Wencai Liu

Mitochondrial dysfunction and consequent bioenergetic collapse in cartilage progenitor cells (CPCs), driven by excessive mitochondrial reactive oxygen species (ROS), constitute a fundamental barrier to endogenous cartilage regeneration and accelerate osteoarthritis (OA) progression. Accordingly, precise modulation of mitochondrial ROS is required to restore mitochondrial metabolic homeostasis. However, conventional antioxidant agents such as N-acetylcysteine (NAC) lack cell specificity and organelle-level precision, and exhibit limited bioavailability, thereby restricting their capacity to effectively reestablish mitochondrial metabolic homeostasis. Here, we engineer zero-dimensional (0D) bioinspired nanoassemblies, CPC membrane-coated (3-carboxypropyl)triphenylphosphonium bromide-functionalized NAC-derived carbon quantum dots (CM@TQDs), with capabilities for homotypic recognition and mitochondria-targeted metabolic reprogramming. Subsequently, CM@TQDs are encapsulated within ROS/pH-responsive hydrogel microspheres (HGCT), permitting inflammation-triggered release within the OA joint. Upon HGCT-mediated delivery and cellular internalization, the 0D nanoassemblies accumulate in mitochondria in a membrane potential–dependent manner, enhancing local mitochondrial bioavailability. Mechanistically, HGCT effectively scavenges mitochondrial ROS, restores oxidative phosphorylation, reestablishes tricarboxylic acid cycle flux, and suppresses aberrant glycolytic dependence. This metabolic restoration reactivates PI3K/AKT signaling, mitigates apoptosis and ferroptosis, and promotes CPC proliferation and chondrogenic differentiation. In vivo, HGCT attenuates synovial inflammation and enhances cartilage regeneration, markedly inhibiting OA progression. Collectively, this work establishes a nanobiomimetic therapeutic platform capable of achieving hierarchical precision from cell-specific targeting to organelle-level metabolic regulation, offering a promising strategy for nanoscale bioenergetic intervention in degenerative diseases.

Structural retear remains common after rotator cuff repair. Persistent inflammation and metabolic dysfunction at the enthesis may impair healing, yet current augmentations rarely address this immunometabolic state. Exploratory single‐cell profiling shows a higher proportion of M1‐like cells among recovered macrophages in injured supraspinatus than in non‐injured semitendinosus specimens (three donors per group). Metabolomics of polarized macrophages identifies glutathione redox imbalance and reduced arginine and proline abundance in the M1 state, supporting redox regulation and arginine metabolism as complementary targets. In ACGQD@pPLG, citrulline‐functionalized, glutathione‐derived carbon dots are encapsulated in an H ₂ O ₂ ‐responsive membrane coating from ultraviolet‐treated tendon‐derived stem cells and embedded in aligned core–shell electrospun fibers. In macrophages, the patch activates nuclear factor erythroid 2‐related factor 2 (NRF2) signaling and increases argininosuccinate synthetase 1 (ASS1) expression and arginine‐associated metabolite abundance. ASS1 knockdown attenuates CD206 induction, whereas NRF2 knockdown reduces HO‐1 and NQO1 induction. In rats, the patch improves enthesis collagen organization and increases ultimate stress, stiffness, and load to failure to 1.6–2.5 times repair‐only values at 8 weeks. Rabbit enthesis histology also improves at 10 weeks. These results support coordinated modulation of macrophage redox signaling and arginine metabolism as a local strategy for tendon‐to‐bone repair.

