Browse Articles

An art style classification network integrating contrastive learning and counterfactual attention

Art style classification is a fundamental task for digital art analysis and intelligent cultural heritage management. However, existing methods often fail to deliver stable and interpretable results due to dispersed style cues, complex local textures, strong interference from background and subject content, and the difficulty of identifying regions that are causally relevant to style recognition. To address these issues, we propose a Multi Source Collaborative Style Network, termed MCS-Net, which jointly models global composition and local brushstroke textures within a unified framework. MCS-Net consists of four modules, namely a basic style feature encoding module, an attention generation module, a style contrastive learning module, and a counterfactual attention module. The attention generation module discovers multiple complementary discriminative regions. The contrastive learning module constructs a more discriminative embedding space to improve the separability of fine-grained styles. The counterfactual attention module builds counterfactual attention samples to explicitly estimate the true contribution of regions to predictions, thereby suppressing spurious activations, improving robustness, and providing counterfactual-supported evidence for model interpretation. Experiments on three public datasets, WikiArt, MultitaskPainting100k, and Pandora18k, demonstrate that MCS-Net outperforms representative baselines across standard evaluation metrics.

Ultranarrow bright single-photon emitters in diamond with strong broadband phonon decoupling

Ultrasound-guided microwave ablation versus conventional surgery for Granulomatous Lobular Mastitis: A two-center retrospective comparative stud

Background Granulomatous lobular mastitis (GLM) is a chronic inflammatory breast disease that causes substantial pain, cosmetic deformity, and psychosocial distress. Evidence directly comparing ultrasound-guided microwave ablation (MWA) with conventional surgery (CS) remains limited. Methods We conducted a two-center retrospective cohort study of consecutive women with core needle biopsy-confirmed GLM treated between March 2020 and February 2025. Treatment allocation was non-random and reflected routine clinical decision-making. The primary outcomes were overall cure and recurrence. Secondary outcomes included perioperative resource utilization and patient-reported outcomes assessed with a BREAST-Q domain-based questionnaire adapted for GLM follow-up. Baseline balance was described with both P values and standardized mean differences (SMDs). Results Among 233 patients (MWA, n = 102; CS, n = 131), overall cure was achieved in 97.06% and 96.18% of patients, respectively (risk difference, 0.88 percentage points; 95% CI, −3.76 to 5.51; P = 1.000). Recurrence occurred in 4.90% of the MWA group and 3.82% of the CS group (risk difference, 1.09 percentage points; 95% CI, −4.24 to 6.41; P = 0.751) during a median follow-up of 18.5 months. Compared with CS, MWA was associated with lower estimated blood loss, shorter procedure time, lower postoperative drainage burden, and shorter hospitalization (all P  < 0.001), while total hospitalization cost did not differ significantly. Questionnaire scores favored MWA for sexual well-being, breast satisfaction, and scar satisfaction, whereas psychological well-being, chest physical well-being, and satisfaction with the medical team were similar between groups. SMDs indicated some baseline imbalance, particularly for multiple-quadrant involvement and ultrasound lesion size. Conclusions In this retrospective observational comparison, MWA and CS were associated with similarly high cure rates and low recurrence. MWA was additionally associated with lower perioperative burden and better breast- and scar-related satisfaction. Prospective studies with standardized co-interventions, confounding adjustment, and validated patient-reported outcome assessment are warranted.

Coupled dual-channel memristors for hardware-native trustworthy Bayesian intelligence

More than a digestive disorder: Perceived stigma in inflammatory bowel disease patients

