The progressive skeletal muscle degeneration observed in Duchenne Muscular Dystrophy (DMD) patients requires multiple cycles of satellite cells (SCs) activation to promote tissue regeneration. Dystrophic SCs present intrinsic defects, and the disrupting fibrotic niche hinders appropriate muscle recovery. Traditional 2D culture systems face challenges in modeling the DMD muscle niche and SCs behavior. Our aim was to validate a 3D culture of skeletal muscle spheroids (iSMS) for DMD modeling, as compared to the traditional 2D culture, while investigating the pathophysiological mechanisms of dystrophin deficiency in vitro. To compare iSMS with traditional 2D myogenic differentiation, we differentiated wild-type (WT), dystrophic (DMD) isogenic induced pluripotent stem cells (iPSCs), as well as iPSCs derived from DMD patients, characterized myogenic markers levels and assessed differences in proliferation and differentiation using RT-qPCR, immunofluorescence, and flow cytometry. Our data showed that iSMS improved PAX7 expression in vitro, while MYOD1, MYOG, MYF5, and MYH3 expression were significantly reduced. These findings suggest that, at three weeks of myogenic differentiation, iSMS cultures retained satellite-like cells in a less activated, progenitor-like state. Accordingly, we identified higher expression of canonical Notch signaling genes such as JAG1 and NOTCH1 in iSMS compared to 2D. We also characterized the response of 2D and iSMS to terminal differentiation medium, providing a valuable comparison with muscle fibers derived from human adult myoblasts. Additionally, we showed that DMD iSMS-derived progenitors proliferated at reduced levels compared with WT, a characteristic not observed in progenitors derived from 2D cultures. Finally, we performed iSMS and 2D myogenic differentiation of iPSC lines from three patients with DMD. Our results highlight important advantages of using the iSMS differentiation platform over 2D for DMD in vitro modeling. Exploring these 3D systems may help to gain a deeper understanding of SCs behavior to advance in novel treatments for DMD, which might be applicable to other forms of muscular disorders.
This study aims to develop a precise predictive model for leaching chalcopyrite concentrates. It employs a leaching system comprising 1-hexyl-3-methylimidazolium hydrogen sulfate ([Hmim][HSO4]) and hydrogen peroxide (H2O2), offering a more efficient alternative to conventional hydrometallurgical approaches. Gene expression programming (GEP) was used to develop this model. To construct these GEP models, 120 experimental data points were collected initially. Input variables included time, acid concentration, temperature, particle size, oxidant concentration, stirring speed, and solid/liquid ratio, while output variables included copper extraction percentage. For modeling purposes, the experimental dataset was randomly partitioned into a training set (84 data points) and a testing set (36 data points). A correlation analysis (BCA) revealed weak linear correlations between input variables, justifying the use of advanced methods such as GEP. Using criteria such as coefficient of determination (R2), mean absolute error (MAE), and root relative square error (RRSE), we proposed the optimal model (GEP-3). As a new model with simplified mathematical expressions for accurate prediction of copper extraction from chalcopyrite concentrate, this model achieves R2 = 0.976, MAE = 2.80, and RRSE = 0.152 in the training set. Sensitivity analysis revealed that temperature, oxidant concentration, and particle size were the most influential parameters on the copper extraction percentage. By taking into account practical or economic constraints, the proposed model enables the optimization of the leaching process to maximize copper extraction and minimize material consumption.
Visual impairment affects over 2.2 billion people worldwide and the major causes include age-related macular degeneration (AMD), glaucoma, and diabetic retinopathy. For research in these areas, although animal models offer a more physiologically complex system than in vitro approaches, their use raises ethical considerations, and species-specific differences such as variations in protein sequences and signaling pathways. This can limit the direct translatability of the outcomes. Traditional 2-D cell cultures, in contrast, lack the multicellular organization and dynamic microenvironment necessary to replicate human retinal complexity. Retinal organoids (ROs), three-dimensional tissue constructs derived from pluripotent stem cells, have emerged as a promising model due to their human origin and complex cellular interactions that cannot be achieved in conventional 2-D/3-D co-culture models. In this review, we provide a brief overview of the evolution from 2-D to 3-D retinal models, highlight the structural and functional features of ROs including the presence of layered retinal architecture, photoreceptor outer segment formation, and light-responsive electrophysiological activity and summarize their applications in disease modeling, drug discovery, and gene and cell therapy. ROs represent a significant advancement over traditional models by enabling the recapitulation of human-specific retinal development, facilitating the study of patient-derived disease phenotypes, and providing a platform for personalized therapeutic screening. Their development has deepened understanding of pathological mechanisms in conditions such as retinitis pigmentosa and AMD, while enabling preclinical testing of targeted interventions like CRISPR-based gene editing and photoreceptor cell replacement. Nonetheless, challenges remain in fully replicating retinal vascularization, long-term functional maturation, and synaptic connectivity, underscoring the need for continued refinement and integration with complementary model systems.
