共找到 20 条结果
暂无摘要（点击查看详情）
To establish and validate interpretable machine-learning (ML) models for early assessment and identification of Chinese women at risk of excessive gestational weight gain (EGWG). We performed a prospective observational study with pregnant women whose gestational age averaged 19 weeks or less. These women attended the obstetric clinic of a tertiary hospital in CentralSouth China between January and June 2023, and again from April to May 2024. Women completed standardized questionnaires, and their gestational weight gain (GWG) was recorded until delivery. We conducted feature selection by applying the Boruta algorithm together with the least absolute shrinkage and selection operator (LASSO) algorithm, We used four ML models-the logistic regression(LR), light gradient boosting machine(LightGBM), extreme gradient boosting(XGBoost), and random forest (RF) models-optimizing its hyperparameters by grid search and 5-fold crossvalidation. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), confusion matrix, kappa statistic, calibration curves, and decision curve analysis. To enhance model interpretability, the SHapley Additive exPlanations (SHAP) framework was applied to quantify and rank the contribution of individual predictors. Based on the key predictive features, a web-based interactive calculator was developed using Python and the Flask micro-framework to facilitate clinical application. We enrolled 578 pregnant women in all. The combined use of the Boruta and LASSO algorithms screened ten critical predictors. The LightGBM model showed superior predictive performance with an accuracy of 88.6%, sensitivity of 87.5%, specificity of 89.9%, kappa statistic of 0.770, and AUC of 0.926 (95% CI: 0.889-0.962) in the test cohort. The SHAP analysis indicated that the body image in pregnancy, protective motivation for gestational weight management, parity, the weekly frequency of consuming sugar-sweetened beverages, desserts, and Western-style fast food, and moderate-intensity physical activity time were the major determinants that influenced model prediction. An online calculator was developed and made available for clinicians at: http://39.103.64.176/. We established an interpretable ML model for predicting the risk of EGWG. The LGBM model exhibited higher predictive accuracy and may serves as a powerful tool for the early detection and individualized management of the EGWG risk among Chinese pregnant women.
Accurate prediction of in-hospital mortality for patients with severe community-acquired pneumonia (SCAP) complicated by respiratory failure admitted to the intensive care unit (ICU) remains a critical challenge. This study aimed to develop and validate a machine learning (ML) model to predict this risk and compare its performance with conventional scoring systems. In this retrospective study, data from 164 patients with SCAP and respiratory failure admitted to the ICU between January 2017 and January 2024 were analyzed. Patients were randomly divided into a training set (n = 114) and a validation (test) set (n = 50). Forty-five clinical features collected at admission were used as candidate predictors. The Least Absolute Shrinkage and Selection Operator (LASSO) regression was employed for feature selection. Six ML models, including Gradient Boosting Decision Tree (GBDT), Random Forest (RF), eXtreme Gradient Boosting (XGBoost), Decision Tree(DT),Support Vector Machine (SVM), and Logistic Regression(LR), were constructed and evaluated.Use SHAP analysis to assess the contribution of each feature in a machine learning model. Construct a nomogram using the top six most influential features. The GBDT model demonstrated the best predictive performance, achieving an area under the receiver operating characteristic curve (AUC) of 0.83 (95% CI: 0.757-0.927) in the internal validation set, significantly outperforming the Acute Physiology and Chronic Health Evaluation II (APACHE-II, AUC = 0.70). Calibration curves demonstrated good agreement between predicted and observed mortality risks, particularly across the mid-probability range. Decision curve analysis indicated that the model provided a higher net benefit than "treat-all" and "treat-none" strategies across a broad range of threshold probabilities. SHapley Additive exPlanations (SHAP) analysis identified lactate, D-dimer, temperature, albumin, Prothrombin Time and Fraction of Inspired Oxygen as the six most influential predictors of in-hospital mortality. Based on these key predictors, we further developed a simplified nomogram to facilitate bedside risk estimation. The GBDT ML model, developed from routinely available clinical data, provides a highly accurate and clinically interpretable tool for predicting in-hospital mortality in SCAP patients with respiratory failure. It outperforms traditional severity scores and holds promise for assisting clinicians in risk stratification and early intervention.
