Explore the vast range of titles available.

Skeleton-based human action recognition has made great progress, especially with the development of a graph convolution network (GCN). The most important work is ST-GCN, which automatically learns both spatial and temporal patterns from skeleton sequences. However, this method still has some imperfections: only short-range correlations are appreciated, due to the limited receptive field of graph convolution. However, long-range dependence is essential for recognizing human action. In this work, we propose the use of a spatial-temporal relative transformer (ST-RT) to overcome these defects. Through introducing relay nodes, ST-RT avoids the transformer architecture, breaking the inherent skeleton topology in spatial and the order of skeleton sequence in temporal dimensions. Furthermore, we mine the dynamic information contained in motion at different scales. Finally, four ST-RTs, which extract spatial-temporal features from four kinds of skeleton sequence, are fused to form the final model, multi-stream spatial-temporal relative transformer (MSST-RT), to enhance performance. Extensive experiments evaluate the proposed methods on three benchmarks for skeleton-based action recognition: NTU RGB+D, NTU RGB+D 120 and UAV-Human. The results demonstrate that MSST-RT is on par with SOTA in terms of performance.

Vision–language models (VLMs) have demonstrated remarkable performance on a wide range of multimodal reasoning tasks, yet their visual grounding mechanisms remain poorly understood and are often unreliable for fine-grained visual concepts. Existing approaches typically rely on raw attention maps or gradient-based saliency, which provide heuristic explanations but lack a causal interpretation of how visual evidence contributes to model predictions. In this paper, we propose an Information Bottleneck Score (IBS) framework that explicitly quantifies the causal importance of visual patches through interventional analysis. By masking candidate image patches and measuring the induced change in the model prediction, the IBS captures patch-level causal contributions rather than correlation-based signals. We further lift patch-level importance to the attention-head level by aggregating the IBS with text-to-image attention, enabling the identification of a small subset of information-transmitting attention heads responsible for visual grounding. Building on the selected heads, we construct refined importance maps that guide visual cropping in a fully training-free manner. Extensive experiments on multiple detail-sensitive benchmarks, including TextVQA, V*, POPE, and DocVQA, demonstrate consistent improvements in fine-grained visual understanding, while evaluations on general-purpose datasets such as GQA, AOKVQA, and VQAv2 confirm that overall reasoning performance is preserved. Additional ablation studies further validate the effectiveness of each component in the proposed framework. Overall, our work provides a causal perspective on visual grounding in VLMs and offers a model-agnostic, training-free approach for both interpreting and enhancing multimodal reasoning.

Background PD-1 and PD-L1 inhibitors have emerged as promising therapies for advanced colorectal cancer (CRC), but their efficacy and safety profiles require further evaluation. This meta-analysis aims to assess the efficacy and safety of PD-1/PD-L1 inhibitors in this patient population. Methods A systematic review and meta-analysis were conducted following PRISMA guidelines, with data sourced from PubMed, Embase, CENTRAL, Web of Science, and CNKI up to August 3, 2024. Nine randomized controlled trials (RCTs) involving 1680 patients were included. The primary outcomes were overall survival (OS), progression-free survival (PFS) and objective response rate (ORR), while safety was assessed through adverse events (AEs) and grade ≥ 3 AEs. Effect sizes were calculated using mean differences (MD) and risk ratios (RR) with 95% confidence intervals (CIs). Results Overall, the meta-analysis showed that PD-1/PD-L1 inhibitors did not significantly extend OS (MD = 0.86, 95% CI: -0.55, 2.27), but they significantly improved PFS (MD = 2.53, 95% CI: 0.92, 4.15). Additionally, PD-1/PD-L1 inhibitors did not significantly increase the ORR compared to controls (RR = 1.19, 95% CI: 0.99, 1.44). In terms of safety, PD-1/PD-L1 inhibitors did not significantly increase the incidence of overall AEs. Subgroup analysis further indicated that PD-1 inhibitors significantly improved OS (MD = 1.24, 95% CI: 0.20, 2.29) and PFS (MD = 6.27, 95% CI: 0.56, 11.97), while PD-L1 inhibitors did not have a significant impact on these outcomes. Additionally, PD-L1 inhibitors were associated with a higher risk of grade ≥ 3 AEs (RR = 1.29, 95% CI: 1.07, 1.57), a risk not observed with PD-1 inhibitors. Conclusion PD-1 inhibitors significantly improve PFS and OS in advanced CRC, making them a preferable option over PD-L1 inhibitors, which show limited efficacy and a higher risk of severe AEs. These findings support prioritizing PD-1 inhibitors in clinical practice for this patient group, while caution is warranted with PD-L1 inhibitors due to their safety concerns. Trial registration PROSPERO (CRD42024611696).

