Objective: To develop a dual-branch deep learning framework for accurate multi-label classification of fundus diseases, addressing the key limitations of insufficient complementary feature extraction and inadequate cross-modal feature fusion in existing automated diagnostic methods. Methods: The fundus multi-label classification dataset with 12 disease categories (FMLC-12) dataset was constructed by integrating complementary samples from Ocular Disease Intelligent Recognition (ODIR) and Retinal Fundus Multi-Disease Image Dataset (RFMiD), yielding 6 936 fundus images across 12 retinal pathology categories, and the framework was validated on both FMLC-12 and ODIR. Inspired by the holistic multi-regional assessment principle of the Five Wheels theory in traditional Chinese medicine (TCM) ophthalmology, the dual-branch multi-label network (DBMNet) was developed as a novel framework integrating complementary visual feature extraction with pathological correlation modeling. The architecture employed a TransNeXt backbone within a dual-branch design: one branch processed red-green-blue (RGB) images to capture color-dependent features, such as vascular patterns and lesion morphology, while the other processed grayscale-converted images to enhance subtle textural details and contrast variations. A feature interaction module (FIM) effectively integrated the multi-scale features from both branches. Comprehensive ablation studies were conducted to evaluate the contributions of the dual-branch architecture and the FIM. The performance of DBMNet was compared against four state-of-the-art methods, including EfficientNet Ensemble, transfer learning-based convolutional neural network (CNN), BFENet, and EyeDeep-Net, using mean average precision (mAP), F1-score, and Cohen's kappa coefficient. Results: The dual-branch architecture improved mAP by 15.44 percentage points over the single-branch TransNeXt baseline, increasing from 34.41% to 44.24%, and the addition of FIM further boosted mAP to 49.85%. On FMLC-12, DBMNet achieved an mAP of 49.85%, a Cohen’s kappa coefficient of 62.14%, and an F1-score of 70.21%. Compared with BFENet (mAP: 45.42%, kappa: 46.64%, F1-score: 71.34%), DBMNet outperformed it by 4.43 percentage points in mAP and 15.50 percentage points in kappa, while BFENet achieved a marginally higher F1-score. On ODIR, DBMNet achieved an F1-score of 85.50%, comparable to state-of-the-art methods. Conclusion DBMNet effectively integrates RGB and grayscale visual modalities through a dual-branch architecture, significantly improving multi-label fundus disease classification. The framework not only addresses the issue of insufficient feature fusion in existing methods but also demonstrates outstanding performance in balancing detection across both common and rare diseases, providing a promising and clinically applicable pathway for standardized, intelligent fundus disease classification.

ObjectiveThe essence of syndrome manifestation recognition in traditional Chinese medicine (TCM) is to infer the body’s latent pathogenesis state from clinical observational information, rather than to perform simple label matching. However, previous studies have largely modeled this task as syndrome pattern classification within a fixed label space, which does not adequately reflect the cognition process of TCM syndrome differentiation centered on pathogenesis reasoning, and is also insufficient to capture the openness, semantic variability, and cross-disease reusability of syndrome manifestation expression. This study aimed to investigate whether introducing pathogenesis reasoning chain-of-thought (PR-CoT) supervision into large language models (LLMs) could improve the quality and cognitive consistency of syndrome manifestation recognition and support cross-disease transfer. MethodsSyndrome manifestation recognition was formulated as a conditional generation task under the framework of clinical observational information (X)→pathogenesis structure (Z)→syndrome pattern output (Y), where Z serves as an explicit intermediate structural variable linking the clinical evidence and syndrome judgment. Within this framework, a PR-CoT-supervised dataset for syndrome manifestation recognition was constructed based on medical case records of spleen-stomach disorders. After preprocessing, information extraction, manual proofreading, and data cleaning, the dataset comprised 4 800 training cases, 400 development cases, and 400 test cases. Each sample was annotated with a structured PR-CoT consisting of three progressive levels: clinical information summarization, comprehensive pathogenesis analysis, and syndrome pattern output. Supervised fine-tuning was conducted on open-source LLMs, with an end-to-end model serving as the