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.

Objective To develop a personalized recommendation strategy for tranexamic acid administration during the perioperative period of orthopedic surgery based on machine learning,aiming to reduce perioperative bleeding and related complications and improving clinical outcomes.Methods A total of 11 727 patients undergoing orthopedic surgery from the INSPIRE database were subjected in this study.Missing data were handled using multiple imputation methods,and relevant feature variables were screened using Boruta analysis.We constructed various machine learning models,including Gradient Boosting Machine(GBM),Generalized Linear Model(GLM),eXtreme Gradient Boosting(XGBoost),K-Nearest Neighbors(KNN),Neural Network(NNET),Naive Bayes(NB),and Random Forest(RF),to evaluate their performance in predicting intraoperative bleeding and prolonged postoperative length of hospital stay.The optimal model was then selected and further integrated using a weighted ensemble,aiming to achieve the best prognosis by recommending usage strategies for tranexamic acid.The predictive performance of the constructed model was then verified against the testing set,and compared with the physician decision-making to complete the evaluation.Results In predicting intraoperative bleeding,the RF model achieved an area under the receiver operating characteristic curve(AUC)of 0.73,which was significantly better than other models.In predicting the prolonged postoperative length of hospital stay,the XGBoost model performed the best,with an AUC value of 0.84.Based on the above best-performing models,an ensemble strategy was implemented.The patients who followed the recommended strategy had reduced intraoperative bleeding and shorter postoperative length of hospital stay.Conclusion The use of tranexamic acid is associated with intraoperative bleeding and postoperative length of hospital stay.Personalized decision-making recommendation based on our constructed model can effectively improve the outcomes of the patients undergoing orthopedic surgery.

Haloacetic acids(HAAs),as a class of disinfection byproducts in drinking water,pose potential threats to human health,so the rapid,accurate and simultaneous detection of HAAs is of great significance for ensuring drinking water safety.Aiming at the challenges in HAAs detection and risk analysis,a novel method for synchronous rapid detection of seven kinds of HAAs in drinking water based on in situ derivatization technology and headspace gas chromatography was developed in this study.Through single-factor optimization experiments,the optimal reaction parameters for in situ derivatization were determined,including the type and dosage of salting-out agent,the acidity of reaction system,the amount of phase transfer catalyst,the dosage of derivatization agent,and the extraction solvent volume.Methodologic validation showed that the seven kinds of HAAs exhibited excellent linear relationships within their respective detection concentration ranges(R2>0.998).The method detection limits(MDLs)ranged from 0.04 to 0.33 μg/L,and the limits of quantification(LOQs)were between 0.14 and 1.34 μg/L.For real water samples,the average spiked recoveries of the seven HAAs ranged from 90.9%to 107.7%,with relative standard deviation(RSDs)between 1.55%and 6.49%,and the HAAs contents in all tested samples were below the limits specified in the Standards for Drinking Water Quality(GB 5749-2022)of China.This method was featured with simple operation,fast analysis speed,high sensitivity,and good accuracy,providing an efficient and reliable technical support for routine monitoring of HAAs contaminants in drinking water and showing promising application value for widespread promotion.

A solid-phase extraction-ultra-performance liquid chromatography-tandem mass spectrometry(SPE-UPLC-MS/MS)method was established for simultaneous determination of 10 kinds of neonicotinoid pesticide residues in aquaculture water.Based on the chemical properties of neonicotinoid pesticides and the matrix characteristics of aquaculture water,suitable temporary storage methods for water samples and appropriate solid-phase extraction columns were selected,and the extraction conditions(including elution solvents and sample loading volumes)were optimized.The method employed acetonitrile and 5 mmol/L ammonium acetate solution(containing 0.1%formic acid)as the mobile phase and an Oasis HLB solid-phase extraction column combined with PSA as a dispersive sorbent for sample purification.The method exhibited good linearity in detection of neonicotinoid in concentration range of 0.2-50 ng/mL(R2>0.99797),with a detection limit of 0.5 ng/L and a quantification limit of 1 ng/L,which were significantly lower than the maximum acceptable method detection limits(9-500 ng/L)for neonicotinoid insecticides in water published by the European Commission.In pond water,rice-fish water,and seawater,the average recoveries of the 10 target analytes were 74.6%-114.1%,with relative standard deviations ranging from 0.3%to 9.6%.Using this method,actual sample tests were conducted on the Pearl River water,Zhaoqing pond water,and Qingyuan rice-fish aquaculture water.The total concentration of five neonicotinoid pesticides in the Pearl River water ranged from 154.8 to 246.6 ng/L,the total concentration of four neonicotinoid pesticides in the Zhaoqing pond water was 95.0-176.1 ng/L,and the total concentration of three neonicotinoid pesticides in the Qingyuan rice-fish aquaculture water was 2.3-11.7 ng/L.This method was simple in operation,highly sensitive,and had strong resistance to interference.It was suitable for detection of trace neonicotinoid pesticides in aquaculture water and could provide technical support for construction of a green aquaculture environment and resolution of international trade disputes.