Hearing is a critical sensory function for humans, and the dynamic behavior of the human ear is influenced by its constitutive relation. In this paper, a finite element model (FEM) of the middle ear with a nonlinear constitutive relation and the inner ear with a viscoelastic constitutive relation was developed based on CT scanning data of the clinical human ear. The frequency-response at the tympanic membrane (TM) and stapes footplate (SF), as well as the amplitude-frequency response and time-domain response curves at different positions of the basilar membrane (BM) were obtained by using the acoustic-solid and fluid-solid coupling dynamic analysis. The results indicated that the previous linear elastic model exhibits an error of 28% relative to experimental data, while the nonlinear-viscoelastic model proposed in this paper demonstrates an error of only 5% relative to experimental data. This model not only better align with the actual biomaterial properties of ear tissue, but also more accurately simulates the mechanical behavior of the human ear during sound perception. The realistic biomaterial model provides a reasonable and accurate numerical simulation platform for studying the biomechanical behavior of the real human ear, laying an applied foundation for related research into hearing damage.
 : Cystic fibrosis (CF) is likely underdiagnosed in Caribbean populations due to non-representative cystic fibrosis transmembrane conductance regulator (CFTR) variant screening panels, limited newborn screening programs, and structural healthcare barriers. Data from 2022 indicate substantial populations with European ancestry in Puerto Rico (1.4 M, 42.7%) and the Dominican Republic (1.4 M, 57.9%), yet the true burden of CF in the broader Caribbean remains largely undocumented. Current diagnostic frameworks, largely based on European-derived CFTR variant distributions, fail to capture the true burden of CF in Caribbean populations, leading to underestimated prevalence and delayed or missed diagnoses. To synthesize registry, clinical, and published data to identify barriers to accurately assessing CF prevalence in Caribbean populations. This narrative literature review integrates CF registries, published data on CFTR variant distribution, population ancestry data, and clinical observations from CF centers in Puerto Rico and the Dominican Republic. Clinical insights were derived from pediatric patients evaluated at the Pediatric Rare Lung and Asthma Institute in Puerto Rico and the CF Clinic at Robert Reid Cabral Children's Hospital in the Dominican Republic. CFTR variant patterns differ from those in the United States, with higher frequencies of rare variants such as p.Ala559Thr. Standard screening panels may miss these variants, contributing to underdiagnosis. Limited newborn screening, misdiagnosis, and restricted access to CFTR modulator therapies further exacerbate disparities. Structural, diagnostic, and genetic factors hinder accurate CF prevalence estimates in the Caribbean, highlighting the need for region-specific research, improved screening, and expanded access to therapies.
We introduce a graph neural network framework that integrates density functional theory (DFT) calculations and scanning tunneling microscopy experimental data to predict molecular self-assembly behavior on metal surfaces. We constructed a data set comprising 20 diverse aromatic precursor molecules and their corresponding assembled structures on Au, Ag, and Cu substrates. Leveraging DFT-derived descriptors, feature importance analysis identified the molecule-substrate interactions and interfacial charge transfer as the dominant factors governing assembly behavior. A modified graph attention network model, trained on this multisource data set, achieved exceptional predictive accuracy, which exceeded 95% in classifying molecular arrangements and attained a coefficient of determination (R2) of 0.985 for adsorption energy regression. The model's generalizability was further validated by accurately classifying the self-assembled layers of three previously unseen molecules. This study establishes a machine learning framework that bridges computational and experimental insights, paving the way for the rational design of surface-supported functional nanostructures.