baseline. Qwen3-32B was used as the primary experimental model, and Qwen3-14B as the scale comparison model. A progressive multidimensional evaluation framework was further established, comprising a structural parsing level, a semantic similarity level, and an expert blind review level. At the structural parsing level, syndrome pattern expressions were decomposed into structural elements and evaluated using Precision, Recall, F1 score, and Jaccard similarity. At the semantic similarity level, independent LLMs scored the theoretical proximity between predicted and reference syndrome patterns. At the expert blind review level, three TCM experts independently evaluated model outputs on two dimensions: syndrome differentiation consistency and terminology standardization of syndrome patterns. In addition, zero-shot cross-disease transfer evaluation was conducted on gynecological and heart-system disorder test sets. ResultsAt the structural parsing level, PR-CoT supervision did not lead to a stable improvement in the element-wise overlap of syndrome pattern structural components. Compared with the corresponding baselines, neither Qwen3-32B nor Qwen3-14B showed consistent advantages in structural matching metrics after the introduction of PR-CoT supervision. In contrast, at the semantic similarity level, PR-CoT supervision produced stable positive gains across different model scales and evaluation systems. The average semantic score of Qwen3-32B increased from 6.425 8 in the baseline model to 6.585 0 after PR-CoT supervision, and that of Qwen3-14B increased from 5.870 0 to 5.964 2. At the expert blind review level, the overall score of Qwen3-32B (PR-CoT) was 7.026 0±0.107 7, higher than 6.416 3±0.288 9 for its baseline. In zero-shot cross-disease testing, the PR-CoT model still showed advantages in semantic evaluation and expert evaluation on both gynecological and heart-system disorder test sets, indicating a certain degree of transferability. ConclusionThe benefits of PR-CoT supervision are mainly reflected in TCM semantic consistency and clinical plausibility, rather than in improved hard matching of structural elements. These findings support understanding syndrome manifestation recognition as a process of generating and expressing latent pathogenesis structures, rather than as a classification task within a traditional fixed label space. By introducing pathogenesis reasoning as an explicit intermediate structure into the modeling process and combining it with a progressive multidimensional evaluation framework, this study provides a methodological pathway for intelligent TCM syndrome differentiation that integrates theoretical alignment, interpretability, and multi-level evaluation.

ObjectiveThe essence of syndrome manifestation recognition in traditional Chinese medicine (TCM) is to infer the body’s latent pathogenesis state from clinical observational information, rather than to perform simple label matching. However, previous studies have largely modeled this task as syndrome pattern classification within a fixed label space, which does not adequately reflect the cognition process of TCM syndrome differentiation centered on pathogenesis reasoning, and is also insufficient to capture the openness, semantic variability, and cross-disease reusability of syndrome manifestation expression. This study aimed to investigate whether introducing pathogenesis reasoning chain-of-thought (PR-CoT) supervision into large language models (LLMs) could improve the quality and cognitive consistency of syndrome manifestation recognition and support cross-disease transfer. MethodsSyndrome manifestation recognition was formulated as a conditional generation task under the framework of clinical observational information (X)→pathogenesis structure (Z)→syndrome pattern output (Y), where Z serves as an explicit intermediate structural variable linking the clinical evidence and syndrome judgment. Within this framework, a PR-CoT-supervised dataset for syndrome manifestation recognition was constructed based on medical case records of spleen-stomach disorders. After preprocessing, information extraction, manual proofreading, and data cleaning, the dataset comprised 4 800 training cases, 400 development cases, and 400 test cases. Each sample was annotated with a structured PR-CoT consisting of three progressive levels: clinical information summarization, comprehensive pathogenesis analysis, and syndrome pattern output. Supervised fine-tuning was conducted on open-source LLMs, with an end-to-end model serving as the baseline. Qwen3-32B was used as the primary experimental model, and Qwen3-14B as the scale comparison model. A progressive multidimensional evaluation framework was further established, comprising