Shenzhen Sixth People's Hospital

Lengthy acquisition time remains a key bottleneck for the widespread use of MRI in clinics. While accelerated MRI can reduce scan duration, it often introduces increased noise, compromising image quality and diagnostic reliability. In this study, we present a unified deep learning-based denoising model for multi-organ accelerated MRI, designed to operate directly on reconstructed images from commercial MRI systems. Our model was trained on a prospectively collected, large-scale real-world dataset comprising 148,930 noisy-clean image pairs from six clinical centers and four major MRI vendors, spanning six organs and 96 MRI protocols. On a test set of 20,143 real-world image pairs, our model consistently outperforms state-of-the-art denoising methods. Importantly, downstream evaluation using tissue segmentation demonstrates a 7.05% improvement in Dice score across multiple organs compared to noisy images. The model further generalizes effectively to 46,870 external clinical images from four independent cohorts, highlighting its robustness across various scanners and acquisition protocols. To assess clinical utility, two experienced radiologists conducted blinded evaluations across multiple organs, focusing on overall image quality, diagnostic confidence, and disease diagnosis. The denoised images retained high visual fidelity and yielded diagnostic performance equivalent to clean images even with acceleration factor of 3× compared to clinical scanning setup, such that many acquisitions can be completed within one minute. This unified MRI denoising model holds great potential for various clinical applications.

Post-stroke depression (PSD) is a common and serious complication following stroke, affecting approximately one-third of survivors and contributing to poor functional recovery. Its pathophysiology is multifactorial, with recent evidence highlighting decreased hippocampal neurogenesis as a key mechanistic contributor. This review aims to examine the specific role of hippocampal neurogenesis in PSD, focusing on (1) the evidence linking impaired neurogenesis to PSD onset (2), the underlying molecular and cellular mechanisms—including dysregulation of the hypothalamic–pituitary–adrenal(HPA) axis, neuroinflammation, altered neurotrophic signaling, and neurotransmitter disturbances—and (3) current and emerging therapeutic strategies that promote neurogenesis for PSD management. Pharmacological agents (e.g., antidepressants), neuroregulatory interventions, and lifestyle-based approaches show promise in restoring neurogenic activity and alleviating depressive symptoms. In conclusion, impaired hippocampal neurogenesis represents a central pathway in PSD pathogenesis, offering a valuable target for future research and therapeutic development. Further studies are needed to fully elucidate these mechanisms and translate neurogenesis-focused treatments into clinical practice.

The endogenous cannabinoid system, comprising endocannabinoids, cannabinoid receptors, and enzymes responsible for synthesis and degradation of endocannabinoids, is integral to the regulation of skin inflammation. However, the precise contributions of cannabinoid-metabolizing enzymes including fatty acid amide hydrolase and monoacylglycerol lipase (MAGL) remain poorly defined in psoriatic pathogenesis. In this study, we demonstrated the significant MAGL upregulation in psoriatic lesions and established that pharmacological inhibition of MAGL markedly reduced psoriasiform inflammation and pruritus in imiquimod-challenged C57BL/6 mice. Mechanically, inhibition of MAGL decreased the expression of 12-lipoxygenase (ALOX12), subsequently reducing the synthesis of 12-hydroxyeicosatetraenoic acid (12-HETE), which is the principal metabolite of ALOX12. Importantly, these effects were reversed by cannabinoid receptor 2 (CB2R) antagonism or knockout, establishing a signaling axis in which MAGL inhibition activates CB2R to suppress ALOX12. Collectively, our findings identified MAGL as a potentially targetable nexus in psoriasiform inflammation and itch.