In terms of precise diagnosis and risk assessment, this special issue highlights pivotal progress from morphological judgment to quantitative prediction. Artificial intelligence models based on multimodal imaging and clinical data have significantly improved diagnostic accuracy and prognostic prediction. For instance, KneeXNet integrates graph convolutional networks (GCNs) with a multi-scale feature fusion module to enable automated, high-precision MRI diagnosis of knee injuries (Sun et Overall, the 24 research papers in this special issue cover a complete research chain, ranging from competitive sports injuries to age-related fragility fractures, from surgical techniques to rehabilitation protocols, and from fundamental mechanisms to clinical translation. They fully embody the interdisciplinary innovation philosophy of clinical demand-oriented, engineering technology-enabled, and functional recovery-targeted. These findings not only address numerous current clinical dilemmas but also lay a solid foundation for intelligent diagnosis, digital orthopedics, regenerative repair and smart rehabilitation. They also demonstrate that medical and engineering innovations will drive earlier, smarter, more precise, more efficient and personalized prevention and treatment in sports medicine. Moving forward, the diagnosis and management of sports injuries will continue to evolve toward full-process digitalization, normalized regenerative medicine, interpretable artificial intelligence, and integrated injury prevention and health maintenance.We hope readers can gain more than just technological progress, but also a clear perspective on how to solve practical clinical problems and foster broader interdisciplinary collaboration through this issue. Ultimately, we aim to enable all individuals affected by sports injuries-whether elite athletes or the general public-to truly benefit from these innovations, allowing them to exercise safely, recover efficiently, and return to their cherished lives with dignity. This embodies the original mission of sports medicine and fuels our ongoing research and endeavor.

Operative duration reflects surgical complexity and is valuable for perioperative planning. To predict prolonged operative time in posterior spinal deformity correction, we developed a preoperative machine‐learning model. Data from 162 patients across 11 hospitals were analyzed. Ten algorithms were trained and compared, and the best‐performing model was implemented as a web calculator. Boruta identified seven key predictors: interval from illness to surgery, fibrinogen, white blood cell count, activated partial thromboplastin time, American Society of Anesthesiologists class, albumin, and age. LightGBM showed the most consistent performance, achieving an AUC of approximately 0.70. SHAP analyses improved interpretability, and calibration and decision‐curve analyses indicated good calibration and net clinical benefit. We developed an interpretable, internally validated model to estimate operative duration in posterior spinal deformity surgery. The web calculator may support individualized planning and more efficient operating‐room resource allocation. Clinical Trial Registration: NCT05867732

Background: Elbow stiffness is a condition that causes mobility dysfunction and severe suffering in patients. Patients with posttraumatic elbow stiffness who undergo open elbow arthrolysis are at risk for poor postoperative flexion-extension outcome, which leads us to suspect that these patients possess a unique “elbow stiffness predisposition.” Purpose: To develop an innovative predictive model that combines clinical and laboratory indicators to forecast poor flexion-extension outcomes after open elbow arthrolysis, thereby interpreting the “elbow stiffness predisposition.” Study Design: Cohort study; Level of evidence, 3. Methods: Patients who underwent open elbow arthrolysis between 2019 and 2022 were selected for model training and validation (n = 254), while those who underwent open elbow arthrolysis between 2016 and 2017 served as a test set (n = 35). The study assessed 19 clinical features and 58 laboratory parameters. A comparative analysis of several machine learning models—logistic regression, Naive Bayes, decision trees, random forest, gradient boosting, and XGBoost—was performed to identify the most effective approach. SHapley Additive exPlanations (SHAP) were employed to prioritize the key factors. Results: Using univariate analysis and the least absolute shrinkage and selection operator (LASSO) regression, 14 key variables were selected for inclusion in the model. The XGBoost model demonstrated superior performance, reaching an area under the curve of 0.909 on the test dataset. The most important indicator identified by LASSO was alkaline phosphatase. Indicators ranked by SHAP were lipoprotein(a), alkaline phosphatase, visual analog scale score, serum calcium, basophil count, serum sodium, previous arthrolysis, alanine aminotransferase, blood glucose, preoperative elbow range of motion, Cystatin C, uric acid, tobacco use, and serum cholinesterase. Conclusion: We have successfully developed a machine learning model using 14 key indicators to predict poor flexion-extension outcomes, which preliminarily explains the “elbow stiffness predisposition.”