a structural parsing level, a semantic similarity level, and an expert blind review level. At the structural parsing level, syndrome pattern expressions were decomposed into structural elements and evaluated using Precision, Recall, F1 score, and Jaccard similarity. At the semantic similarity level, independent LLMs scored the theoretical proximity between predicted and reference syndrome patterns. At the expert blind review level, three TCM experts independently evaluated model outputs on two dimensions: syndrome differentiation consistency and terminology standardization of syndrome patterns. In addition, zero-shot cross-disease transfer evaluation was conducted on gynecological and heart-system disorder test sets. ResultsAt the structural parsing level, PR-CoT supervision did not lead to a stable improvement in the element-wise overlap of syndrome pattern structural components. Compared with the corresponding baselines, neither Qwen3-32B nor Qwen3-14B showed consistent advantages in structural matching metrics after the introduction of PR-CoT supervision. In contrast, at the semantic similarity level, PR-CoT supervision produced stable positive gains across different model scales and evaluation systems. The average semantic score of Qwen3-32B increased from 6.425 8 in the baseline model to 6.585 0 after PR-CoT supervision, and that of Qwen3-14B increased from 5.870 0 to 5.964 2. At the expert blind review level, the overall score of Qwen3-32B (PR-CoT) was 7.026 0±0.107 7, higher than 6.416 3±0.288 9 for its baseline. In zero-shot cross-disease testing, the PR-CoT model still showed advantages in semantic evaluation and expert evaluation on both gynecological and heart-system disorder test sets, indicating a certain degree of transferability. ConclusionThe benefits of PR-CoT supervision are mainly reflected in TCM semantic consistency and clinical plausibility, rather than in improved hard matching of structural elements. These findings support understanding syndrome manifestation recognition as a process of generating and expressing latent pathogenesis structures, rather than as a classification task within a traditional fixed label space. By introducing pathogenesis reasoning as an explicit intermediate structure into the modeling process and combining it with a progressive multidimensional evaluation framework, this study provides a methodological pathway for intelligent TCM syndrome differentiation that integrates theoretical alignment, interpretability, and multi-level evaluation.

Introduction: The 2021 IDF-DAR risk calculator had been previously evaluated in multiple studies and subsequently widely accepted and applied in clinical practice as a practical standardized tool for patient risk stratification. Recently updated, the 2026 IDF-DAR Risk calculator enables a more individualized, evidence-related evaluation of patientrelated and disease-related risk factors, incorporating modern diabetes technologies, including continuous glucose monitoring (CGM), automated insulin delivery (AID) systems, and advanced insulin formulations to enhance risk stratification. This tool allows medical professionals to tailor Ramadan practices based on overall factors toward promoting safe fasting. Methodology: This prospective multicentre observational study recruited adults with Type 1 and Type 2 diabetes attending public hospitals nationwide. People with diabetes (PwD) intending to perform Ramadan fasting were invited to participate and assessed using the 2026 IDF-DAR Risk Calculator in the 6-week pre-Ramadan period between 30th January and 19th March 2026. Results: A total of 458 PwD were evaluated and stratified into low (15.7%), moderate (41%), and high risk (43.3%) categories. Most participants had Type 2 diabetes (83.6%), with 60.3% having a disease duration exceeding 10 years and 43% exhibiting poor glycemic control (hemoglobin A1c >9%). Insulin therapy was used by 76.4% of participants, including two individuals with Type 1 diabetes using AID systems. Most participants reported no recent hypoglycemia (76.4%), 81.0% performed glucose monitoring, and 3.3% used CGM. Severe comorbidities were uncommon, with 1.1% having unstable macrovascular disease and 4.4% advanced chronic kidney disease (estimated glomerular filtration rate <30). Notably, 72.2% received structured Ramadan education. Conclusion Majority of PwD attending tertiary diabetes clinics were in the moderate- to high-risk category and intended to fast despite medical advice against fasting in some cases. Although most participants were on insulin therapy, hypoglycemia was low in the pre-Ramadan period. Integration of modern technologies, advanced insulin therapies, and structured education may support safer fasting practices.