Depth image-based rendering (DIBR) is an important technology in the process of 2D-to-3D conversion. It uses texture images and related depth maps to render virtual views. While there are still some challenging problems in the current DIBR systems, such as disocclusion occurrences. Inpainting methods based on deep learning have recently shown significant improvements and generated plausible images. However, most of these methods may not deal well with the disocclusion holes in the synthesized views, because on the one hand they only treat this issue as generative inpainting after 3D warping, rather than following the full DIBR processing procedures. While on the other hand the distributions of holes on the virtual views are always around the transition regions of foreground and background, which makes them more difficult to distinguish without special constraints. Motivated by these observations, this paper proposes a novel learning-based method for stereoscopic view synthesis, in which the disocclusion regions are restored by a progressive structure reconstruction strategy instead of direct texture inpainting. Additionally, some special cues in the synthesized scenes are further exploited as constraints for the network to alleviate hallucinated structure mixtures among different layers. Extensive empirical evaluations and comparisons validate the strengths of the proposed approach and demonstrate that the model is more suitable for stereoscopic synthesis in the 2D-to-3D conversion applications.

Inflammatory bowel disease (IBD) comprises chronic relapsing disorders of the gastrointestinal tract characterized pathologically by intestinal inflammation and epithelial injury. Here, we uncover a function of extracellular matrix protein 1 (ECM1) in promoting the pathogenesis of human and mouse IBD. ECM1 was highly expressed in macrophages, particularly tissue-infiltrated macrophages under inflammatory conditions, and ECM1 expression was significantly induced during IBD progression. The macrophagespecific knockout of ECM1 resulted in increased arginase 1 (ARG1) expression and impaired polarization into the M1 macrophage phenotype after lipopolysaccharide (LPS) treatment. A mechanistic study showed that ECM1 can regulate M1 macrophage polarization through the granulocyte-macrophage colony-stimulating factor/ STAT5 signaling pathway. Pathological changes in mice with dextran sodium sulfate-induced IBD were alleviated by the specific knockout of the ECM1 gene in macrophages. Taken together, our findings show that ECM1 has an important function in promoting M1 macrophage polarization, which is critical for controlling inflammation and tissue repair in the intestine.

Depth image-based rendering (DIBR) is an important technology in the process of 2D-to-3D conversion. It uses texture images and related depth maps to render virtual views. While there are still some challenging problems in the current DIBR systems, such as disocclusion occurrences. Inpainting methods based on deep learning have recently shown significant improvements and generated plausible images. However, most of these methods may not deal well with the disocclusion holes in the synthesized views, because on the one hand they only treat this issue as generative inpainting after 3D warping, rather than following the full DIBR processing procedures. While on the other hand the distributions of holes on the virtual views are always around the transition regions of foreground and background, which makes them more difficult to distinguish without special constraints. Motivated by these observations, this paper proposes a novel learning-based method for stereoscopic view synthesis, in which the disocclusion regions are restored by a progressive structure reconstruction strategy instead of direct texture inpainting. Additionally, some special cues in the synthesized scenes are further exploited as constraints for the network to alleviate hallucinated structure mixtures among different layers. Extensive empirical evaluations and comparisons validate the strengths of the proposed approach and demonstrate that the model is more suitable for stereoscopic synthesis in the 2D-to-3D conversion applications.

