GuoYou Wang

Background: Generative language and multimodal foundation models could connect clinical text, bone imaging, and longitudinal records in osteoporosis care. Whether published evidence supports changing clinical management remains unknown.,Methods: We conducted a PubMed/PMC-based evidence map and reproducibility audit (January 1, 2022–August 3, 2026). Eligible studies evaluated generative large language models, vision-language models, or multimodal foundation models in osteoporosis, osteopenia, bone mineral density, or fragility fractures. Clinical readiness was classified from R1 (output capability) to R4 (prospective live comparative workflow). Public reruns required seven prespecified compatibility criteria. Quality, threshold sensitivity, and reproducibility were assessed.,Findings: Of 50 unique records, 21 target studies were included: 14 R1, five R2, one R3, and one R4. No study evaluated an osteoporosis-specific vision-language or multimodal model on paired bone images and reports with treatment or fracture endpoints. The R3 same-centre temporal validation showed zero-shot F1 0.857–0.871 for bone cement leakage but 0.309 for new vertebral fracture (MCC 0.044). The R4 formative randomized trial (100 randomized; 88 analysed) of an education chatbot improved knowledge (80 vs 75; p=0.01) and reduced nurse time (5 vs 23 min; p<0.001), but did not assess screening, treatment initiation, fracture, function, quality of life, or harm. No study reported improved live clinical outcomes; none of six rerun candidates passed all gates.,Interpretation: The prespecified practice-change threshold was not met. Evidence supports supervised narrow-task investigation, not autonomous osteoporosis or fragility-fracture care. Future studies should freeze models, validate across institutions, and measure treatment initiation, fractures, harms, and implementation costs.

Background:The biomechanical consequences of L4–L5 transforaminal lumbar interbody fusion (TLIF) versus decompression alone remain quantitatively under‐resolved, particularly regarding adjacent‐segment load transfer. Finite‐element (FE) modeling can estimate internal variables such as intradiscal pressure (IDP) and facet contact forces, but published comparisons often lack transparent validation, morphology conditioning, or explicit linkage to patient‐specific imaging. We therefore designed a validation‐gated FE framework to compare these two surgical strategies using MRI‐derived lumbar phenotypes, with pre‐specified primary endpoints at L3–L4 and L5–S1.Methods:We audited two public MRI repositories (AISSLab, LSMAPQR) for segmentation, grading, and geometry extraction; external benchmarks from the FE literature were catalogued as validation comparators. The primary AISSLab sagittal MRI dataset (2.04 GB) could not be downloaded at usable speed, precluding segmentation‐based FE execution in the current phase. Concurrently, we analysed seven independent public transcriptomic datasets (GEO) to localise degeneration‐ and loading‐associated cell states, using five pre‐defined gene signatures (mechanotransduction, ECM anabolism/catabolism, inflammation, senescence/stress). Analyses spanned single‐donor and multi‐patient single‐cell RNA‐seq, 18‐donor and 4‐donor microarrays, severity‐stratified bulk profiles, porcine annulus RNA‐seq (loading model), and mouse spatial transcriptomics (negative control). No patient‐level mechanogenomic linkage was assumed or enforced. All transcriptomic tests were corrected for multiple comparisons (Benjamini–Hochberg); machine‐learning stress tests and signed‐context audits were performed for sensitivity.

Existing finite element (FE) models of the foot-ankle complex frequently oversimplify soft-tissue structures, compromising accuracy for clinical assessment. This study aimed to develop and comprehensively validate a high-fidelity three-dimensional FE model through systematic integration of all key anatomical components. Computed tomography data from one healthy male volunteer (age 25 years, height 173 cm, weight 60 kg) were used to reconstruct a geometric model comprising 28 bones, 15 cartilages, 50 ligaments, 19 bony junctions, and 36 tendons as three-dimensional solid elements. Tissue-specific linear elastic material properties were assigned from published experimental data. The model was validated against experimental measurements across four loading scenarios: static stance (600 N axial load), range of motion (10 N·m/4 N·m moments), stability testing (67 N drawer forces), and gait cycle simulation. Under static loading, model-predicted tibiotalar joint contact stress demonstrated excellent agreement with experimental measurements (maximum stress: 2.81 MPa vs. 2.92 MPa experimental; relative error 3.9%). The model accurately reproduced physiological joint kinematics (plantarflexion 10.18°, dorsiflexion 11.47° under 10 N·m) and captured the characteristic mechanical behavior of lateral ankle instability (lateral drawer displacement 1.45 mm vs. anterior 0.45 mm, posterior 0.60 mm, medial 0.84 mm). Gait simulation revealed dynamic stress variation consistent with established biomechanical principles (peak stress 6.62 MPa at mid-stance). This validated platform provides a foundation for patient-specific biomechanical assessment in foot-ankle orthopaedics.