In the present study, a mouse model of coronary artery ligation was employed to evaluate the effects of Jiming Powder on mitophagy in the mouse model of myocardial infarction and elucidate its underlying mechanisms. A mouse model of myocardial infarction post heart failure was constructed by ligating the left anterior descending branch of the coronary artery. The therapeutic efficacy of Jiming Powder was assessed from multiple perspectives, including ultrasonographic imaging, hematoxylin-eosin(HE) staining, Masson staining, and serum cardiac enzyme profiling. Dihydroethidium(DHE) staining was employed to evaluate the oxidative stress levels in the hearts of mice from each group. Mitophagy levels were assessed by scanning electron microscopy and immunofluorescence co-localization. Western blot was employed to determine the levels of key proteins involved in mitophagy, including Bcl-2-interacting protein beclin 1(BECN1), sequestosome 1(SQSTM1), microtubule-associated protein 1 light chain 3 beta(LC3B), PTEN-induced putative kinase 1(PINK1), phospho-Parkinson disease protein(p-Parkin), and Parkinson disease protein(Parkin). The results demonstrated that compared with the model group, high and low doses of Jiming Powder significantly reduced the left ventricular internal diameter in systole(LVIDs) and left ventricular internal diameter in diastole(LVIDd) and markedly improved the left ventricular ejection fraction(LVEF) and left ventricular fractional shortening(LVFS), effectively improving the cardiac function in post-myocardial infarction mice. Jiming Powder effectively reduced the levels of myocardial injury markers such as creatine kinase(CK), creatine kinase isoenzyme(CK-MB), and lactate dehydrogenase(LDH), thereby protecting ischemic myocardium. HE staining revealed that Jiming Powder attenuated inflammatory cell infiltration after myocardial infarction. Masson staining indicated that Jiming Powder effectively inhibited ventricular remodeling. Western blot results showed that Jiming Powder activated the PINK1-Parkin pathway, up-regulated the protein level of BECN1, down-regulated the protein level of SQSTM1, and increased the LC3Ⅱ/LC3Ⅰ ratio to promote mitophagy. In conclusion, Jiming Powder exerts therapeutic effects on myocardial infarction by inhibiting ventricular remodeling. The findings pave the way for subsequent pharmacological studies on the active components of Jiming Powder.
Animals ; Myocardial Infarction/physiopathology* ; Mitophagy/drug effects* ; Mice ; Drugs, Chinese Herbal/administration & dosage* ; Protein Kinases/genetics* ; Male ; Ubiquitin-Protein Ligases/genetics* ; Humans ; Disease Models, Animal ; Mice, Inbred C57BL ; Signal Transduction/drug effects*

This study employed a mouse model of coronary artery ligation to assess the effect and mechanism of Jiming Powder on mitochondrial autophagy in mice with myocardial infarction. The mouse model of heart failure post-myocardial infarction was established by ligating the left anterior descending coronary artery. The pharmacological efficacy of Jiming Powder was evaluated through echocardiographic imaging, hematoxylin-eosin(HE) staining, and Masson staining. The levels of malondialdehyde(MDA), Fe~(2+), reduced glutathione(GSH), and superoxide dismutase(SOD) in heart tissues, as well as MDA immunofluorescence of heart tissues, were measured to assess lipid peroxidation and Fe~(2+) levels in the hearts of mice in different groups. Ferroptosis levels in the groups were evaluated using scanning electron microscopy and Prussian blue staining. Western blot analysis was conducted to detect the levels of key ferroptosis-related proteins, including nuclear factor erythroid 2-related factor 2(NRF2), ferritin heavy chain(FTH), glutathione peroxidase 4(GPX4), solute carrier family 7 member 11(SLC7A11), heme oxygenase 1(HO-1), and Kelch-like ECH-associated protein 1(KEAP1). The results showed that compared with the model group, both the high-and low-dose Jiming Powder groups exhibited significantly reduced left ventricular internal diameter in systole(LVIDs) and left ventricular internal diameter in diastole(LVIDd), while the left ventricular ejection fraction(EF) and left ventricular fractional shortening(FS) were significantly improved, effectively enhancing cardiac function in mice post-myocardial infarction. HE staining revealed that Jiming Powder attenuated myocardial inflammatory cell infiltration post-infarction, and Masson staining indicated that Jiming Powder effectively reduced fibrosis in the infarct margin area. Treatment with Jiming Powder reduced the levels of MDA and Fe~(2+), indicators of lipid peroxidation post-myocardial infarction, while increasing GSH and SOD levels, thus protecting ischemic myocardium. Western blot results demonstrated that Jiming Powder reduced KEAP1 protein accumulation, activated the NRF2/HO-1/GPX4 pathway, and up-regulated the protein expression of FTH and SLC7A11, exerting an inhibitory effect on ferroptosis. This study reveals that Jiming Powder exerts a therapeutic effect on myocardial infarction by inhibiting ferroptosis through the NRF2/HO-1/GPX4 pathway, providing a foundation for subsequent research on the pharmacological effects of Jiming Powder.
Animals ; Ferroptosis/drug effects* ; Myocardial Infarction/physiopathology* ; NF-E2-Related Factor 2/genetics* ; Mice ; Drugs, Chinese Herbal/administration & dosage* ; Male ; Heme Oxygenase-1/genetics* ; Phospholipid Hydroperoxide Glutathione Peroxidase/genetics* ; Humans ; Mice, Inbred C57BL ; Signal Transduction/drug effects* ; Disease Models, Animal