Risk Assessment ; Diabetes Mellitus ; Hospitals ; Fasting

Objective To investigate the epidemiological characteristics and changing trend of typhus in Shaanxi Province from 2005 to 2023 to provide a scientific basis for its prevention and control. Methods Excel 2007, SPSS 25.0, Joinpoint 4.9.1.0, and Geoda 1.6 were used for data collection and statistical analysis. Super Map was used for data visualization to describe the changing characteristics of the disease. Results A total of 394 typhus cases were reported in Shaanxi Province from 2005 to 2023. The average annual incidence of typhus was 0.054/100 000, showing a dynamic fluctuation trend(AAPC=-3.3, t=-0.3, P>0.05). The cases were mainly concentrated in Baoji, Hanzhong and Xi′an, accounting for 78.68%. The incidence peak was from May to October, accounting for 64.21% of annual incidence. The epidemic season was from May to October and December. The incidence of the disease was concentrated in the 40-69 age group, accounting for 58.88%, and the sex was 1.07:1. The main occupation was farmers, accounting for 72.08%. The median time from onset to diagnosis was 7 days. Global spatial autocorrelation analysis showed that there were significant spatial autocorrelations in 11 years from 2005 to 2023(P<0.05). Local spatial autocorrelation analysis detected a total of 43“high-high”clustering areas, mainly concentrated in Baoji City. Conclusions The overall incidence of typhus in Shaanxi Province showed a dynamic fluctuation trend, with notable seasonal and regional aggregation. The incidence of typhus was higher in middle-aged and elderly people in rural areas. Surveillance should be strengthened in typhus endemic areas in summer and autumn, and health education should be conducted for key population to form good health habits and reduce the incidence of typhus.

Panretinal photocoagulation(PRP)is the first-line treatment for severe non-proliferative diabetic retinopathy and proliferative diabetic retinopathy(PDR).Its therapeutic efficacy varies due to multiple patient-specific systemic and ocular factors. Despite standardized PRP, some patients fail to show adequate regression of neovascularization and subsequently develop vitreous hemorrhage(VH). In such cases, pars plana vitrectomy(PPV)effectively clears the hemorrhage, restores transparency of the ocular media, supplement PRP, and delays the progression of PDR, thus becoming a critical intervention. With recent advances in surgical techniques, concepts, and equipment, early PPV intervention for VH following PRP has emerged as a growing research focus. Nevertheless, consensus remains lacking regarding its indications, optimal timing, and surgical protocols. This article systematically reviews relevant domestic and international studies, summarizing the surgical timing, approaches, and prognoses of PPV for VH in PDR patients after PRP, aiming to provide evidence-based reference for clinical decision-making and reduce the risk of PDR-related blindness.

Autism spectrum disorder (ASD) is a neurodevelopmental condition with a steadily rising global prevalence, yet effective pharmacological interventions remain notably limited, highlighting an urgent need for safe, accessible, and mechanism-based therapeutic strategies. Physical exercise has emerged as a promising non-pharmacological intervention that ameliorates both core symptoms—social communication deficits and restricted repetitive behaviors—and associated features including cognitive dysfunction and motor impairments, in children and adolescents with ASD. However, the molecular mechanisms mediating these beneficial effects remain incompletely defined, impeding the development of evidence-based exercise prescriptions and biomarker-driven rehabilitation protocols. MicroRNAs (miRNAs) are evolutionarily conserved small non-coding RNAs that post-transcriptionally regulate approximately 60% of protein-coding genes. Within the central nervous system, miRNAs orchestrate diverse neurobiological processes including neural progenitor proliferation, neuronal differentiation, dendritic spine morphogenesis, synaptic plasticity, and neuroinflammatory homeostasis. Notably, miRNAs are remarkably stable in biological fluids and can be packaged into extracellular vesicles, rendering them attractive candidates as both mechanistic mediators and non-invasive peripheral biomarkers. Among the hundreds of miRNAs expressed in the brain, three—miR-132, miR-34a, and miR-146a—have emerged as particularly relevant to ASD pathophysiology. This review focuses on these three miRNAs for the following reasons: miR-132 is a master regulator of activity-dependent synaptic plasticity through its modulation of BDNF/MeCP2/PTEN signaling and has been consistently downregulated in ASD prefrontal cortex; miR-34a functions as a pro-apoptotic factor that suppresses Bcl-2-mediated neuronal survival pathways and is upregulated in ASD cerebellum; and miR-146a serves as a key brake on neuroinflammation via TLR7/IRAK1 signaling and shows region-specific dysregulation in ASD temporal lobe. We first summarize evidence from human post-mortem brain tissues and ASD animal models demonstrating the consistent dysregulation of these three miRNAs. Notably, the pathological consequences of these miRNA alterations—impaired synaptic plasticity, excessive neuronal apoptosis, and sustained