Jinling Institute of Technology

Remote sensing object detection (RSOD) models play an increasingly important role in modern remote sensing systems. However, during model delivery, sharing, and deployment, RSOD models face increasing risks of unauthorized redistribution, illegal replication, and intellectual property infringement. To mitigate these threats, this paper proposes a white-box watermarking framework for RSOD models that enables reliable copyright verification while preserving the performance of the primary detection task. Specifically, a gradient-based sensitivity analysis of the detection loss is first performed to adaptively identify model parameters that minimally affect detection performance, which are then selected as watermark carriers. Subsequently, a parameter-ranking-based watermark encoding scheme is developed, where watermark bits are embedded by enforcing relative ordering constraints between parameter pairs. To further improve robustness under practical deployment conditions, an attack-simulation-driven training strategy is introduced, in which common perturbations and watermark removal attacks are simulated during the embedding process. In addition, a stealthiness enhancement strategy based on statistical distribution constraints is designed to maintain consistency between the distribution of watermarked parameters and those of the original model, thereby reducing the risk of watermark exposure and localization. Extensive experiments across multiple RSOD datasets and detection architectures demonstrate that the proposed method achieves a high copyright verification success rate with negligible impact on detection accuracy and exhibits strong robustness and stealthiness against a variety of watermark removal attacks.

Excessive mechanical stress is a major cause of intervertebral disc degeneration (IVDD). Macrophages can sense physical signals, but their role in responding to mechanical stress within the disc to maintain homeostasis is unclear. This study investigates the function of macrophage-derived legumain (LGMN) in IVDD. Single-cell RNA sequencing data of human disc samples were analyzed. Macrophage-specific Lgmn knockout (LgmnF/F;LysMCre) and nucleus pulposus cell (NPC)-specific Yap1 knockin (Yap1LSL/LSL; Col2a1Cre) mice were generated to study IVDD progression in vivo using a lumbar spine instability model. In vitro, NPCs and macrophages were cultured under mechanical compression. Molecular interactions were predicted with AlphaFold3 and validated by coimmunoprecipitation and mass spectrometry. Signaling pathways were analyzed via RNA sequencing, western blot, and chromatin immunoprecipitation. Engineered LGMN-overexpressing small extracellular vesicles (sEVs) were tested therapeutically in a rat compression model. LGMN was significantly upregulated in human and animal degenerate discs, primarily in macrophages. Conditional knockout in macrophages accelerated IVDD in mice. Mechanistically, macrophage-derived LGMN bound to integrin αvβ3 on NPCs, inhibiting RhoA activity and activating the Hippo pathway. This led to phosphorylation and cytoplasmic retention of YAP1, which suppressed mechanical stress-induced ferroptosis in NPCs. Mechanical stress promoted STAT3 nuclear translocation in macrophages, directly enhancing LGMN transcription. Intradiscal delivery of LGMN-enriched sEVs alleviated IVDD in rats. Macrophage-derived LGMN is a key mechanosensitive regulator that ameliorates IVDD by inhibiting NPC ferroptosis via the integrin αvβ3–Hippo pathway, revealing a novel endogenous protective mechanism and a potential therapeutic strategy.

The implementation of the new national electric bike (e-bike) standard has imposed stringent compliance requirements on equipment and e-bikes in the instant delivery sector, which directly affects the delivery efficiency and the work adaptability of food delivery riders. This study aims to investigate food delivery riders’ continued usage intention of e-bikes under China’s new e-bike regulation. Based on valid data collected from food delivery riders in Nanjing, this study employs ordered logit regression to examine the primary factors influencing their continued usage intention of e-bikes. The findings reveal that: (1) Male riders’ willingness to continue using e-bikes is comparatively lower, whereas older riders show a stronger intention. (2) Food delivery riders with higher incomes and those who need to replace their e-bikes show a stronger inclination to continue using them. (3) Limited e-bike options have a significant negative effect on riders’ continued usage intention, while speed limits exert no significant influence. Based on these empirical findings, corresponding policy recommendations are proposed to promote riders’ continued use of e-bikes, such as developing age-friendly delivery models, establishing an income guarantee mechanism for riders, and optimizing platform delivery time allocation. The findings could provide a theoretical basis and practical insights for policymakers and food delivery platforms to improve e-bike management policies.

Adaptive beamforming (ABF) can improve the signal-to-interference-plus-noise ratio (SINR) of radar systems through the suppression of interference and by maintaining the desired signal. However, unavoidable array defects will cause significant performance degradation in real scenarios because of sensor position error. To address this challenge, an effective ABF based on the flamingo search algorithm (FSA) is established, referred to as FSABF. The strong global search capability and fast convergence of FSA are exploited to optimize the beamforming weights. As a result, the main lobe is accurately directed toward the target, while deep nulls are imposed in interference directions, thereby significantly improving the output SINR. In addition, an early-stopping strategy is introduced to optimize the iteration process. The resulting beamformer, namely, FSABFE, maintains excellent beamforming performance while reducing computational overhead to 11.90% of that of FSABF, thereby significantly enhancing overall efficiency. This advantage makes the proposed approach more suitable for practical radar applications in situations featuring limited computational resources. The simulation results show that the proposed FSABF and FSABFE achieve robust beam control under sensor position error.

Introduction: Current guidelines lack consensus on fibrinogen replacement thresholds in critical hemorrhage. This study aims to establish evidence-based supplementation thresholds and optimal targets for ICU patients with major bleeding, providing a framework for individualized coagulation management. Methods: This prospective multicenter cohort study enrolled 197 patients with severe hemorrhage and coagulopathy from 12 tertiary ICUs in China. Standardized systems collected demographics, laboratory parameters, and interventions. Analysis comprised two phases: (1) Restricted cubic spline (RCS) models evaluated nonlinear associations between baseline fibrinogen and outcomes (hemostasis, ICU mortality), identifying a critical threshold at 1.54 g/L; (2) Post-treatment fibrinogen peaks (within 72h) were stratified into subgroups based on prior MIMIC-IV validation: < 1.54 g/L (low-risk), 1.54–2.0 g/L (intermediate), 2.0–2.5 g/L (target), ≥2.5 g/L (high-level). Statistical analyses included Cochran-Armitage test (dose-response), Firth’s regression (small-sample bias correction), and Kruskal-Wallis tests with Benjamini-Hochberg FDR correction (FDR< 0.05; transfusion heterogeneity). Results: We enrolled 197 patients with massive hemorrhage (treatment: n=98; control: n=99). Baseline data showed the treatment group had significantly greater blood loss (1735vs.967 mL, P=0.013), lower fibrinogen (1.49vs.2.04 g/L, P=0.001), and higher APACHE II scores (19.5vs.16.7, P=0.047). Pretreatment fibrinogen showed nonlinear associations with hemostasis and ICU mortality (both Pnon-linear< 0.001), identifying a critical threshold at 1.54 g/L. Post-treatment fibrinogen levels of 2.0-2.5 g/L achieved a hemostasis rate of 75%-80%; higher levels (2.5-3.5 g/L) showed slightly reduced efficacy. Mortality was 35% for levels < 1.5 g/L, significantly decreasing to 15% at 2.0-2.5 g/L, and approaching zero at 2.5-3.5 g/L. Post-treatment levels also showed nonlinear associations with outcomes (both Pnon-linear< 0.001), with a critical threshold at 2.5 g/L. Conclusions: Fibrinogen levels predict outcomes in critically ill patients with massive hemorrhage. Initiation of fibrinogen supplementation is recommended when levels fall below 1.5g/L, targeting 2.0–2.5g/L as the optimal therapeutic window to improve prognosis.