Browse Articles
Discover research articles across all indexed journals
Long-term outcomes after endoscopic resection of gastric adenocarcinoma of the fundic gland type and oxyntic gland adenoma: a retrospective cohort study
Talquetamab–Daratumumab in Relapsed or Refractory Myeloma
Initial data from the prospectively randomized G-MEMBRANE trial and systematic review on the embolization of the middle meningeal artery in the treatment of chronic subdural hematomas
Periodontitis aggravates high-fat diet-induced MASLD via gut microbiota dysbiosis and metabolic dysfunction in mice
Collagen gene expression profiles predict recurrence and progression of DCIS to IDC
One-part slag-mullite geopolymer: role of solid activators on connected porosity and mechanical properties
Visually challenging conditions on sign language intelligibility show behavioural analogies with spoken language
Abstract This study aimed to understand how visually degraded conditions affect the intelligibility of isolated signs in sign language, and how these conditions influence perceived difficulty. Twenty-nine fluent users of Swedish Sign Language viewed 100 isolated signs presented under two types of visual degradation: spectral degradation (pixelation) and background noise (salt-and-pepper noise), each across five levels of degradation. Participants identified each sign and rated the perceived difficulty. Generalized linear mixed models were used to evaluate the effect of level of degradation on intelligibility and perceived difficulty, separately and for each type of degradation. Psychometric functions were estimated, and generalized linear mixed models were used to analyse the relationship between intelligibility and perceived difficulty. Both types of degradation significantly reduced intelligibility and increased perceived difficulty. Psychometric curves showed that intelligibility plateaued around 70% accuracy, mirroring results from speech-in-noise research. Correct responses were consistently rated as easier to perceive, and an interaction between degradation level and response accuracy was observed only in the spectral degradation condition, suggesting distinct perceptual processing mechanisms. Visual degradations affect sign language intelligibility in ways comparable to auditory distortions in speech, with differences between degradation types suggesting distinct perceptual strategies. These findings have implications for theories about linguistic processing in perceptually challenging conditions and call for further investigation into cognitive and linguistic factors that influence sign perception in degraded conditions.
On the smart coordination of flexibility scheduling in multi-carrier integrated energy systems
Abstract Coordinating the interactions among flexibility assets in multi-carrier integrated energy systems (MIES) can lead to a cost-efficient energy transition. However, the proliferation of flexibility assets and their growing participation in active demand response increases the complexity of coordinating these interactions. This paper investigates several approaches to model the coordination of flexibility scheduling in MIES with many autonomous flexibility providers. We propose runtime model coupling as an alternative modeling paradigm to overcome the limitations of monolithic centralized co-optimization. Specifically, we introduced two model coupling approaches—a distributed price-response and a decentralized market auction approach—to address practical challenges such as preserving the autonomy and privacy of flexibility providers while ensuring scalability. We conduct a quantitative benchmark of these approaches against co-optimization across varying problem sizes, complexities, and computing infrastructures. This benchmark provides new empirical insights into the trade-offs between optimality, autonomy, and scalability that have so far remained unquantified in the energy system literature. We show that model coupling offers a method to balance optimality and realism (autonomy and privacy) while delivering substantial scalability gains. The proposed model coupling approaches are formalized as open-source software with several practical applications: modelers can experiment with different flexibility modeling approaches and choose the one that best matches their modeling objectives and constraints; flexibility providers can couple their models to simulate interactions between their systems to make informed operational decisions without disclosing any confidential information.