neuroinflammation—are interconnected and collectively contribute to the heterogeneous symptomatology of ASD. We then present a synthesis of emerging evidence demonstrating that various exercise modalities, including swimming, treadmill running, and voluntary wheel running, can concurrently reverse these ASD-like behavioral phenotypes and normalize the expression of the three key miRNAs. These data provide the first direct experimental evidence linking exercise-induced miRNA modulation to ASD symptom improvement. On the basis of these findings, we propose an integrative “exercise-miRNA-ASD” framework wherein exercise functions as a multi-targeted modulator—simultaneously enhancing synaptic plasticity, promoting neuronal survival, and attenuating neuroinflammation—through coordinated regulation of the three miRNAs. Importantly, this framework is not merely descriptive but offers testable predictions: exercise-induced miRNA changes should be dose-dependent, show temporal correlation with behavioral improvements, and be blunted by miRNA-specific antagonists or CRISPR/Cas9-mediated knockout. Beyond its specific application to ASD, this framework has broader implications. The miR-132/BDNF, miR-34a/Bcl-2, and miR-146a/TLR7 pathways are not ASD-specific but represent fundamental neural stress and repair mechanisms that are dysregulated across Alzheimer’s disease (AD), traumatic brain injury, Parkinson’s disease (PD), and major depressive disorder. Exercise has been shown to modulate these same miRNAs in several of these conditions, suggesting that the “exercise-miRNA-neural function” axis may represent a conserved neuroprotective mechanism that transcends diagnostic boundaries. Thus, we propose that ASD serves as an ideal model disease for elucidating this universal mechanism, with findings potentially generalizable to other neurological disorders. We also critically evaluate current translational barriers: the near-absence of human clinical trials with serial miRNA profiling; the undefined dose-response relationships between exercise parameters and miRNA expression; the unresolved causality issue (current evidence demonstrates association, not causation); and the uncertain correlation between peripheral exosomal miRNA levels and brain miRNA dynamics. We argue that future research must prioritize CRISPR/Cas9-based miRNA manipulation in animal models combined with longitudinal exercise interventions to establish causality, and that cross-disease validation studies are essential to determine whether exercise-induced miRNA changes represent a shared neuroprotective signature or disease-specific responses. Ultimately, we envision a paradigm where a simple blood test measuring exosomal miR-132, miR-34a, and miR-146a levels could guide personalized exercise prescriptions, enabling precision rehabilitation for individuals with ASD and potentially other neurological conditions.

In this paper, near-infrared spectroscopy(NIRS) was employed to analyze 129 batches of commercial products of Reduning Injection. The batch reporting rate was estimated according to the report of Reduning Injection in the direct adverse drug reaction(ADR) reporting system of the drug marketing authorization holder of the Center for Drug Reevaluation of the National Medical Products Administration(National Center for ADR Monitoring) from August 2021 to August 2022. According to the batch reporting rate, the samples of Reduning Injection were classified into those with potential risks and those being safe. No processing, random oversampling(ROS), random undersampling(RUS), and synthetic minority over-sampling technique(SMOTE) were then employed to balance the unbalanced data. After the samples were classified according to appropriate sampling methods, competitive adaptive reweighted sampling(CARS), successive projections algorithm(SPA), uninformative variables elimination(UVE), and genetic algorithm(GA) were respectively adopted to screen the features of spectral data. Then, support vector machine(SVM), logistic regression(LR), k-nearest neighbors(KNN), naive bayes(NB), random forest(RF), and artificial neural network(ANN) were adopted to establish the risk prediction models. The effects of the four feature extraction methods on the accuracy of the models were compared. The optimal method was selected, and bayesian optimization was performned to optimize the model parameters to improve the accuracy and robustness of model prediction. To explore the correlations between potential risks of clinical use and quality test data, TreeNet was employed to identify potential quality parameters affecting the clinical safety of Reduning Injection. The results showed that the models established with the SVM, LR, KNN, NB, RF, and ANN algorithms had the F1 scores of 0.85, 0.85, 0.86, 0.80, 0.88, and 0.85 and the accuracy of 88%, 88%, 88%, 85%, 91%, and 88%, respectively, and the prediction time was less than 5 s. The results indicated that the established models were accurate and efficient. Therefore, near infrared spectroscopy combined with machine learning algorithms can quickly predict the potential risks of clinical use of Reduning Injection in batches. Three key quality parameters that may affect clinical safety were identified by TreeNet, which provided a scientific basis for improving the safety standards of Reduning Injection.