Background Osteoarthritis (OA) is a common chronic inflammatory orthopedic disease, and ferroptosis (a novel OA-associated cell death mechanism) remains insufficiently understood. Saikosaponin A (a traditional Chinese medicine monomer) shows OA therapeutic effects, but its role in OA-related ferroptosis is unclear. This study aimed to elucidate its mechanism via bioinformatics, single-cell analysis, and in vitro/in vivo experiments. Methods OA expression profiles (GEO: GSE1919, GSE55235, GSE55475) and ferroptosis genes (FerrDb) were collected. Machine learning identified OA ferroptosis hub genes and pathways; scRNA-seq (GSE169454, Seurat/SCENIC) defined chondrocyte subsets. Network pharmacology and experiments (human primary chondrocytes, MIA-KOA rats) validated targets. Results Bioinformatics and network pharmacology analysis revealed that Saikosaponin A mediates chondrocyte ferroptosis via the TNF signaling pathway and FoxO signaling pathway through the regulation of MAPK14, MAPK8, SOX2, IL1B, NR4A1, and PTGS2. Subsequent in vitro and in vivo experiments confirmed that Saikosaponin A can alleviate the progression of KOA by regulating chondrocyte ferroptosis through the FoxO signaling pathway. Finally, single-cell combined bioinformatics analysis further expanded the relationship between core genes and chondrocyte subtypes. Conclusion This study systematically elucidated that Saikosaponin A can regulate FoxO signaling pathway to modulate chondrocyte ferroptosis and alleviate the progression of KOA, and that MAPK14, MAPK8, SOX2, IL1B, NR4A1, and PTGS2 are involved in the regulation of chondrocyte phenotype. Key Points • Integrated multi-omics approach identifies the FoxO-ferroptosis axis in OA. Combining bioinformatics, network pharmacology, and single-cell sequencing revealed the crucial role of the FoxO signaling pathway in regulating chondrocyte ferroptosis during osteoarthritis (OA) progression. • Saikosaponin A alleviates chondrocyte ferroptosis via the FoxO pathway. Both in vitro and in vivo experiments demonstrate that the natural compound Saikosaponin A inhibits erastin-induced ferroptosis in chondrocytes by targeting key components of the FoxO signaling pathway (FoxO1 and PI3K). • MAPK14, MAPK8, SOX2, IL1B, NR4A1, and PTGS2 are key regulatory genes. Multi-dataset analysis pinpointed six hub genes linking ferroptosis to OA, which are involved in Saikosaponin A’s therapeutic mechanism. • Single-cell analysis maps ferroptosis hub genes to specific chondrocyte subtypes. The study unveils the distinct distribution of ferroptosis-related regulators (e.g., high PTGS2 in ProCs and RegCs) across chondrocyte subpopulations, providing cellular-level insights for targeted therapy.

To test whether brief bedside photoplethysmography-derived heart-rate variability (PPG-HRV) provides objective information about pain, sleep, and affect in adults with chronic pain, and whether sleep moderates the pain–emotion association. We conducted a cross-sectional study of consecutive inpatients with chronic pain (duration ≥ 3 months) admitted to an orthopedics/traumatology service from March 31 to September 28, 2025. Adults (≥ 18 years; right-handed) were excluded for severe internal disease or acute trauma, major psychiatric disorders, marked cognitive impairment, or recent use of anxiolytics/hypnotics. At admission, 5-min fingertip PPG-HRV was recorded under standardized bedside conditions, with artifact correction, stationarity checking, and quality-control screening applied before analysis. Outcomes were the Brief Pain Inventory–Short Form (BPI-SF), Hamilton Depression Rating Scale (HAMD-24), Hamilton Anxiety Rating Scale (HAMA-14), and a composite “sleep problems” index computed as the mean of z-scored HAMD insomnia items and the BPI-SF sleep-interference item. Analyses included correlations, cross-validated random-forest regression incorporating logarithmically transformed HF power (lnHF), and Hayes’ PROCESS Model 1 testing sleep as a moderator of the pain→emotion pathway. A total of 172 patients were enrolled (37.2% men; mean age 59.2 years). lnHF was significantly associated with HFnorm, pain, sleep disturbance, and depression, but not anxiety. Random-forest models showed the strongest predictive performance for sleep disturbance (R²_CV = 0.844), followed by depression (R²_CV = 0.696) and pain-related burden (R²_CV = 0.506), whereas anxiety showed limited predictive performance (R²_CV = 0.009). HFnorm and lnHF emerged as recurrently important predictors of pain, sleep disturbance, and depressive symptoms. In moderation analyses, pain×sleep interactions were significant for anxiety (model R² = 0.465; β = 0.159; ΔR² = 0.026; p < 0.01) and depression (model R² = 0.513; β = 0.252; ΔR² = 0.066; p < 0.001). Brief bedside PPG-HRV, particularly HF-related indices such as lnHF and HFnorm, provides auxiliary physiological information related to pain, sleep disturbance, and depressive symptoms in hospitalized adults with chronic pain. Sleep disturbance amplified the association between pain and emotional symptoms, supporting the clinical relevance of the pain–sleep–emotion pathway. HRV is not suitable for diagnosis or stand-alone decision-making, but may serve as a practical adjunctive marker for risk stratification and symptom monitoring.