Visual deep learning approaches for alphabetic sign language interpretation
LaRHP: latent-aware reconstruction via hypersphere projection for industrial image anomaly detection
Whole transcriptome sequencing and ceRNA regulatory network in diabetic peripheral neuropathy
Propensity-matched comparison of early femoral complications after cementless total hip arthroplasty with different stem designs
Regional transcriptomic divergence reveals thermal adaptation mechanisms in the giant kelp Macrocystis pyrifera
Contrastive unlearning via representation editing for graph neural networks
Cassipourol and β-sitosterol from Malva parviflora L.: a mechanistic study of dual anti-inflammatory action against COX/LOX and TNF-α/BCL-2
Abstract Although inflammation protects our bodies against harmful stimuli, uncontrolled inflammation drives serious chronic disorders. Malva parviflora L. (family Malvaceae) may represent a source for anti-inflammatory metabolites based on its potent anti-inflammatory activity. Its total ethanol extract demonstrated notable inhibition of cyclooxygenase-1/2 (COX-1/COX-2) and 5-lipoxygenase (5-LOX), with IC 50 values comparable to those of the reference drugs indomethacin and zileuton, respectively. The hexane fraction was the most active fraction (lowest IC₅₀) among the solvent partitions (dichloromethane, ethyl acetate, and butanol). Subsequent column chromatography of the hexane fraction produced two compounds: Cassipourol, isolated for the first time from the Malvaceae family, and β -sitosterol. Their structures were confirmed by matching their NMR and mass spectrometry data with literature. Both were validated in vitro as dual COX/LOX inhibitors, exhibiting IC₅₀ values comparable to those of the standards. To uncover additional mechanisms, a compound–target–inflammation network was constructed using network pharmacology approaches, revealing 178 shared targets. Among these, tumor necrosis factor (TNF- α ) and the antiapoptotic protein B-cell lymphoma 2 (BCL-2) emerged as central nodes linked to inflammatory pathways. Subsequent assays in human colon carcinoma (Caco-2) and lung adenocarcinoma (A549) cell lines showed that β -sitosterol suppressed TNF- α and BCL-2 by approximately 55%, whereas cassipourol displayed only modest inhibition (~20%). Molecular docking predicted moderate ( ca. –4.5) and strong (> –5) binding affinities of both compounds to key inflammatory targets. Collectively, these results suggest that β -sitosterol from M. parviflora is a promising multitarget lead for inflammatory disorders, including cancer, whereas cassipourol requires further structural optimization and mechanistic investigation to improve its unfavorable physicochemical properties.
Enhanced antimicrobial activity, mechanical and dielectric properties of styrene butadiene rubber vulcanizates designed for protection products
Abstract This study focuses on fabricating and evaluating novel, ecofriendly, green, flexible, antimicrobial Styrene Butadiene Rubber vulcanizates for producing safety products. Styrene butadiene rubber (SBR) was compounded with ginger extract, a plant rich in bioactive phenolic compounds, and the conventional drugs, berberine and chitosan, in the ordinary mixer of rubber. Ginger ( Zingiber officinale ) extract was analyzed using HPLC. The rheological properties of the SBR mixes were analyzed to determine the optimal curing time. The compounded rubber mixes were then vulcanized at 152 °C. The prepared vulcanizates were evaluated using Fourier-transform infrared spectroscopy (FTIR), specific surface area measurements, mechanical, swelling, antimicrobial properties, and cytotoxicity analysis. The results showed that the physicochemical and mechanical properties of all vulcanizates remained robust, i.e., no significant degradation, even after exposure to thermal oxidative aging at 90 °C for seven days. Therefore, the change of tensile strength percentage after aging was 25, 5.56, 13, and 10% for free SBR, SBR/30 phr chitosan, SBR/20 phr ginger, and SBR/7 phr berberine, respectively. The cytotoxicity tests confirmed the safety of the investigated vulcanizates due to the negative results against the normal human fibroblast cell line (BJ1). Furthermore, the antimicrobial activity results demonstrated that the release of the various antimicrobial agents successfully inhibited the growth of different bacteria and fungi on the SBR rubber surface. The dielectric and electric findings highlighted the ability of the presence of bioactive agents to modify dielectric relaxation and conductivity in the investigated SBR vulcanizates and confirmed the utility of SBR composites contained 30 phr chitosan, 20 phr ginger, and 7 phr berberine as a good viable choice for antistatic and flexible electronic applications.
Analysis of the ERG25 gene family based on the whole-genome sequence in Trametes versicolor and its response to light stress
Convolutional low-rank adaptation for efficient semantic segmentation in vision transformers
Abstract Vision Transformers (ViTs) have shown remarkable performance across various computer vision tasks, but their fine-tuning for dense prediction tasks such as semantic segmentation remains computationally intensive. This work proposes a novel dual-task architectural application of the LyCORIS Low-Rank Adaptation for Convolutions (LyCORIS LoCon) framework, which introduces learnable low-rank convolutional modules into pre-trained ViTs. This method is applied to Depth Anything V2 (DAV2), augmenting its decoder to support dual-task outputs; monocular depth estimation and binary human semantic segmentation, without disrupting its original capabilities. By injecting only 150K trainable parameters, this approach significantly reduces the adaptation cost while achieving segmentation performance comparable to state-of-the-art models like SAM, MaskFormer, and SegFormer. Extensive experiments on filtered COCO 1 and ImageNet subsets show that Conv-LoRA enhances task-specific learning with minimal computational overhead. The method achieves an mAP of 89.69% and an mIoU of 79.17% for human segmentation, performing competitively alongside state-of-the-art models like Mask2Former, while preserving the depth prediction accuracy of the base model.