This study compared knowledge-based and deep-learning dose prediction models with clinically approved plans for prostate radiotherapy with focal boosting. The knowledge-based model accurately reproduced clinical plans. Deep-learning predictions showed median variations of 5% (IQR < 5%) for bladder and rectum in high-dose regions and 3.8 Gy higher mean dose (IQR 1.2 Gy) for the femoral heads. However, pudendal artery mean doses were 15 Gy higher (IQR: 10-19 Gy) and urethral V62.4Gy exceeded 50%, above the 2% constraint, likely reflecting their absence in the training dataset. While deep-learning models provide a consistent spatial framework for plan optimization, expert review remains essential.
Cystic hepatic lesions are a group of heterogeneous entities commonly encountered in clinical practice. The prevalence of cystic hepatic lesions has been reported to be as high as 15%-18% in the United States. Recent advances in imaging have led to the early incidental detection of hepatic cysts. Most of them are benign with no clinical significance. However, a few malignant and potentially lethal conditions can also cause cystic lesions in the liver. Clinical, radiological, and pathological correlation is crucial in accurate diagnosis and treatment. Treatment modalities for hepatic cysts range from simple fenestration to aspiration sclerotherapy, to surgical resection. In the current review, we classified the hepatic cystic lesions as developmental, neoplastic, inflammatory, post-traumatic, and miscellaneous. The unique clinical features, radiological, and histological findings, and treatment modalities of various cystic hepatic lesions are discussed in detail in the review.
MicroRNAs (miRNAs) serve as crucial regulators of gene expression and are involved in many fundamental biological processes, including cell growth, differentiation, and programmed cell death. In recent years, a growing body of evidence has highlighted the vital role of miRNAs in the pathogenesis, prognosis, and therapeutic response of glioma tumors. Given the significant increase in research in this field over the past two decades, a comprehensive bibliometric analysis is essential to evaluate scientific trends, identify key researchers, assess international collaborations, and uncover emerging topics. Such an analysis can provide a clear overview of scientific advancements and existing knowledge gaps. This study presents a systematic bibliometric review, with data collected from the Scopus database. The search strategy combined the keywords "microRNA," "Glioma," "Research Trends," and "Brain Tumor" in article titles, abstracts, and keywords. The timeframe for this review was from 2007 to 2025, and only peer-reviewed articles published in English were considered. The extracted data were analyzed based on several metrics, including the number of annual publications, research growth trends, prominent authors, national and international scientific collaborations, and keyword co-occurrence frequency. Data visualization and analysis were performed using VOSviewer software to map co-occurrence networks. The analysis of publication trends revealed that research on microRNAs in glioma showed a consistent growth from 2010 onwards, peaking in 2020 with approximately 280 published articles, but has followed a downward trend since 2021. The co-authorship analysis by country identified China and the United States as the main hubs for scientific output and international collaboration in this domain. Among authors, Galina Gabriely (Center of Neurologic Diseases, Brigham and Women's Hospital, USA), Li Gang (Department of Neurosurgery, Huashan Hospital, Fudan University, China), Wang Y (Department of Neurosurgery, Capital Medical University, China), and You Yongping (Department of Neurosurgery, Nanjing Medical University, China) were recognized as the most prolific and influential researchers based on publication volume and centrality in the co-authorship network (Gabriely et al., 2008, Li et al., 2013, Wang et al., 2025). The use of full names and institutional affiliations facilitates accurate identification of these researchers in international databases such as PubMed. The author co-authorship map revealed several active and focused research clusters. In the keyword co-occurrence analysis, terms with the highest frequency and centrality were "glioma" (n = 653), "microRNA" (n = 589), "glioblastoma" (n = 413), "mir-21" (n = 201), "migration" (n = 180), "biomarker" (n = 164), "prognosis" (n = 139), and "therapy" (n = 132), establishing them as the core