Spectroscopy, Near-Infrared/methods* ; Drugs, Chinese Herbal/administration & dosage* ; Machine Learning ; Algorithms ; Humans ; Quality Control

This study aims to explore the effects of Buzhong Yiqi Decoction(BYD) on the cyclic guanosine monophosphate-adenosine monophosphate synthase(cGAS)-stimulator of interferon genes(STING) signaling pathway in the mouse model of autoimmune thyroiditis(AIT) and the mechanism of BYD in alleviating the immune injury. Forty-eight NOD.H-2~(h4) mice were assigned into normal, model, low-, medium-, and high-dose BYD, and selenium yeast tablets groups(n=8). Mice of 8 weeks old were treated with 0.05% sodium iodide solution for 8 weeks for the modeling of AIT and then administrated with corresponding drugs by gavage for 8 weeks before sampling. High performance liquid chromatography was employed to measure the astragaloside Ⅳ content in BYD. Hematoxylin-eosin staining was employed to observe the pathological changes in the mouse thyroid tissue. Enzyme-linked immunosorbent assay was employed to measure the serum levels of thyroid peroxidase antibody(TPO-Ab), thyroglobulin antibody(TgAb), and interferon-γ(IFN-γ). Flow cytometry was employed to detect the distribution of T cell subsets in the spleen. The immunohistochemical method was used to detect the expression of cGAS, STING, TANK-binding kinase 1(TBK1), and interferon regulatory factor 3(IRF3). Real-time PCR and Western blot were employed to determine the mRNA and protein levels, respectively, of markers related to the cGAS-STING signaling pathway in the thyroid tissue. The results showed that the content of astragaloside Ⅳ in BYD was(7.06±0.08) mg·mL~(-1). Compared with the normal group, the model group showed disrupted structures of thyroid follicular epithelial cells, massive infiltration of lymphocytes, and elevated levels of TgAb and TPO-Ab. Compared with the model group, the four treatment groups showed intact epithelial cells, reduced lymphocyte infiltration, and lowered levels of TgAb and TPO-Ab. Compared with the normal group, the model group showed increases in the proportions of Th1 and Th17 cells, a decrease in the proportion of Th2 cells, and an increase in the IFN-γ level. Compared with the model group, the four treatment groups presented decreased proportions of Th1 and Th17 cells and lowered levels of IFN-γ, and the medium-dose BYD group showed an increase in the proportion of Th2 cells. Compared with the normal group, the modeling up-regulated the mRNA levels of cGAS, STING, TBK1, and IRF3 and the protein levels of cGAS, p-STING, p-TBK1, and p-IRF3. Compared with the model group, the four treatment groups showed reduced levels of cGAS, STING, TBK1, and IRF3-positive products, down-regulated mRNA levels of cGAS, STING, and TBK1, and down-regulated protein levels of cGAS and p-STING. The high-dose BYD group showed down-regulations in the mRNA level of IRF3 and the protein levels of p-TBK1 and p-IRF3. The above results indicate that BYD can repair the imbalance of T cell subsets, alleviate immune injury, and reduce thyroid lymphocyte infiltration in AIT mice by inhibiting the cGAS-STING signaling pathway.
Animals ; Drugs, Chinese Herbal/administration & dosage* ; Signal Transduction/drug effects* ; Thyroiditis, Autoimmune/metabolism* ; Mice ; Membrane Proteins/metabolism* ; Mice, Inbred NOD ; Humans ; Female ; Nucleotidyltransferases/metabolism* ; Male ; Disease Models, Animal