Unsupervised domain adaptation (UDA) aims to adapt a model learned from the source domain to the target domain. Thus, the model can obtain transferable knowledge even in target domain that does not have ground truth in this way. In medical image segmentation scenarios, there exist diverse data distributions caused by intensity in homogeneities and shape variabilities. But multi source data may not be freely accessible, especially medical images with patient identity information. To tackle this issue, we propose a new multi-source and source-free (MSSF) application scenario and a novel domain adaptation framework where in the training stage, we only get access to the well-trained source domain segmentation models without source data. First, we propose a new dual consistency constraint which uses domain-intra and domain-inter consistency to filter those predictions agreed by each individual domain expert and all domain experts. It can serve as a high-quality pseudo label generation method and produce correct supervised signals for target domain supervised learning. Next, we design a progressive entropy loss minimization method to minimize the class-inter distance of features, which is beneficial to enhance domain-intra and domain-inter consistency in turn. Extensive experiments are performed for retinal vessel segmentation under MSSF condition and our approach produces impressive performance. The sensitivity metric of our approach is highest and it surpasses other methods with a large margin. It is the first attempt to conduct researches on the retinal vessel segmentation task under multi-source and source-free scenarios. In medical applications, such adaptation method can avoid the privacy issue. Furthermore, how to balance the high sensitivity and high accuracy need to be further considered.

In this study, we report the antibacterial activities of six polyphenols (i.e., luteolin, quercetin, scutellarin, apigenin, chlorogenic acid, and resveratrol) against 29 clinical isolates of methicillin-resistant Staphylococcus aureus (MRSA), and in vitro antibacterial activities of two-drug combinations. All of the MRSA strains evaluated were clinical isolates from patients with MRSA bacteremia. The antibacterial activities were determined by agar dilution method, and the two-drug antibacterial activities were determined by the checkerboard agar dilution method. It was found that luteolin, quercetin and resveratrol show obvious antibacterial activities against MRSA, and the results of two-drug antibacterial activity show either synergy or additivity, without evidences of antagonistic effects.

This paper reports the occurrence of six kinds of commercial enzyme hydrolysis effects for use in grape juice and cherry juice, which provide a basis for studying the bound aroma compounds in fruit juice and their application in production. Using headspace solid-phase microextraction combined with GC–MS, a reliable procedure for determining the free and glycosidic-bound volatile compounds has been established. Comparison of these free and bound aroma compounds revealed that non-volatile glycosides, known as aroma precursors, occur at high concentrations in grape and cherry juice. Using six different glycosidase enzymes, 67 volatile compounds were identified in these two juices, including terpenes, C13-norisoprenoids, higher alcohols, esters, C6-compounds, C9-compounds, and phenols. The different enzymes had significant effects on varietal aroma. Creative and AR2000 had similar hydrolysis effects, which contribute greatly to the characteristic aroma of grape juice and cherry juice, significantly enhance the floral and fruity features of fruit juice, and improve aroma complexity in the system. The Creative enzyme can be used as a new choice for studying juice-bound aroma and hydrolysis-bound aroma in fruit and wine production.

Antioxidant activity and neuroprotective activity of three stilbenoids, namely, trans-4-hydroxystilbene (THS), trans-3,5,4′-trihydroxy-stilbene (resveratrol, RES), and trans-3′,4′,3,5-tetrahydroxy-stilbene (piceatannol, PIC), against β-amyloid (Aβ)-induced neurotoxicity in rat primary cortex neurons were evaluated. THS, RES, and PIC significantly scavenged DPPH• and •OH radicals. All three stilbenoids were able to inhibit Aβ neurotoxicity by decreasing intracellular reactive oxygen species (ROS) via the PI3K/Akt signalling pathway. Specifically, stilbenoids significantly promoted Akt phosphorylation; suppressed Bcl-2/Bax expression; and inhibited caspase-9, caspase-3, and PARP cleavage. Molecular docking between stilbenoids with Akt indicated that stilbenoids could form hydrogen bond interactions with the COOH-terminal region of Akt. Additionally, the neuroprotective activity of stilbenoids correlated with the number and position of hydroxyl groups. The lack of meta-dihydroxyl groups on THS did not affect its neuroprotective activity in comparison with RES, whereas the ortho-dihydroxyl moiety on PIC significantly enhanced neuroprotective activity. These results provide new insights into the correlation between the biological activity and chemical structure of stilbenoids.