concepts of the studies. Four distinct conceptual clusters were extracted: molecular and cellular mechanisms, clinical applications, signaling pathways, and comparative studies between gliomas and other cancers.To provide readers with a clearer and more comprehensive perspective of these thematic clusters, representative signature publications within each domain are highlighted. In the molecular and cellular mechanisms cluster, studies such as Chen et al. (2021) and Beylerli et al., 2022b, Beylerli et al., 2022a have elucidated how specific microRNAs regulate glioma cell proliferation, migration, invasion, and apoptosis. Within the clinical applications cluster, Tluli et al., 2023a, Tluli et al., 2023 and Mafi et al. (2022) have emphasized the diagnostic, prognostic, and therapeutic potential of microRNA signatures in glioma patients. Regarding signaling pathways, Ahmed et al. (2021) and Makowska et al. (2023) have detailed the involvement of miRNA-mediated modulation of pathways such as PI3K/AKT, p53, and Wnt/β-catenin in glioblastoma progression. Finally, in the comparative oncology cluster, studies examining shared microRNA regulatory patterns across glioma and other malignanciesincluding hepatocellular carcinoma and osteosarcoma have been reported by Faramin Lashkarian et al. (2023) and related works, illustrating the broader oncogenic and tumor-suppressive roles of microRNAs across cancer types. The inclusion of these representative publications strengthens the conceptual interpretation of the bibliometric clusters and situates the findings within the broader scientific literature. The findings of this bibliometric study indicate that research in the field of microRNAs and glioma has experienced significant growth over the last two decades, with several key countries, institutions, and authors playing a prominent role in its advancement. Emerging topics such as diagnostic biomarkers, therapeutic targets, and miRNA-related signaling pathways in glioma tumors are central to recent research. This analysis can assist researchers and scientific policymakers in identifying knowledge gaps, strengthening international collaborations, and directing future research efforts.
Accurate forecasting of tourist arrivals in major urban destinations is critical for optimizing tourism resource allocation and formulating data-driven marketing strategies. To address this need, this study presents a novel prediction framework that integrates deep learning methodologies with online search behavior data. Specifically, we propose the DTN (Dynamic Tourism Network) model, which combines Disentangled Shape and Time series Normalization (Dish-TS) with Temporal Convolutional Networks (TCN), and utilizes Baidu Index data as a key indicator of online search trends to predict tourist arrivals in Sanya, China. Empirical validation across multiple evaluation metrics demonstrates that the DTN model consistently surpasses conventional deep learning approaches, achieving statistically significant improvements in predictive accuracy for tourist volume estimation. This advancement provides a robust analytical foundation for real‑time tourism demand forecasting in destination management systems. Notably, the proposed method has been evaluated only on a popular urban tourist destination with pronounced seasonality and available Baidu Index data; its applicability to other destination types or regions where different search engines dominate therefore requires further validation.
Infra-isthmal femoral fractures in elderly patients are difficult to stabilize due to the broad metaphyseal anatomy. In the absence of consensus on optimal management, this study assesses a minimally invasive technique combining anterior clamp-assisted reduction with retrograde intramedullary nailing. A retrospective case series was conducted on ten patients treated between January 2023 and April 2024. All underwent percutaneous fracture reduction via a limited anterior approach, followed by retrograde nailing. The mean patient age was 84.9 ± 6.2 years. All fractures were caused by low-energy trauma. Six involved native bone and four were periprosthetic. All cases progressed to union without infection or wound complications. One peri-implant fracture occurred at the proximal tip of the nail and was successfully treated with overlapping plate fixation. This technique was safe and effective for infra-isthmal spiral femoral fractures in elderly, frail patients. It allowed accurate reduction, minimized soft tissue disruption, and resulted in reliable fracture healing.