Stroke is a serious condition associated with high rates of incidence and mortality. Timely and accurate risk assessment is critical for improving prevention and clinical intervention. This study investigates the application of machine learning models in evaluating the risk of stroke. The data for this study were drawn from the National Health and Nutrition Examination Survey, covering the years 1999 to 2002. To construct predictive models and select relevant variables, 3 approaches were applied: LASSO regression combined with stepwise selection, random forest, and the boruta algorithm in conjunction with LASSO regression. Model performance was evaluated through multiple metrics, including receiver operating characteristic and precision–recall curves, as well as calibration and decision curve analyses. The study analyzed data from 9922 participants, among whom 358 had a history of stroke. Key predictors were identified through a combination of LASSO and stepwise regression, producing a model with strong discriminative performance (area under the curve [AUC] = 0.843). In comparison, the random forest approach selected fewer predictors and showed lower predictive accuracy (AUC = 0.612). A model developed using the boruta algorithm followed by LASSO regression achieved a similarly high level of performance (AUC = 0.828). These findings illustrate how different variable selection methods can influence the predictive accuracy of the resulting models. The machine learning model, which is created using the National Health and Nutrition Examination Survey database, and serves as a reliable means of predicting stroke risk. There is every possibility to set up accurate predictive models with the help of different variable selection techniques and modeling methods that have high accuracy and clinical value.

Neurodegenerative diseases, such as Alzheimer's disease (AD), Parkinson's disease (PD), and amyotrophic lateral sclerosis, impose an ever-growing burden on patients, families, and healthcare systems worldwide (1). Despite remarkable advances in neuroscience, early diagnosis and effective therapeutic monitoring remain challenging. In recent years, artificial intelligence (AI) and radiomics have emerged as powerful tools to transform the landscape of neurodegenerative disease research (2). AI-driven algorithms can uncover subtle imaging patterns beyond human perception, while radiomics enables quantitative extraction of high-dimensional features from neuroimaging data. Together, these approaches hold promise for improving diagnostic accuracy, predicting disease progression, and guiding personalized treatment strategies (3). This Research Topic brings together contributions that highlight the latest methodological innovations and clinical applications at the intersection of AI, radiomics, and neurodegenerative disorders, aiming to foster dialogue between researchers, clinicians, and data scientists to accelerate translation into clinical practice. The collection "Frontier Research on Artificial Intelligence and Radiomics in Neurodegenerative Diseases" features 18 studies applying artificial intelligence and radiomics to neurodegenerative disorders, including 2 reviews and 16 original research articles. Several of these works highlight the potential of multimodal AI approaches in advancing the diagnosis and management of PD. Zhou et al. ("A predictive model of Parkinsonian brain aging based on brain imaging features") demonstrated that specific structural indices such as temporal lobe gyrification and thalamic volume significantly contribute to brain age prediction in PD. Ye et al. ("Individualized diagnosis of Parkinson's disease based on multivariate magnetic resonance imaging radiomics and clinical indexes") showed that integrating radiomic features with clinical data yields superior diagnostic accuracy. Complementing these imaging-based approaches, Yang et al. ("Identification of Parkinson's disease using MRI and genetic data from the PPMI cohort: an improved machine learning fusion approach") improved early PD diagnosis through MRI–genetic data fusion, while Xu et al. ("Non-invasive detection of Parkinson's disease based on speech analysis and interpretable machine learning") established speech-based machine learning models that achieved high predictive performance with interpretable features. Together, these studies underscore the promise of AI-driven multimodal strategies—spanning imaging, genetics, and speech—for advancing early, non-invasive, and precise diagnosis of PD. Advances of AI and radiomics in AD and related cognitive disorders are demonstrated by several studies included in this collection. Xu et al. ("Accurate and robust prediction of Amyloid-β brain deposition from plasma biomarkers and clinical information using machine learning") developed and validated a low-cost, plasma biomarker–based model across the Alzheimer's Disease Neuroimaging Initiative (ADNI) and the Chinese National Twin Network (CNTN) cohorts, achieving high accuracy [area under the receiver operating characteristic curve (AUC) up to 0.95] with only five core features, offering a potential alternative to PET imaging. Han et al. ("Exploring the changes in functional connectivity of the limbic system in Patients with amnestic mild cognitive impairment treated by acupuncture based on fMRI") revealed abnormal limbic connectivity in amnestic mild cognitive impairment and demonstrated that acupuncture could modulate these connections, providing imaging evidence for its therapeutic mechanism. Aresta et al. ("AI-based staging, causal hypothesis and progression of subjects at risk of Alzheimer's disease: a multicenter study") showed that AI tools achieve staging results consistent with expert assessments and accurately align with biomarker-based causal hypotheses in memory clinic workflows. Mao et al. ("Transcriptional patterns of brain structural abnormalities in CSVD-related cognitive impairment") combined neuroimaging and transcriptomics to identify gray matter volume loss in the temporal lobe and thalamus, linking these alterations with 1,580 gene expression profiles and cognitive decline. Zhen et al. ("Association between Alzheimer's disease pathologic products and age and a pathologic product-based diagnostic model for Alzheimer's disease") analyzed large-scale ADNI data, revealing opposite age-related trends of Tau and Aβ levels, and built a high-performance early diagnostic model (AUC = 0.959) using machine learning. Collectively, these studies provide compelling evidence that AI-driven multimodal approaches—ranging from plasma biomarkers and neuroimaging to transcriptomics—hold great promise for the early detection, mechanistic understanding, and precision diagnosis of AD and related disorders. Beyond Parkinson's and Alzheimer's disease, several studies included here broaden the application of AI and radiomics to other neurological disorders. Fah et al. ("Comparing machine learning classifier models in discriminating cognitively unimpaired older adults from three clinical cohorts in the Alzheimer's disease spectrum: demonstration analyses in the COMPASS-ND study"), systematically compared multiple machine learning classifiers and explainable AI techniques, finding Random Forest and Gradient Boosted Trees to be the most robust for near-cohort classification, while SHAP outperformed LIME in interpretability. Huang et al. ("Knowledge map of artificial intelligence in neurodegenerative diseases: a decade-long bibliometric and visualization study"), analyzed 1,921 publications from the past decade, identifying key journals, institutions, and authors, and highlighting future research directions such as multimodal precision diagnosis, mechanistic insights,...

