network. This network utilizes the combined optimization of depth-separable and point-by-point convolutions to enhance features at multiple scales, thereby dramatically improving the hierarchical characterization of cartilage damage. Finally, five classifications of knee cartilage injuries are performed by classifiers.ResultsTo overcome the limitations of network models trained with single-plane images, this study presents the first hospital-based multidimensional magnetic resonance imaging real dataset for KCI, on which the classification accuracy is 99.7%, the Kappa statistic is 99.6%, the F-measure is 99.7%, the sensitivity is 99.7%, and the specificity is 99.9%. The experimental results validate the feasibility of the proposed method.ConclusionsThe experimental outcomes confirm that the proposed methodology not only achieves exceptional performance in classifying knee cartilage injuries but also offers substantial improvements over existing techniques. This underscores its potential for clinical deployment in enhancing diagnostic precision and efficiency.

The construction industry faces pressure to improve sustainability, but a critical gap persists between conceptual Digital Twin (DT) frameworks and the empirically validated, operational workflows needed for material design. To bridge this gap, this research proposes and empirically evaluates a tangible, end-to-end Industry 4.0 workflow. This study serves as a proof-of-concept, demonstrating how a DT can be developed from a theoretical concept into a practical optimization engine for material design. Our methodology operationalizes this blueprint by utilizing a public dataset as a proxy for digitalization, developing a deep neural network (DNN) as the high-fidelity virtual twin, and benchmarking it against four classical machine learning (ML) models to predict concrete compressive strength. The results provide a strong performance benchmark for this approach: the deep learning-based Digital Twin achieves high predictive fidelity with an R² of 0.9011 and a Root Mean Squared Error (RMSE) of 5.09 Megapascals (MPa) on the held-out test set. In stark contrast, baseline linear models achieved an R-squared (R²) of only 0.632, demonstrating that the complexity of our proposed workflow is essential for capturing non-linear material interactions. This work’s primary contribution is a reproducible blueprint for applying predictive analytics in sustainable construction, enabling the data-driven design of high-performance, low-carbon concrete mixes while significantly reducing the need for physical experimentation.

Acute ischemic stroke (AIS) severely compromises patients' quality of life, making accurate National Institutes of Health Stroke Scale (NIHSS) scoring critical for guiding subsequent treatment. This paper investigates the use of textual data—including patient complaints, current medical history, and past medical history—to achieve precise stroke severity grading. To enhance the accuracy of medical text analysis and support clinical decision-making, we focus on the classification of longtext data and propose an enhanced BERT model, termed the eBERT-AP-DAL model, which significantly improves classification performance. Our contributions are twofold: (1) We integrate an attention pooling mechanism into BERT to replace the original pooling strategy, thereby enhancing feature extraction. (2) We incorporate a category attention weighting scheme into the loss function to mitigate the impact of class imbalance on predictive performance. Experimental results demonstrate that the proposed eBERT-AP-DAL model consistently outperforms the original BERT model on the CLUE benchmark, both with and without full pre-training. Notably, on the iFLYTEK, TNEWS, CLUEWSC2020, and CSL datasets, the model achieves over 20% improvement in accuracy, underscoring its low-cost and highefficiency characteristics. Compared to the pre-trained BERT model, our approach requires only a small amount of pretraining data to achieve higher accuracy on the TNEWS and CLUEWSC2020 datasets. In the application of NIHSS-based stroke grading, the proposed model achieves an accuracy of 93%, further validating its effectiveness in real-world clinical scenarios.

With the rapid advancement of Generative Adversarial Networks (GANs) and Diffusion Models, the threshold for generating deepfake facial images has been significantly reduced. Consequently, forged facial expressions and synthetic face videos are widely disseminated across social media, psychological evaluation, human–computer interaction, and public opinion monitoring, posing serious threats to facial-based affective computing systems. Affective computing aims to recognize human emotional states through images or videos; however, when the input data are untrustworthy or artificially forged, the reliability of emotion recognition results is compromised and may even be maliciously exploited to mislead the model. Therefore, developing a trustworthy affective computing framework with anti-spoofing capability has become an important research direction in AI security.This paper proposes a facial authenticity verification framework for affective computing based on deepfake face detection. A pre-verification module is integrated before the traditional emotion recognition system to discriminate between authentic and forged facial inputs. First, mainstream deepfake detection techniques are systematically analyzed, and three representative models—DenseNet121, ResNet50, and SE-HRNet—are selected for comparative evaluation. Experiments are conducted on the Celeb-DFv2 and DFDC datasets. Results show that the three models achieved accuracy rates exceeding 90%.In summary, the proposed fusion framework of deepfake detection and emotion recognition significantly enhances the security and trustworthiness of affective computing systems. This approach shows great potential for applications in psychological behavior analysis, virtual education, intelligent customer service, and security monitoring.

As a fundamental expression of traditional Chinese culture, Sinology has become an integral component of Chinese language and cultural education programs for international students. However, two key challenges constrain effective Sinology learning: the abstract nature of classical Chinese cultural concepts and the gap between modern pedagogical methods and historical contexts. To address these issues, this paper presents an immersive Sinology learning framework that integrates Extended Reality (XR) technology with AI Agents. The framework aims to bridge cultural and temporal divides in contemporary Sinology instruction.