Traditional Chinese medicine(TCM) processing is a specialized pharmaceutical technique with the primary objective of reducing the toxicity of medicinal substances. Euphorbia pekinensis, E. ebracteolata, and E. fischeriana, all belonging to Euphorbiaceae, are classified as drastic purgative herbs, traditionally used for eliminating retained water, reducing swelling, resolving toxicity, and dispersing masses. However, these herbs are also associated with adverse effects such as abdominal pain and diarrhea. Accordingly, they are commonly processed with vinegar, milk, or Terminalia chebula decoction to reduce the toxicity. This review summarizes the chemical constituents, pharmacological activities, historical evolution of processing methods, and detoxification mechanisms of the three toxic Euphorbia species. The primary toxic constituents are terpenoids. Specifically, E. ebracteolata and E. fischeriana are rich in diterpenoids, while E. pekinensis contains diterpenoids, triterpenoids, and sesquiterpenoids. Studies have shown that vinegar processing promotes structural transformations of diterpenoids, including ether bond hydrolysis, lactone ring opening, esterification, oxidation, and epoxide ring cleavage, thereby reducing the content and toxicity of these compounds. Milk processing facilitates the dissolution of toxic components into the residual liquid of excipients, leading to decreases in their concentrations in the final decoction pieces. Processing with T. chebula decoction raises the levels of tannin-derived phenolic acids, which antagonize the adverse effects of the intestine. These findings reveal a shared detoxification pattern among the three toxic herbs. Accordingly, this review proposes the concept of a shared detoxification mechanism for toxic herbs belonging to the same family or genus. That is, toxic herbs belonging to the same taxon often exhibit similar toxicological profiles and can undergo detoxification through the same processing methods, reflecting common underlying mechanisms. Investigating such shared mechanisms across multiple species of the same genus offers a promising research strategy. Ultimately, the research into processing-induced detoxification mechanisms provides both theoretical and practical support for ensuring the safety of toxic TCM.
Euphorbia/classification* ; Drugs, Chinese Herbal/metabolism* ; Humans ; Animals ; Inactivation, Metabolic ; Medicine, Chinese Traditional

Metabolic-associated fatty liver disease (MAFLD) and osteoporosis (OP) are two very common metabolic diseases. A growing body of experimental evidence supports a pathophysiological link between MAFLD and OP. MAFLD is often associated with the development of OP. Rutaecarpine (RUT) is one of the main active components of Chinese medicine Euodiae Fructus. Our previous studies have demonstrated that RUT has lipid-lowering, anti-inflammatory and anti-atherosclerotic effects, and can improve the OP of rats. However, whether RUT can improve both fatty liver and OP symptoms of MAFLD mice at the same time remains to be investigated. In this study, we used C57BL/6 mice fed a high-fat diet (HFD) for 4 months to construct a MAFLD model, and gave the mice a low dose (5 mg·kg^-1) and a high dose (15 mg·kg^-1) of RUT by gavage for 4 weeks. The effects of RUT on liver steatosis and bone metabolism were then evaluated at the end of the experiment [this experiment was approved by the Experimental Animal Ethics Committee of Institute of Medicinal Biotechnology, Chinese Academy of Medical Sciences (approval number: IMB-20190124D₃03)]. The results showed that RUT treatment significantly reduced hepatic steatosis and lipid accumulation, and significantly reduced bone loss and promoted bone formation. In summary, this study shows that RUT has an effect of improving fatty liver and OP in MAFLD mice.