Background Inflammatory bowel disease (IBD) is a chronic condition characterized by symptoms that are often difficult to conceal and may expose patients to negative social reactions. Beyond physical burden, stigma represents an important but underrecognized psychosocial challenge in IBD, potentially affecting social functioning and disease management. To determine the prevalence of perceived stigma among patients with IBD and to evaluate its association with disease subtype, sociodemographic characteristics, clinical features, and self-esteem. Methods A total of 146 patients with IBD under regular follow-up were included in this cross-sectional study. Perceived stigma was assessed using ten disease-specific yes/no items reflecting real-life stigma-related experiences. Sociodemographic data, clinical characteristics, and self-esteem levels measured by the Rosenberg Self-Esteem Scale were recorded. Patients were classified as experiencing stigma if they endorsed at least one stigma item. Results Overall, 63.0% of patients reported at least one stigma-related experience. The most frequently reported manifestations were social avoidance and discomfort in public settings, while overt discrimination was less common. The distribution of stigma-related experiences was largely similar between Crohn’s disease and ulcerative colitis patients. Individuals reporting stigma were significantly younger and had a younger age at diagnosis compared to those without stigma. No significant differences were observed with respect to gender, marital status, education, employment, income, disease subtype, surgical history, or comorbidities. Patients experiencing stigma had significantly lower self-esteem scores than those without stigma. Conclusion Perceived stigma is common among patients with IBD, particularly among younger individuals, and is more often expressed as social discomfort and avoidance rather than explicit discrimination. These findings highlight stigma as an important psychosocial dimension of IBD that warrants greater clinical attention. Integrating psychosocial support into routine care may help mitigate stigma-related consequences and improve overall disease management. Longitudinal studies are needed to clarify the long-term impact of stigma on disease outcomes.

Giant tunable barocaloric effects using guest-molecule adsorption in a porous metal-organic framework

Abstract Metal-organic frameworks (MOFs) that exhibit first-order structural phase transitions with significant volume changes are promising candidates for barocaloric cooling and heating applications. Here we report giant barocaloric effects in a zeolitic-imidazolate framework [ZIF-4(Zn)] that can be reversibly driven over a wide range of temperatures with pressure, near a starting temperature that is highly tunable using gas adsorption. For nitrogen-loaded frameworks of ZIF-4(Zn), we report isothermal entropy changes |∆ S it | > 38 J K -1 kg -1 and adiabatic temperature changes |∆ T ad | > 10 K driven reversibly by |∆ p | = | p - p atm | ~ | p | ~ 1 kbar ( p atm  ~ 1 bar) near T 0  ~ 190 K. For desolvated frameworks of ZIF-4(Zn), we estimate larger barocaloric effects up to |∆ S it | ~ 210 J K -1 kg -1 and |∆ T ad | ~ 60 K near T 0  ~ 137 K driven using the same pressure change. Due to the dynamic behaviour of this porous MOF, a combination of quasi-direct, indirect and direct methods was used to determine the caloric performance. This combination of methods underlines the cooling and heating potential of porous MOFs, while highlighting the necessity of quantifying the presence of any inactive mass. Our findings should establish a benchmark for MOFs as environmentally benign solid-state refrigerants that can be tuned using guest molecules as a caloric vector.

Cross-market spillovers amid policy uncertainty era of China’s economy

This study meticulously discusses spillover effects among agricultural futures (AGF), crude oil prices (CRO), and the U.S. dollar (USD) during the volatile phase of China’s economic policy (CEPU), utilizing the quantile connectedness approach. The empirical findings show that, in the face of economic instabilities and tail incidents, AGF, CRO, USD, and CEPU exhibit significantly higher return spillover at extreme quantiles in comparison to mean one. Furthermore, the dynamic properties of AGF, CRO, USD, and CEPU are empirically substantiated, and certain crises obviously enhance these spillover effects. The primary innovation involves the integration of CEPU into the analytical framework of the agriculture-energy-financial system. This integration illustrates the interdependencies of the system across a variety of market conditions through dynamic quantile shifts, thereby expanding the theoretical scope. This study recommends actions for different market participants, such as an cross-nation economic cooperation and financial market tracking mechanism.

Retraction Note: Targeted production of reactive oxygen species in mitochondria to overcome cancer drug resistance

Quantitative correlation of rock fragment gradation parameters and TBM disc cutter efficiency: Insights from DEM simulations

This study explores the relationship between rock fragment size distribution and disc cutter breaking efficiency of a tunnel boring machine (TBM) using discrete element simulations of indentation tests. Fragmentation behavior was analyzed under varying joint dip angles, joint spacings, and confining pressures, and cutter efficiency was evaluated by correlating fragment gradation parameters with specific energy (SE). The results show that cracks preferentially propagate along the weak joint interfaces, demonstrating that joint geometry controls stress redistribution and fracture evolution. With increasing joint dip angle, crack density and crushed zone extent first increase and then decrease. Reduced joint spacing enhances crack interaction and coalescence, resulting in lower energy demand. In contrast, increasing confining pressure suppresses tensile crack development, restricts crack opening displacement, and promotes shear-dominated failure, leading to finer fragments and higher SE. Strong linear correlations (R² > 0.80) between gradation parameters and SE indicate that fragment size distribution quantitatively reflects variations in energy dissipation during cutter and rock interaction.