This study aimed to characterize pediatric nasal button battery impactions and evaluate, in an in vitro septum model, the efficacy of household reagents in mitigating alkaline-induced injury. We retrospectively reviewed 230 pediatric cases over ten years. Demographic characteristics, presentation times, and complications were recorded. In vitro, sheep septal cartilage segments (2-3mm thick) were immersed in saline at 36°C and exposed to a 3V button battery for 15, 30, 45, or 60min to determine the critical injury window. In the 30min group, interventions with honey, MOH (a honey-olive oil mixture), vinegar, vitamin D, glycerin, and saline were applied, and the tissue pH, residual battery voltage, and injury depth were measured. A retrospective analysis of 230 children revealed that 61.7% were male, 97.8% were aged 1-5years, and 87.4% presented within 24h of battery impaction; the incidence of nasal septal perforation was 16.5%. In vitro experiments identified 30min as the critical window for rapid tissue injury progression, with a mean injury depth of 2160 ± 43.8μm. Intervention assays demonstrated that honey (685.3 ± 92.2μm), MOH (1038 ± 145.1μm), and glycerin (1275 ± 39.6μm) significantly reduced injury depth without breaching the perichondrial layer. Honey and MOH best neutralized the pH, whereas glycerin and vitamin D inhibited battery discharge the most. Saline, vinegar, or vitamin D alone had no protective effect on depth. Alkaline burns from nasal button batteries progress rapidly within 30min. Household reagents-especially honey, MOH, or glycerin-neutralize pH, inhibit discharge, and attenuate tissue damage. In prehospital settings where immediate care is delayed, early application of honey, MOH, or glycerin is recommended to minimize severe complications.

This study aimed to identify metabolic signatures reflecting air pollution and further explore their associations with gout risk. We retrieved 207,908 gout-free participants from a large cohort study. Annual average concentrations of fine particulate matter (PM2.5), inhalable particulate matter (PM10), nitrogen dioxide (NO2), and nitrogen oxides (NOx) were estimated using bilinear interpolation based on the residential address. Elastic net regression was used to identify air pollutant-related metabolites and construct metabolic signatures. Cox regression models were performed to evaluate the association between air pollutants, related metabolic signatures, and the risk of gout. Mediation analyses were performed to explore how metabolites mediate the relationships linking air pollutants to gout. During a median follow-up of 12.56 years, 2,474 incident gout cases were identified. We identified metabolite biomarkers reflecting air pollutants, including 87 metabolites for PM2.5, 76 for PM10, 68 for NO2, and 71 for NOx. Air pollution-related metabolic signatures were associated with elevated gout risk, with hazard ratios of 1.10 (95% confidence interval [CI] 1.05-1.14) for PM2.5, 1.12 (95% CI 1.08-1.17) for PM10, 1.09 (95% CI 1.04-1.13) for NO2, and 1.11 (95% CI 1.06-1.15) for NOx. Consistently, per-interquartile rangeincrease in PM2.5, PM10, NO2, and NOx was associated with 17%, 15%, 6%, and 6% increased gout risk, respectively. Additionally, metabolic signatures mediated the air pollutants-gout associations, with the mediation proportions varying from 2.36% for PM2.5 to 4.30% for NOx. Our study indicated that air pollutants and metabolic signatures were associated with elevated gout risk, with metabolomic signatures playing a mediating role in the air pollutants-gout relationship.

Abstract Objective Despite the central role of deep brain structures such as the striatum in motor learning, existing noninvasive stimulation methods are hindered by limited depth and precision. Temporal interference (TI) stimulation presents the potential for precise, individualized modulation of deep regions. However, how TI stimulation influences deep brain activity and large-scale network reorganization to facilitate motor learning is still unclear. Therefore, this study aimed to clarify these mechanisms by investigating how personalized, high-intensity TI targeting the striatum modulates neural activity and enhances motor learning. Methods Twenty-six healthy right-handed male participants were enrolled in a randomized, double-blind, sham-controlled crossover study. Each participant received both TI and sham stimulation targeting the right striatum (10 mA, Δf = 20 Hz) through an individualized electrode montage. Resting-state functional magnetic resonance imaging (fMRI), diffusion tensor imaging (DTI), and serial reaction time (SRT) task performance were assessed before and after each intervention. Neural analyses included static and dynamic fractional amplitude of low-frequency fluctuation (fALFF) in the target region, structural connectivity (SC)–functional connectivity (FC) coupling and topological metrics across six major brain networks, as well as brain-behavior correlations related to learning performance. Results (1) Target region activity: TI stimulation significantly increased both static and dynamic fALFF in the right striatum (Cohen’s d = 0.49, p = 0.016; Cohen’s d = 0.39, p = 0.035). (2) Brain network reorganization: Compared with the Sham group, the TI group exhibited significantly reduced SC-FC coupling in the cerebellar network (CN) (t = – 2.279, Cohen’s d = -0.82, p = 0.027, FDR corrected). TI stimulation significantly enhanced nodal degree in the cerebellar network (CN) and nodal efficiency the cingulo-opercular network (CON) (FDR corrected with Bonferroni correction, p < 0.0083) following the stimulation. Behavioral performance: The TI stimulation significantly improved performance in second implicit learning (SIL) after stimulation (Cohen’s d = 0.47, p = 0.043). Brain-behavior correlation: Changes in SC-FC coupling in the CN were significantly negatively correlated with improvement in SIL (r = – 0.372, p = 0.040). Conclusion Personalized high-intensity TI of the striatum enhances deep target activity and promotes selective network reorganization, particularly by reducing SC-FC coupling and strengthening intra-network connectivity in the CN. These network-level modulations underlie improved implicit learning performance, highlighting the potential of TI neuromodulation as a precise and effective approach for promoting motor learning by targeting deep nuclei and large-scale brain networks. Trial registration number ChiCTR2500098699.