The growing worldwide population has increased the use of electric arc furnaces (EAF), resulting in a surge of EAF slag and a huge environmental concern. EAF slag's complex physical qualities have a considerable impact on concrete's mechanical performance, mainly its compressive strength (CS). This study introduces a novel framework for forecasting the CS of EAF slag concrete that uses advanced machine learning models such as Extreme Gradient Boosting (XGB), AdaBoost (ADB), Random Forest (RF), Hybrid XGB-RF, and Hybrid XGB-ADB. A full dataset of 730 samples was meticulously created, containing essential input parameters such as binders, aggregates, and other necessary variables, with CS as the desired outcome. Based on the findings, the XGB model showed highest accuracy, with an test R2 of 0.951, MAPE of 1.128, and RMSE of 1.393 MPa, indicating its potential for dependable performance forecasting. In addition, the hybrid XGB-RF model also demonstrated strong predictive accuracy, achieving an R2 value of 0.947 during the testing phase. Furthermore, explainability tools such as SHapley Additive ExPlanations (SHAP) and partial dependence curves (PDCs) identified the curing period and content of cement as the most influential factors to predict the CS of EAF slag concrete. The methodologies and outcomes of this study will help to reduce reliance on resource-intensive experimental methods, pave the way for efficient, precise, and ecologically conscientious concrete design.
Health-related quality of life (HRQoL) is a vital indicator of evaluating care outcomes and prognosis, yet little is understood about its developmental trajectories in older patients with chronic pain. This study aimed to identify latent HRQoL trajectories and their predictors, and to develop explainable machine learning models for predicting HRQoL deterioration. This prospective cohort study assessed 608 older patients with chronic pain at admission and at 1, 3, and 6 months post-admission, collecting data on HRQoL, general characteristics, pain level, activities of daily living (ADL), depression, and perceived social support. Growth mixture modeling was applied to identify trajectories of physical and mental HRQoL. Predictors were selected using LASSO regression and SVM-RFE. Nine explainable machine learning models were developed for both components, and SHAP interpreted the outputs. An HRQoL decision-support dashboard was developed to facilitate potential clinical application. Three physical HRQoL trajectories were identified: Stable High, Decline and Low Stability, alongside two mental HRQoL trajectories: Improvement and Decline. Key predictors included education level, pain duration, pain level, ADL, depression, and perceived social support, with ADL and pain level being the most influential for physical and mental HRQoL, respectively. This dual-trajectory study identified five distinct HRQoL patterns in older patients with chronic pain, elucidating key predictors via explainable machine learning. The proposed HRQoL decision-support dashboard may provide an interpretable tool to support understanding of predictive relationships and assist healthcare professionals in HRQoL assessment. Not applicable.
Breast cancer patients often experience significant psychological distress. This study examined distress trajectories from diagnosis to 6 months post-treatment and explored differences across demographic, medical, and psychosocial subgroups. In this prospective cohort study, 528 patients with breast cancer were recruited between 1 December 2023 and 31 December 2024. Assessments were conducted at baseline (at diagnosis, T0), after the first treatment (T1), mid-treatment (T2), at treatment completion (T3), and at three (T4) and six months (T5) post-treatment. Growth mixture modeling (GMM) was used to identify distinct trajectories of psychological distress. Multinomial logistic regression analysis was performed to examine associations between patient-related factors and trajectory membership. Three psychological distress trajectories were identified: a high-distress remission group (17.05%), a moderate-stable distress group (11.93%), and a low-fluctuating distress group (71.02%). Multivariable analyses showed that higher educational attainment, breast-conserving surgery, early disease stage, partial self-management ability, and strong social support were associated with membership in the moderate-stable or low-fluctuating groups (p < 0.05). Employment, health insurance coverage, avoidant medical coping style, and higher baseline anxiety and depression scores were concurrently associated with membership in the high-distress remission group (p < 0.05). Although psychological distress generally decreased over time, 71.02% of patients followed a low-fluctuating trajectory, 11.93% maintained moderate distress with potential risk of persistence, and 17.05% showed high initial distress that remitted substantially within 6 months. Continuous monitoring and early psychosocial support are recommended, particularly for patients with moderate- or high-risk trajectories.