Personalized targeting of BCL2 family proteins overcomes acquired resistance to BRAF-MEK inhibitors in preclinical melanoma

Abstract There are currently no effective targeted therapies for BRAF-mutant metastatic melanoma patients with acquired resistance to approved BRAF and MEK inhibitors (BRAFi and MEKi), and very few ongoing clinical trials. Anti-apoptotic BCL2 family proteins promote de novo resistance to several therapies, including single-agent BRAFi in BRAF-mutant melanomas. In this study, in vivo testing of a large collection of patient-derived xenograft (PDX) models from melanoma patients with acquired resistance to BRAFi or BRAFi+MEKi shows that combining BCL2 inhibitors (BCL2i; navitoclax or venetoclax) with BRAFi+MEKi induces tumor regressions in a subset of these PDXs. High basal BCL2 predicts response whereas high basal MCL1 predicts resistance to this strategy. MCL1 overexpression studies functionally validate its role in resistance. Further, combining BRAFi+MEKi with an MCL1 inhibitor (MCL1i) counteracts resistance and interestingly decreases MCL1i-associated markers of cardiotoxicity. Together these studies identify potential personalized strategies to improve outcomes in this challenging patient population.

Cross-cultural adaptation of evidence-based practice measure among Hong Kong healthcare providers

Objective This study aimed to adapt and validate an Evidence-Based Practice (EBP) tool for healthcare providers in Hong Kong, addressing the need for reliable and culturally relevant tools. Methods The original EBP measure, developed for recent Canadian physiotherapy and occupational therapy graduates, was systematically adapted for use by healthcare providers in Hong Kong following established cross-cultural adaptation guidelines. The adaptation process consisted of two phases: Phase I involved forward and backward translation, expert panel review, and cognitive debriefing interviews with a purposive sample of 36 healthcare providers to ensure linguistic and cultural relevance. In Phase II, the finalized instrument was administered to a convenience sample of 248 registered healthcare providers in Hong Kong. The structure of the adapted instrument remained consistent with the original, comprising two models encompassing six constructs: (1) a formative model (use of EBP and EBP activities) and (2) a reflective model (knowledge, self-efficacy, attitudes, and resources). Construct validity was assessed using the Rasch model, internal consistency reliability was evaluated with the Person Separation Index (PSI), and differential item functioning (DIF) was examined. Results The formative model only required linguistic modifications. The Rasch model was applied to the reflective component. For knowledge, 45% (5/11) items fit the Rasch model with a chi-square fit statistic (χ² = 17.90, p = 0.268; PSI = 0.84). For self-efficacy, 89% (8/9) items fit the Rasch model with χ² = 26.48, p = 0.33; PSI = 0.93. The attitudes construct was divided into positively and negatively worded subscales due to multidimensionality. The resources construct showed a good fit with χ² = 26.60, p = 0.49; PSI = 0.84. DIF was not observed in the final measures. Conclusions The adapted EBP measure demonstrated evidence of construct validity and internal consistency reliability among healthcare providers in Hong Kong. However, further research is needed to assess additional aspects of validity, such as test–retest reliability and responsiveness, as well as to evaluate its applicability in broader healthcare settings.

Experience-dependent modulation of collective behavior in larval zebrafish

Prevalence of hyperuricemia in preeclampsia: A systematic review and meta-analysis of studies from low - and middle - income countries