BackgroundT lymphocyte dysfunction is closely associated with immunosuppression in sepsis, whereas the underlying mechanisms are not fully understood.ResultsIn this study, we established a mouse model of cecal ligation and puncture (CLP)-induced sepsis and observed immunometabolic alterations in splenic T cells. Serum energy metabolites related to glycolysis and the tricarboxylic acid (TCA) cycle were imbalanced. Splenic T cells from septic mice showed a shift in subset distribution, with decreased naïve T cells and increased effector populations, along with concurrent activation and exhaustion phenotypes. Notably, mitochondrial mass and mitochondrial membrane potential were significantly diminished in both CD4+ and CD8+ T cell, correlated with increased programmed cell death protein 1 (PD-1) expression. Transmission electron microscopy further confirmed mitochondrial morphological alterations in CLP-derived CD3+ T cells. Furthermore, seahorse assays demonstrated impaired metabolic reprogramming capacity in activated CLP splenic CD3+ T cells, with suppressed glycolytic and oxidative phosphorylation responses. This impairment was coupled with reduced fold-increases in mitochondrial mass and mitochondrial membrane potential levels upon activation in both CD4+ and CD8+ T cell compared to controls. Clinically, peripheral T cells from septic patients showed elevated CD69 and PD-1 expression, a significant increase in CD39 and a decrease in CD73, increased mitochondrial mass and decreased mitochondrial membrane potential, particularly in those with septic shock.ConclusionsOur findings provide several layers of T cell dysfunction in sepsis, linking subset redistribution, an exhausted phenotype, mitochondrial impairment, and reduced proliferative capacity, suggesting that future therapeutic interventions aiming to reverse sepsis-induced immunosuppression may require a combinatorial approach.

Early diagnosis of breast cancer (BC) remains pivotal for enhancing patient outcomes, yet current imaging modalities face inherent limitations. This retrospective study, encompassing 1366 BC patients and 1186 healthy controls (HC) from Shenzhen Nanshan People’s Hospital (2022–2025), developed and validated a serum tumor marker-based predictive model to address this gap. Eight markers—CEA, AFP, CA199, CA242, CA15-3, CA125, FER, and NSE—were analyzed, with multivariate logistic regression identifying AFP, CA242, CA15-3, and NSE as independent diagnostic predictors. The resulting four-marker panel demonstrated robust performance, achieving an AUC of 0.90 (95% CI: 0.89–0.92) with 77.3% sensitivity and 89.7% specificity in the development cohort (n = 1535), and sustained efficacy in validation (AUC = 0.82, 87.5% sensitivity, 72.8% specificity; n = 761). The model effectively stratified early-stage (T1/T2: 86.49% accuracy) versus advanced-stage (T3/T4: 92.00%) tumors, lymph node involvement (N0: 86.49%; N+: 88.00%), metastatic status (M0: 87.32%; M+: 85.71%), and molecular subtypes (Luminal A: 91.33%; Luminal B: 82.35%; HER2+: 84.61%; triple-negative: 87.75%). Notably, longitudinal risk score trends correlated with therapeutic response, declining in remission and rising with progression. These findings collectively highlight the model’s dual utility as a high-accuracy diagnostic tool for early BC detection and a dynamic biomarker for monitoring treatment efficacy. This novel model may provide an auxiliary approach in current screening paradigms, underscoring its transformative potential in oncology practice.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-025-30087-1.