Proximal femoral fractures are highly prevalent in Japan, with over 200,000 cases annually and a rising trend. Fracture liaison service (FLS) interventions improve osteoporosis treatment initiation and reduce refracture rates. The content of FLS interventions varies by institution, and the effectiveness of our intervention remains unclear. The aim of this study was to evaluate the effectiveness of our FLS intervention in preventing fragility fractures within 1 year after proximal femoral fracture surgery. A retrospective case-control study was performed on patients aged ≥ 50 undergoing surgery for proximal femoral fracture between February 2021 and January 2024. Patients were divided into non-FLS (pre-August 2022) and FLS groups. Data including demographics, comorbidities, fracture type, medication initiation, and refracture occurrence within 1 year were extracted. Statistical analyses involved Mann-Whitney U, χ2 tests, and Cox proportional hazards modeling. Among 521 eligible patients, osteoporosis medication initiation within 3 months improved from 14% in the non-FLS group to 100% in the FLS group (p < 0.05). Time to medication initiation decreased from 20 to 12 days (p < 0.05). The refracture rate was significantly lower in the FLS group (1.8% vs. 5.7%, p < 0.05). Multivariate analysis showed FLS intervention significantly reduced refracture risk (HR 0.32, 95% CI 0.12-0.89, p = 0.03) and robust in sensitivity analyses for cognition, walking ability, and discharge destination. FLS intervention effectively reduced fragility fractures within 1 year postoperatively by enhancing early osteoporosis treatment initiation. Continued FLS programs and long-term follow-up are recommended to sustain benefits.
As an extension of standard graphs, hypergraphs have demonstrated significant advantages in modeling high-order complex relationships compared with standard graphs. Existing literature has witnessed the great success of hypergraph representation learning methods in classifying nodes. However, most of them seek to obtain low-dimensional crisp representations, overlooking the fuzzy and uncertain nature of node attributes. In fact, node attributes such as paper keywords may contain noise or be incomplete, which leads to uncertain semantics. To address this issue, in this paper, we propose learning fuzzy representations for hypergraph node classification. Specifically, we develop a novel method called Hypergraph Collaborative Fuzzy Network (HyperCFN), which studies hypergraph representations with fuzzy logic. Firstly, HyperCFN augments the original hypergraph into two hypergraphs, which are then put into the proposed fuzzy hypergraph encoders. The fuzzy hypergraph encoders consist of hypergraph collaborative networks and fuzzy logic to learn fuzzy representations for every node and hyperedge. Subsequently, the learned representations are enforced node-, hyperedge-, and membership-level contrast. Lastly, to further preserve the hypergraph structure, we develop decoders to reconstruct the augmented hypergraphs. We perform extensive experiments on several datasets, and the promising results demonstrate that the effectiveness of the proposed model and learning fuzzy representations for hypergraphs is valid.
Fluorinated gel polymer electrolytes (FGPEs) prepared via in situ polymerization are expected to expedite the large-scale application of lithium metal batteries (LMBs) by enabling stable LiF-rich solid electrolyte interphases (SEIs) and good compatibility with high-voltage cathodes. However, the electron-withdrawing nature of fluorine units retards polymerization kinetics of such monomers, resulting in GPEs with compromised mechanical performance and cycling durability. Herein, a design principle for in situ formation of fluorinated copolymers is proposed to regulate the polymerization kinetics of trifluoroethyl methacrylate (TFEMA)-typed monomers. Such strategy yields relatively uniform polymer chains with moderate molecular weights, which are subsequently crosslinked to form a robust fluorinated-nitrogenated copolymer network (FNPE). The tailored polymer matrix integrates the capabilities to form a LiF-containing SEI promoted by fluorinated segments, enhanced mechanical robustness, and a Li3N-rich interphase contributed by the N-isopropylacrylamide (NIPAM) domains. Consequently, the FNPE achieves NCM811(6.8 mg cm-2, 1.2 mAh cm-2)//Li full cells with high capacity retention (> 80%, 225 cycles), and applicable in wide temperature range (-15 to 60°C) and pouch cell configuration (40 µm Li). Through experimental and multiscale modeling investigations, this work elucidates the intrinsic kinetic challenge for in situ formed FGPEs and provides a new design principle of copolymer-type electrolytes for durable LMBs.