Background Hyperuricemia is a recognized biochemical finding in preeclampsia (PE), but the reported frequency varies across low- and middle-income countries (LMICs). A pooled estimate from LMIC settings may help clarify the extent of this finding in hospital-based populations. Objective To estimate the pooled prevalence of hyperuricemia among women with preeclampsia in hospital-based studies from LMICs. Methods A systematic review and meta-analysis was conducted according to PRISMA guidelines and registered in PROSPERO (CRD420251107624). PubMed, Scopus, Web of Science, and Lens.org were searched for observational studies published from 2010 to 2025 reporting hyperuricemia among women with preeclampsia in LMIC hospital settings. Prevalence estimates were pooled using a random-effects model with logit transformation. Heterogeneity was assessed using Cochran’s Q, Tau², and I². Sensitivity analyses included subgrouping by hyperuricemia threshold and leave-one-out analysis. Results Eleven studies involving 1,099 women with preeclampsia from seven LMICs were included. The pooled prevalence of hyperuricemia was 53.47% (95% CI: 45.17% to 61.58%), with low-to-moderate heterogeneity (I² = 27.47%, p = 0.204). Included studies used different diagnostic thresholds for hyperuricemia, ranging from 5.0 to 7.0 mg/dL. All studies were hospital-based, and reporting of gestational age at the time of uric acid measurement was inconsistent. Conclusions Hyperuricemia was common among women with preeclampsia in hospital-based LMIC studies. These findings describe prevalence only and should not be interpreted as evidence of diagnostic accuracy, prognostic performance, or clinical utility. Future studies should standardize diagnostic thresholds and report gestational timing of uric acid measurement more consistently.

Gut bacterial metabolite imidazole propionate potentiates Alzheimer’s disease pathology

Abstract The gut microbiome modulates metabolic and neurovascular processes implicated in Alzheimer’s disease and related dementias (ADRD), but the underlying mechanisms remain unclear. Here, we identify the bacterial metabolite imidazole propionate (ImP) as a modifier of ADRD pathology. In a cohort of 1196 cognitively unimpaired adults, higher plasma ImP levels were associated with lower preclinical cognitive scores and biomarkers of ADRD, both cross-sectionally and longitudinally. Fecal metagenomic analysis linked putative ImP producers to ADRD phenotypes. Genome-wide integrative analysis revealed a locus on chromosome 12 associated with both plasma ImP levels and AD risk in humans, supporting a host genetic contribution to ImP regulation and a causal role of this metabolite in AD. In mice, chronic ImP administration exacerbated AD-like pathology. ImP impaired brain endothelial barrier and promoted tau hyperphosphorylation in primary neurons, an effect blocked by glycogen synthase kinase-3β inhibition. Together, this study links ImP to hallmarks of neurodegeneration and suggests that targeting ImP may represent a potential strategy to modify ADRD risk.

Investigating compensatory adjustments induced by rhythmic auditory stimulation for changes in temporal gait symmetry in lower-limb prosthetic users

Gait training is an important part of rehabilitation for prosthetic users, improving mobility and long-term health outcomes. This study investigated the effects of rhythmic auditory stimulation (RAS) on gait symmetry and overall movement patterns in lower-limb prosthesis users (LLPUs). We aimed to determine how RAS, designed to enhance temporal gait symmetry, influenced overall lower-limb kinematic patterns and whether such changes were beneficial or detrimental to gait. Analysis of discrete parameters and the Gait Profile Score (GPS) indicated that while RAS improved temporal symmetry, it did not significantly alter movement patterns for both transfemoral (TFA) and transtibial amputees (TTA). Subject-specific analyses revealed varied responses, with some LLPU participants showing improved gait symmetry without significant movement changes, while others exhibited compensatory adjustments. To further explore lower-limb kinematic changes associated with temporal symmetry, a perturbation-based task involving able-bodied (AB) individuals demonstrated the presence of unique kinematic changes, including reduced ankle plantarflexion and increased hip flexion, reflecting adjustments to synchronize with the asymmetric beat. Finally, subject-specific analyses revealed user experience as a key factor influencing the LLPU response. Participants with more years of prosthesis use demonstrated greater improvements in their overall gait pattern (GPS). This study highlights that RAS can improve gait symmetry in LLPUs without inducing other significant, compensatory changes in their movement patterns, a positive finding for clinical rehabilitation. However, our results show that individual responses can vary depending on factors such as prosthesis type and user experience and these should be considered in clinical care. Future research should explore a broader range of gait parameters and long-term effects of RAS to gain a comprehensive understanding of its impact on gait and rehabilitation.