NF2 gene alterations are significant drivers in a subset of renal cell carcinomas (RCCs), associated with high-grade morphology, aggressive behavior, and features overlapping with biphasic hyalinizing psammomatous RCC (BHP RCC). We report two examples of NF2-mutated RCC to advance understanding of this entity. Both were female patients in their sixth decade with incidental, solid renal masses (3.2 and 3.0 cm). Both tumors had high-grade nuclei and infiltrative growth but distinct architectures: tumor #1 showed solid nests, tubules, and a focal BHP RCC-like biphasic pattern; tumor #2 featured solid, elongated tubulotrabeculae with spindling and peritubular basement membrane material. Both had sclerotic stroma. Immunohistochemistry showed positivity for PAX8, keratin 7, vimentin, and AMACR (tumor #2: focal TFE3). Targeted next-generation sequencing identified pathogenic somatic NF2 mutations in both tumors: a nonsense mutation (c.235A>T, p.Lys79*) in tumor #1 (validated by Sanger sequencing) and a splice-site mutation (c.600-1G>A) with concurrent chromosome 22 deletion (confirming biallelic inactivation) in tumor #2. Subsequent merlin immunohistochemistry showed loss of expression. TFE3/TFEB rearrangements were absent. Patient #1 developed widespread metastases within 7 months and received immunotherapy; patient #2 remains disease-free at short-term follow-up. These tumors highlight the broad morphological heterogeneity within NF2-mutated RCC and underscore the necessity of an integrated diagnostic approach combining histology, immunohistochemistry (especially merlin loss), and molecular confirmation. Recognizing this entity is critical for accurate classification and for guiding potential therapeutic strategies, including immune checkpoint inhibitors.
Accurate maxillary positioning is critical in digital prosthodontics, particularly when transferring the spatial relationship of the maxilla to a virtual articulator. This study aimed to compare the accuracy of the two most common registration methods, three-dimensional (3D) match and UV (landmark-based point-pair registration) match, for virtual facebow transfer and to investigate whether the type of facial scanner used influences the accuracy of these registration techniques. A cone beam computed tomography (CBCT) scan of a manikin was used as the reference. The maxillary arch was scanned with three anatomical landmarks and aligned to the CBCT. A computer-aided design and computer-aided manufacturing (CAD-CAM)-printed facebow fork was affixed to the arch, followed by facial scans using the EinScan HX and iPad (10 scans each). For each scan, two registration methods (3D match and UV match) were performed, generating 40 virtual alignments. Superimpositions were completed in Exocad, and accuracy was evaluated by comparing each registration to the CBCT reference using a Python script to measure linear and angular deviations. Trueness and precision were analyzed using a linear mixed model. 3D match generally outperformed UV in both trueness and precision. The industrial scanner combined with 3D match yielded the highest accuracy, with linear trueness of 0.98 ± 0.67 mm and precision of 1.31 ± 0.79 mm (p < 0.001). With the iPad, 3D registration demonstrated significantly superior angular accuracy (p < 0.050). 3D registration provides higher accuracy than UV match, and 3D registration, in combination with an industrial-grade scanner, offers higher accuracy in virtual facebow transfer than using iPad.
Accurate maize yield prediction is critical for food security planning, particularly in sub-Saharan Africa, where maize is essential to national economies and livelihoods. This systematic review assesses the use of machine learning (ML) techniques in maize yield estimation, focusing on the methodologies, predictor variables, and results in peer-reviewed studies. The review followed the PRISMA 2021 guidelines, synthesizing 81 peer-reviewed studies published between 2014 and 2025. The analysis examined the ML algorithms, predictor variables, evaluation metrics, and methodological gaps identified in these studies. The review found a significant increase in publications after 2021, reflecting growing confidence in the application of ML for agronomic decision-support. Random Forest (49.4%), XGBoost (16.1%), and Support Vector Machines (12.4%) were the most common algorithms, with hybrid deep-learning frameworks showing superior performance. Environmental variables, remote-sensing indices, and soil properties were the most frequently used predictors. RMSE and R 2 were the primary evaluation metrics. The findings underscore the challenges of data scarcity, limited interpretability, and geographical imbalance in the research, with Africa contributing less than 25% of the studies. There is a need for open-access agricultural data systems, hybrid explainable AI frameworks, and capacity building in computational agronomy to improve the effectiveness of ML applications in maize yield prediction.