Complications after pancreatoduodenectomy for pancreatic ductal adenocarcinoma (PDAC) are associated with delays or omission of adjuvant chemotherapy (AC). Similar data for patients who undergo distal pancreatectomy (DP) are lacking. A retrospective cohort study was conducted using the SEER-Medicare database to identify patients who underwent upfront DP for PDAC (2010-2019). Multilevel logistic regression and Cox proportional hazards models were used to evaluate the association of postoperative complications with AC omission and delay as well as survival endpoints based on receipt of AC. Of 1029 patients identified, 613 (59.6 %) received AC. Patients with complications had lower rates of AC (50.0 % vs 61.1 %; p = 0.013) and multi-agent AC (25.0 % vs 28.2 %; p = 0.039) and higher rates of delays in AC (42.9 % vs 21.4 %; p < 0.001) than those without complications. In multivariable analysis, complications were associated with a lower rate of AC (hazard ratio [HR], 0.67; 95 % confidence interval [CI], 0.54-0.84; p < 0.001) and a higher rate of delayed AC (odds ratio [OR], 3.36; 95 % CI 1.92-5.91; p < 0.001). For survival, receipt of AC overall (HR, 0.56; 95 % CI 0.47-0.67; p < 0.001), even when delayed (HR, 0.72; 95 % CI 0.57-0.90; p = 0.005), was associated with better overall survival (OS) than no AC. However, delayed AC was associated with worse OS than timely AC (HR, 1.27; 95 % CI 1.01-1.62; p = 0.04). Patients who experienced a postoperative complication after DP for left-side PDAC had lower rates of AC overall and higher rates of delayed AC, both associated with worse OS.
Functional validation of candidate genes in congenital anomalies of the kidneys and urinary tract (CAKUT) and other disorders is essential for translating genetic discoveries into clinical applications. Conditional knockout mouse models are indispensable for studying gene function in complex organ systems. The Short Conditional intrON (SCON) system accelerates the generation of such models by inserting the artificial SCON into a coding exon. SCON is designed to be spliced out after transcription, without affecting gene expression. Upon Cre activity, SCON is converted into the ΔSCON allele which cannot be spliced out, introducing premature termination codons (PTCs) to inactivate the gene. Previous validation of the SCON system in mice has focused primarily on phenotypic outcomes. Here, we provide a molecular characterization of the SCON system in Cdh12-a candidate gene implicated in kidney damage in CAKUT. We found that both Cdh12SCON and Cdh12ΔSCON alleles caused unintended skipping of the exon downstream of the insertion site, culminating in a frameshift and PTC. Consequently, the Cdh12SCON allele led to a ~ 25% reduction in mRNA expression, indicating that it was not transcriptionally inert as designed. Despite unintended exon skipping, the Cdh12ΔSCON allele still effectively suppressed mRNA expression. These findings highlight the importance of transcript-level characterization of engineered alleles prior to functional studies, as artefactual splicing events may occur across multiple gene-targeting strategies, including artificial intron-based conditional alleles as shown here.