The aim of this study was to compare the performance of 4 machine learning models—Lasso regression model, random forest model, Boruta algorithm model, and the Boruta algorithm combined with Lasso regression—in predicting stroke risk among hypertensive patients. The study evaluated the strengths and weaknesses of each model to provide a more clinically valuable prediction model for stroke risk. The study included 3472 hypertensive patients, of which 312 had experienced a stroke, and 3160 had not. Various health indicators were analyzed using Lasso regression, random forest, Boruta algorithm, and the Boruta algorithm combined with Lasso regression. Model performance was evaluated based on the area under the curve (AUC) of the receiver operating characteristic curve, the precision–recall curve, calibration curve, and decision curve analysis to assess classification ability, precision, calibration, and clinical benefit. The Lasso regression and Boruta algorithm models both have an AUC of 0.716, making them the best-performing models in terms of classification ability. The Boruta algorithm combined with Lasso regression model has an AUC of 0.705, slightly lower than the previous 2 models but still shows good predictive capability, with better interpretability due to feature selection. The random forest model has an AUC of 0.626, which is the lowest among the models, indicating weaker classification performance compared to the others. Among the 4 models, the Lasso regression model and Boruta algorithm model performed similarly in terms of classification ability, both demonstrating moderate predictive power, while the random forest model performed relatively poorly. The Boruta combined with Lasso regression model was precise in variable selection but had limited clinical utility. Therefore, the Lasso regression model appears to be the most balanced in predicting stroke risk and is the recommended model based on this study.