X-ray spectroscopy provides sensitive, element-specific insight into local geometric and electronic structures, but predictive first-principles simulations can be computationally expensive for large and chemically diverse molecular systems. Recent machine-learning approaches have shown promise in accelerating structure-to-spectrum prediction; however, most directly regress discretized spectral intensities and rely on hand-crafted geometric descriptors centered on the absorbing atom. Herein, we introduce a machine learning framework that encodes a detailed, environment-aware representation of the nuclear structure beyond the absorbing site. The model combines these descriptors with a physically motivated, multiscale Gaussian spectral basis whose coefficients are obtained via ridge projection, enforcing smoothness and spectral consistency. To further enhance robustness across chemical and conformational diversity, we employ a multiscale structural similarity loss that couples geometric and spectral resolution. This integrated approach yields accurate and transferable predictions across a wide range of molecular geometries and chemical environments while maintaining physical interpretability. The proposed framework establishes a physically structured and scalable route to machine-learned X-ray spectroscopy.
Developments in fusion imaging (FI) software have facilitated easy use of three-dimensional (3D) roadmaps based on preregistered computed tomography (CT) or magnetic resonance imaging (MRI) datasets for guidance of cardiac catheterizations. The aim of this study was to report the initial results from the first international prospective registry of cardiac catheterizations guided with fusion of CT and MRI datasets. A multi-center prospective registry was set up to evaluate fusion of fluoroscopic two-dimensional (2D) images and the CT- or MRI-derived 3D roadmaps for guidance of cardiac catheterizations in congenital heart disease. Fusion imaging was applied in 205 patients for guidance (n = 182) or planning (n = 23) of cardiac catheterization. Successful fusion of CT or MRI images was achieved in all cases. In 176 (96.7%) patients, 2D-3D registration was performed. In the remaining 6 patients, 3D-3D registration was utilized. Accurate initial 3D roadmap alignment was achieved in 142 (78%) patients. Seventeen (9.3%) patients required intra-procedural readjustment of the 3D roadmap due to distortion of the anatomy. Interventional procedures were performed in 137 (75.3%) patients. In 37 (20.3%) patients, catheterization was performed using only 3D guidance without additional angiography. Overall, 3D guidance with FI was deemed at least useful in 98.3% of patients and not useful or misleading in 3 (1.7%) patients. Direct 2D-3D registration of pre-catheterization CT or MRI is a safe and effective method of guidance of cardiac catheterization in selected congenital heart disease cases. In selected patients, FI facilitates percutaneous interventions without contrast angiography.
Prenatal detection of corpus callosum (CC) abnormalities is essential for assessing fetal neurodevelopment, yet conventional ultrasound diagnosis faces challenges from operator variability and suboptimal fetal positioning. We developed a novel deep learning framework CC-FocusNet that integrates automated region localization with an anatomy-aware dual-stream architecture for multi-view analysis. The model was trained on 496 cases and validated on an independent external cohort of 93 cases. We assessed both diagnostic performance and clinical interpretability through attention visualization. Our framework achieved 97.36% accuracy on the external test set. Grad-CAM++ heatmaps revealed that model attention consistently focused on clinically relevant anatomical landmarks, demonstrating strong interpretability. When integrated into clinical workflows, the AI system enhanced diagnostic accuracy and efficiency, particularly reducing misdiagnosis rates in challenging cases. This interpretable AI system provides accurate and efficient prenatal detection of CC abnormalities, offering substantial potential to support clinical decision-making and enable timely intervention for at-risk pregnancies.