The association between preoperative peripheral nerve block (PNB), major adverse cardiovascular events (MACE), and postoperative length of hospital stay (LOS) in elderly patients who underwent major thoracic and abdominal surgery remains unclear. This study aims to explore the potential mediating effect of MACE on the association between preoperative PNB and postoperative LOS using a statistical mediation framework. In this retrospective cohort study, perioperative data were collected from elderly patients (aged over 65 years) who underwent major thoracic and abdominal surgery. Mediation analysis was employed to examine the relationships between PNB, MACE, and postoperative LOS. A total of 1915 patients were included in the analysis, with 68.7% (1316/1915) receiving preoperative PNB. Compared to patients who did not receive PNB, those who did had a significantly lower incidence of MACE (P < 0.001) and a shorter postoperative LOS (P < 0.001). The adjusted total and direct associations of PNB with postoperative LOS were - 0.809 days (95% confidence interval [CI], -1.236 to -0.390; P < 0.001) and - 0.661 days (95% CI, -1.077 to -0.250; P = 0.003), respectively. A statistically significant indirect association via MACE was observed (adjusted β=-0.149 days; 95% CI, -0.271 to -0.060; P < 0.001), indicating that 18.1% (95% CI, 6.7% to 41.0%) of the total association was statistically attributable to the indirect pathway through MACE under the model assumptions. A sensitivity analysis excluding postoperative covariates yielded consistent results (proportion mediated: 25.3%). Our findings suggest that the observed association between preoperative PNB and reduced postoperative LOS in elderly patients following major thoracic and abdominal surgery may be partly explained by a statistically significant indirect pathway through a reduction in MACE, potentially accounting for approximately 18% of the total effect. These findings are hypothesis-generating and represent statistical associations rather than demonstrated causal mechanisms. ChiCTR2400087610; https://www.chictr.org.cn.
Visual information extraction (VIE) from visually rich documents remains challenging due to high layout variability and real-world impairments. Existing methods typically rely on sequential OCR pipelines or end-to-end models requiring extensive labeled data and layout-specific training, limiting their scalability. We propose a classification-guided large vision-language model (LVLM) framework for multi-type VIE that achieves high accuracy with minimal supervision. The approach decouples document-type classification from content extraction and employs in-context learning (ICL)-based dynamic prompt engineering to inject task-specific knowledge, enabling robust zero-shot inference across diverse layouts. From a theoretical perspective, the proposed method can be viewed as a form of conditional computation that reduces task uncertainty and improves information efficiency during prompt-based inference. Evaluated on a real-world bidding dataset with 16 certificate types, our zero-shot method (based on Qwen2.5-VL-7B) outperforms a strong supervised baseline by 18.35 percentage points in F1-score (86.43% vs. 68.08%) and 0.23 in normalized edit distance (0.90 vs. 0.67). Optional domain-specific fine-tuning further improves performance to 93.65% F1 and 0.93 NED, demonstrating superior robustness against seals, watermarks, and low contrast. The framework offers an efficient, scalable solution for complex document understanding in office automation. Code is available at https://github.com/FairmeHIT/Multi-VIE, and fine-tuned models at https://huggingface.co/fairme/Qwen2.5-VL-7B-SFT.
Insurance fraud detection remains challenging to predict in reality because claims data is often uneven among classes, and the information concerning claims is often multidimensional and nonhomogeneous. The present research used a unified evaluation framework to assess the predictive and interpretive capabilities of three distinct model families: CatBoost (tree-based ensemble learning), Bi-GRU with Attention (sequence-oriented learning), and TabTransformer (categorical feature contextual). The model families were tested using a standardised experimental protocol.The study is novel in the sense of a cross-model interpretability framework that unites Shapley Additive Explanation (SHAP)-based feature attribution with attention-based contextual analysis to enable a clear comparison of model reasoning between the suggested frameworks. The data on which the experiments were done consisted of 4,000 life insurance claims that were characterized in terms of 83 attributes. Common preprocessing procedures like missing values, scaling numerical variables, and selecting highly correlated variables were used before training the models. Experimentally, CatBoost is proven to be the most precise on legitimate claims, Bi-GRU is the most recall on fraudulent claims, and TabTransformer is the best in terms of tradeoff between accuracy, interpretability, and computational efficiency. Practical characteristics such as the quantity of claim, tenure in a policy, and diagnosis were repeatedly emphasized in both SHAP and attention analyses. Combined, the current research study provides a consistent and explainable benchmark that may be applied to conduct fraud detection research reliably and assist practitioners in choosing models that are accurate and understandable.