Introduction: Modified Yaotong Tang (MYT) is used as an adjunctive postoperative herbal decoction in traditional Chinese medicine practice for lumbar disorders, but its association with early recovery after unilateral biportal endoscopy (UBE) and the reproducibility of its clinical preparation remain insufficiently characterized. Methods: This single-center retrospective comparative cohort study screened 235 patients with symptomatic single-level lumbar disc herniation (LDH) who underwent surgical treatment between September 2022 and September 2025. Of these, 182 met the eligibility criteria, including 122 patients in the primary UBE analytic cohort (62 UBE + MYT and 60 UBE alone) and 60 patients in a transforaminal lumbar interbody fusion cohort retained only as a descriptive external benchmark. The primary UBE comparison was adjusted using inverse probability of treatment weighting (IPTW). Results: Compared with UBE alone, the UBE + MYT cohort showed lower IPTW-adjusted Visual Analog Scale scores at postoperative day (POD) 3 (3.88 vs. 4.67; mean difference, −0.79; 95% CI, −1.06 to −0.53; P < 0.001) and POD7 (2.85 vs. 3.73; mean difference, −0.88; 95% CI, −1.14 to −0.62; P < 0.001). IPTW-adjusted IL-6 and CRP levels were lower during the first postoperative week, and follow-up Japanese Orthopaedic Association scores were higher in the UBE + MYT cohort. MYT administration was verified using de-identified medical, pharmacy/decoction-room, nursing, and follow-up records. No Clavien-Dindo grade III or higher adverse events were observed. Discussion: Postoperative MYT use was associated with lower early postoperative pain scores, lower inflammatory biomarker levels, and better subsequent functional recovery after UBE for LDH. No severe safety signal was observed, but causality and rare herb-related safety events require prospective validation.

Anterior cruciate ligament (ACL) reconstruction conventionally prioritizes immediate mechanical stability; however, it does not fully prevent long-term post-traumatic osteoarthritis, residual neuromuscular deficits, or incomplete biological integration. This limitation reflects the gap between structural graft substitution and true functional regeneration, including insufficient ligament-bone interface healing, delayed graft vascularization and remodeling, loss of proprioceptive afferents, and the demanding intra-articular environment characterized by synovial fluid washout and high cyclic mechanical loading. To bridge this gap between conceptual biomaterials and clinical orthopedic practice, this review critically deconstructs the physiological bottlenecks of current ACL repair and reconstruction and outlines a translational roadmap for next-generation biointelligent ligament grafts. We evaluate emerging strategies including: (1) cross-scale hierarchical biomaterials designed to mitigate graft-tunnel micromotion, interfacial shear, and tunnel widening; (2) spatially retained and cell-responsive biological release systems that reduce the risk of uncontrolled intra-articular diffusion and ectopic ossification; (3) orthobiologic adjuncts, including platelet-rich plasma and platelet-rich fibrin, as clinically accessible but still insufficiently standardized tools for biological augmentation; (4) piezoelectric and mechanically shielded niches aimed at supporting proprioceptive neuralization; and (5) fluid-stable implantable sensors coupled with data-driven rehabilitation and digital twin concepts. By integrating biomimetic design, immune modulation, sensory restoration, and technology-readiness considerations, this review provides a clinically oriented framework for shifting ACL therapy from passive mechanical replacement toward active neuro-mechanical regeneration.