Elucidating protein-ligand interactions is pivotal for understanding biological mechanisms and accelerating drug discovery. Blind docking, which identifies binding sites without prior knowledge, has become an indispensable computational strategy for analyzing the surge of protein structures generated by Cryo-EM and AI-based prediction tools like AlphaFold3. Our previous server, CB-Dock2, has been widely adopted by the global research community, averaging over 1000 daily submissions since July 2022 due to its accuracy and user-friendliness. Building on this foundation and incorporating extensive user feedback, we present CB-Dock3, a substantially enhanced platform. Key upgrades include a refined docking engine, an expanded template library, and support for diverse file formats. Benchmark evaluations on CASF-2016 demonstrate that CB-Dock3 achieves a success rate of 67.4% (RMSD ≤ 2.0 Å), representing a 10.6 percentage-point absolute improvement over its predecessor and outperforming other popular blind docking tools. Additionally, CB-Dock3 introduces critical new features driven by community needs: support for user-defined docking regions to handle large complexes, and a metal-aware protocol that explicitly retains essential metal ions and cofactors during simulation. CB-Dock3 stands as an accurate, rapid, and accessible resource for the scientific community, freely available at https://cadd.labshare.cn/cb-dock3/.
Femoral diaphyseal fractures are commonly treated with intramedullary nailing; however, malunion remains a clinically significant complication, particularly following delayed presentation, non-operative management, or treatment in resource-limited settings. Femoral diaphyseal malunion may be angular, rotational, longitudinal, or multiplanar and can result in gait disturbance, limb-length discrepancy, patellofemoral overload, secondary joint degeneration, and chronic pain. Given the substantial functional impact, corrective osteotomy is often indicated. This narrative review presents a structured approach to the assessment and management of femoral diaphyseal malunion. Key elements include comprehensive patient evaluation, detailed clinical and radiological deformity analysis, and contemporary operative strategies aimed at restoring alignment, rotation, and limb length while minimising morbidity. Successful correction requires meticulous preoperative planning and careful intraoperative execution, according to available resources. Treatment options include intramedullary nailing, plate fixation, hybrid constructs, and external fixation techniques, each with specific indications. Although circular external fixators allow accurate multiplanar correction, their prolonged use in the femur is poorly tolerated. Fixator-assisted correction has evolved to improve intraoperative control while avoiding long-term external fixation. Modern techniques such as Computer Hexapod-Assisted Orthopaedic Surgery (CHAOS) utilise a temporary intraoperative hexapod frame to achieve precise multiplanar correction, followed by definitive internal fixation within the same procedure. Adjuncts including poller screws, three-dimensional planning, and patient-specific guides further enhance accuracy and stability. Femoral diaphyseal malunion presents complex diagnostic and technical challenges. A systematic, patient-specific approach incorporating modern fixator-assisted and computer-guided techniques enables reliable restoration of alignment and rotation, improving function and quality of life.
Central lymph nodes metastasis (CLNM) is common in papillary thyroid. Microcarcinoma (PTMC). Whilst prophylactic central lymph node dissection (CLND) can prevent further CLNM, it remains controversial. An accurate model to predict CLNM is therefore necessary for patients with PTMC. This study incorporated 228 patients with general clinical information, thyroid related serological examination and ultrasound of CLNM prediction, divided into training and validation sets randomly at 7:3 ratio. Least Absolute Shrinkage and Selection Operator (LASSO) regression was used for key features screening. Eight machine learning models were developed, evaluated by cross-validation and performance comparison (the area under curve, calibration curve and decision curve analysis). Shapley Additive exPlanations (SHAP) value analysis provided the interpretability of the model. Age, gender, tumor diameter, T3, T4, TPOAb and ultrasound of CLNM prediction were identified as key features of CLNM in patients. Support Vector Machine (SVM) model with 0.783 accuracy and 0.805 specificity in validation set was considered as the most favorable performance. Age, gender and tumor diameter were the top three contributing variables in SVM model. This study established a machine learning-based framework for predicting CLNM in PTMC, with the SVM model demonstrating superior stability and clinical utility among the evaluated algorithms. While these results are preliminary, they provide a promising tool to assist in tailoring prophylactic CLND strategies, potentially reducing unnecessary surgical intervention.