Periodontitis, a chronic inflammatory disease, is increasingly prevalent among young people and impairs their quality of life. Adverse childhood experiences (ACE), depressive symptoms, and suboptimal health status (SHS) are linked to health risks and chronic diseases, but their interrelationships with periodontitis in Chinese young adults remain unclear. This study aimed to explore associations among these factors. From December 2024 to May 2025, 2,888 participants (aged 18-35) from Tongji Hospital completed surveys on demographics, ACE, depressive symptoms, and SHS. Periodontitis was diagnosed according to the 2018 criteria. Simple, parallel, and chain mediation models were used, controlling for age, sex, marital status, and smoking. Periodontitis prevalence was 25.00% and higher in married individuals (P < 0.001) and smokers (P = 0.004). ACE correlated positively with depressive symptoms (r = 0.28, P < 0.001), SHS (r = 0.19, P < 0.001), and periodontitis (r = 0.16, P < 0.001). Mediation analyses showed: Simple model: Depressive symptoms and SHS partially mediated the effect of ACE on periodontitis (indirect effect = 0.011 for both). Parallel model: Only SHS significantly mediated the effect (indirect effect = 0.011). Chain model: ACE was related to periodontitis via "depressive symptoms → SHS" (indirect effect = 0.010), with significant direct and indirect effects. ACE associated with higher periodontitis risk in young people. This association included both a direct link between ACE and periodontitis, and an indirect link through the chain pathway of "depressive symptoms → SHS"; among these pathways, SHS was a key mediator. The study was registered in the Chinese Clinical Trial Registry (ChiCTR) with the registration number ChiCTR2500103464. Childhood trauma can exert long‐term impacts on health, including oral health. This study involving 2,888 Chinese young adults aged 18‐35 found that 25% of the participants had periodontitis. Those who experienced childhood abuse, neglect, or family issues showed a higher association with the disease. The research revealed two pathways linking early trauma to periodontitis: a direct association and an indirect chain of “depressive symptoms → suboptimal health status (e.g., persistent fatigue).” While depressive symptoms played a role, suboptimal health status was the critical mediator. Higher periodontitis rates in married individuals and smokers may relate to stress or lifestyle factors. The findings suggested that early identification of childhood trauma, combined with interventions targeting mental health or overall well‐being (e.g., counseling, health management), could be more effective than oral care alone in prevention. This underscored the association between early‐life experiences and long‐term health and the need for integrated interventions.
暂无摘要（点击查看详情）
In this review we comprehensively discuss organic cation transporter novel 1 (OCTN1), encoded by the SLC22A4 gene as a member in the solute carrier 22 (SLC22) family, which facilitates the cellular transport of diverse cationic and zwitterionic substrates. OCTN1 is highly expressed in many vital organs in humans, where it facilitates absorption and distribution of both endogenous compounds and therapeutic drugs. Among its substrates, ergothioneine (EGT) serves as the primary antioxidant and anti-inflammatory molecule, underscoring the essential role of OCTN1 in cellular defense and inflammation control. Genetic polymorphisms in SLC22A4 significantly alter OCTN1 expression, substrate affinity, and drug pharmacokinetics, with strong associations to susceptibility and treatment outcomes in human diseases. Insights from knockout models revealed that OCTN1 deficiency leads to reduced EGT availability, heightened oxidative stress, and aggravated inflammation, particularly in the tissues such as intestine, liver and lung. Moreover, OCTN1 activity is dynamically regulated by epigenetic modifications, cytokines, and hormones, linking it to immune modulation and disease progression. Put together, OCTN1 plays a defined role via high-affinity EGT transport, while its broader transport capacity and pharmacological relevance remain under investigation, with possible - though not yet established - implications for inflammation-associated biomarker development.
The deltoid ligament (DL) is the primary stabilizer of the medial ankle, but its injury mechanisms remain poorly understood. This study aimed to investigate the injury risk and mechanisms of individual DL bundles under both acute and chronic conditions to inform prevention and treatment strategies. A validated finite element model of the human foot was used to examine peak stresses in DL bundles under four acute loading scenarios. Chronic loading was simulated by applying gait loads after transecting the lateral ligaments, and the resulting DL stresses were compared with those of the intact model. Additionally, thirty-nine rats were assigned to three groups: a lateral ligament rupture group (LR, n = 13), a tibialis posterior tendon rupture group (TPR, n = 13), and a sham group (n = 13). After 6 weeks of treadmill running, the mechanical properties and histological characteristics of the DL, along with ankle joint morphology and articular stresses, were evaluated to further verify the hypothesized mechanisms of chronic injury. Under acute loadings, the tibiocalcaneal ligament (TCL), anterior tibiotalar ligament (ATTL), and deep posterior tibiotalar ligament (dPTTL) showed the highest stress under pronation-external rotation loading. Lateral ligament rupture increased DL stress during gait. After 6 weeks of treadmill running, the LR and TPR groups showed roughened articular surfaces with osteophyte formation, increased articular stress, decreased talar bone volume fraction, lower failure load and stiffness ratios of the DL (p < 0.01), reduced fluorescence intensity of COL1, and elevated levels of COL3, MMP-2 and IL-1β compared with the sham group (p < 0.01). The TCL, ATTL, and dPTTL bundles are particularly susceptible to acute injury, with pronation-external rotation posing the greatest risk. Chronic degeneration of the DL occurs following rupture of the lateral ligament or tibialis posterior tendon, with a more pronounced effect after lateral ligament rupture.