Observation of localization reversal and harmonic generation in nonlinear non-Hermitian skin effect

Abstract The interplay between band topology and material nonlinearity gives rise to a variety of novel phenomena, such as topological solitons and nonlinearity-induced topological phase transitions. However, most previous studies fall within the Hermitian regime, leaving the impact of nonlinearity on non-Hermitian topology seldom explored. Here, we investigate the effects of nonlinearity on the non-Hermitian skin effect, a well-known non-Hermitian phenomenon induced by the point-gap topology unique to non-Hermitian systems. We discover a nonlinearity-induced point-gap topological phase transition accompanied by a reversal of the skin mode localization. This phenomenon is experimentally demonstrated in a nonlinear microwave metamaterial, where electromagnetic waves are localized around one end of the sample under a low-intensity pump, whereas at a high-intensity pump, the waves are localized around the other end. We also observe third harmonic generation signal induced by the skin modes, whose spatial distribution consistently shows the localization reversal. Furthermore, we extend our setup to build a nonlinear non-Hermitian Su-Schrieffer-Heeger model with both nonlinearity-induced point-gap and line-gap phase transitions. Our results open a new route towards nonlinear topological physics in non-Hermitian systems and are promising for reconfigurable topological wave manipulation and frequency conversion.

Correction: In vitro dissolution and release kinetics of multisource ciprofloxacin 500 mg tablets in West Gondar, Ethiopia: Implications for interchangeability and regulatory oversight

Spatially explicit temperature optima improve climate impact assessment of global crop productivity

Integrated single-cell RNA sequencing and Bulk-RNA technologies reveal the immunological characteristics of lactylation related-genes in glioblastoma

Objective Glioblastoma (GBM) is the most aggressive type of intracranial malignant tumor, known for its extremely poor prognosis. Lactylation, a newly identified post-translational modification, has been linked to tumorigenesis, though its specific role in GBM remains unclear. This study aims to integrate single-cell RNA sequencing (scRNA-seq) and bulk RNA sequencing (RNA-seq) data to create a novel prognostic model for GBM, focusing on lactylation-related factors. Methods We studied lactate metabolism genes as markers in GBM. We obtained bulk transcriptomic data from TCGA and the GSE141383 and GSE162631 cohorts in the GEO databases. We used the R package Seurat to analyze scRNA-seq data, CellChat for cell communication analysis, and AUCell to assess lactate metabolism gene set scores across cell types. We developed a prognostic model using machine learning algorithms and tested its efficacy across multiple cohorts. Additionally, we investigated differences in immune infiltration, predicted sensitivity, and other factors between high and low-risk groups. We validated the function of the key gene CD93 at the cellular level. Results The scRNA-seq data identified nine major cell types in GBM, with FCGBP+ macrophages showing the highest score in the lactate metabolism gene set. Authors designed a model informed by machine learning pinpointed three key genes（CD93, FCER1G, and GRB2）and developed a model with optimal prognostic value across cohorts.The high-risk group presented significantly poorer clinical outcomes. Immune-related bioinformatic analysis revealed significant differences in immune cell infiltration and checkpoint gene expression between risk groups. High-risk patients demonstrated lower immune infiltration and higher immunosuppression, rendering them less suitable for immunotherapy. Predictive algorithms indicated that axitinib and imatinib could be potential therapeutic drugs for these high-risk patients. In GBM tissue and cells, CD93 expression was significantly elevated, identifying it as a key risk gene in this model. Inhibition of CD93 expression via siRNA significantly reduced the proliferation, invasion, and migration of U87 and U251 cells. Conclusion In summary, we developed a novel characterization of lactylation-related clusters using single-cell sequencing technology. This study provided insights into the prognostic significance of lactate metabolism-related genes in GBM.

Advancing fair and explainable machine learning for neuroimaging dementia pattern classification in multi-racial and multi-ethnic populations

Abstract Dementia, a degenerative disease affecting millions globally, is projected to triple by 2050. Early and precise diagnosis is essential for effective treatment and improved quality of life. However, current diagnostic approaches often show inconsistent performance across multi-racial and multi-ethnic groups, raising concerns about fairness and clinical reliability. This study investigates performance discrepancies in dementia classification among 6584 Non-Hispanic White, 1263 Non-Hispanic African American, and 713 Hispanic White populations. We observed significant cross-group bias, particularly when models trained on one group are tested on another. To address this, we evaluated RegAlign, a few-shot domain adaptation objective that combines source-side focal learning, target-side class-weighted supervision, and class-conditional alignment to improve adaptation to underrepresented populations. Our results show that this approach substantially reduces inter-group performance gaps, especially between Non-Hispanic White and Hispanic populations. Here, we show the importance of fairness-aware learning strategies and diverse training data for improving the accuracy and equity of MRI-based dementia classification.