Optical Chemical Structure Recognition (OCSR) aims to convert two-dimensional molecular images into machine-readable formats such as SMILES strings. Deep learning has substantially improved OCSR performance, yet most methods rely on synthetic training data and struggle to generalize to real-world inputs, especially hand-drawn diagrams, where stroke width, geometry, and drawing conventions vary widely across individuals. In this work, we propose an image-to-graph model AdaptMol that enables effective transfer from synthetic to real-world data without requiring manual graph annotations in the target domains. AdaptMol is an integrated pipeline that starts with training a base model on synthetic data, and then refines model representations through unsupervised domain adaptation and self-training. Our key insight is that bond features are domain-invariant in nature; they encode structural relationships between atoms that are independent of visual variations across domains. Thus, during domain adaptation, we align bond-level feature distributions via class-conditional Maximum Mean Discrepancy (MMD) to enforce cross-domain consistency. We also design a comprehensive data augmentation strategy to enhance the robustness of the base model, facilitating stable self-training on unlabeled target samples. On hand-drawn molecular images, our model achieves 82.6% accuracy and outperforms the best prior method by 10.7 points, while maintaining competitive performance across four benchmarks comprising molecular images from scientific literature and patent documents.Scientific contributionWe propose AdaptMol, an image-to-graph model that predicts molecular structures as graphs of atoms and bonds, achieving effective transfer from synthetic to hand-drawn molecular images without requiring target domain graph annotations. We combine class-conditional Maximum Mean Discrepancy to align bond features across domains with comprehensive data augmentation to increase training data variation, jointly improving base model accuracy sufficiently for self-training and addressing the critical failure mode of prior approaches that begin with insufficient accuracy. We further introduce a dual position representation that supervises atom positions through both discrete coordinate tokens and continuous spatial heatmaps to reduce false positives in atom localization.
Postoperative gastrointestinal (GI) bleeding is a serious complication after hip fracture surgery in older adults, yet perioperative risk stratification remains limited because commonly used GI-bleeding scores are not tailored to orthopedic settings. This study aimed to develop and internally validate an interpretable model to predict postoperative GI bleeding risk in elderly hip fracture patients, using data routinely available during the perioperative period. We retrospectively included 342 elderly patients who underwent hip fracture surgery at the Third Hospital of Hebei Medical University from January to December 2023. The outcome was GI bleeding within 1 month after surgery, confirmed by medical records and/or telephone follow-up. Patients were randomly split into a training set (n = 242) and a validation set (n = 100). Predictors were screened using LASSO with 10-fold cross-validation, followed by multivariable logistic regression to identify independent risk factors. Ten prediction algorithms were trained and compared. Model performance was assessed by AUC, calibration, and decision curve analysis, and interpretability was evaluated using SHAP. GI bleeding occurred in 38 patients (11.1%). Multivariable analysis identified four independent predictors: alcohol consumption history (OR 8.109, 95% CI 2.463-26.69), glucocorticoid use (OR 4.922, 95% CI 1.055-22.97), NSAID use (OR 6.851, 95% CI 1.811-25.915), and higher systemic immune-inflammation index (SII) (OR 1.001, 95% CI 1.000-1.002). Among the tested models, LightGBM showed the best overall performance, with AUCs of 0.843 (training) and 0.817 (validation), good calibration, and the highest net benefit on decision curve analysis. SHAP results ranked feature importance as SII, NSAID use, alcohol consumption history, and glucocorticoid use, consistent with regression findings. We developed and validated an interpretable LightGBM model that predicts postoperative GI bleeding risk in elderly hip fracture patients using routinely available clinical data. The final model incorporates only preoperative variables, systemic inflammation, NSAID use, alcohol history, and glucocorticoid use, supporting its application for early risk stratification prior to surgery.