The assessment of running kinematics is essential for injury prevention and rehabilitation, including anterior cruciate ligament sprains. Recent advances in computer vision have enabled the development of tools for quantifying kinematics in research and clinical settings. This study evaluated the accuracy of an OpenPifPaf-based markerless method for assessing sagittal plane kinematics of the ankle, knee, and hip during treadmill running using smartphone video footage and examined the impact of clothing on the results. Thirty healthy participants ran at 2.5 and 3.6 m/s under two conditions: (1) wearing minimal clothing with markers to record kinematics by using both a smartphone and a marker-based system, and (2) wearing usual running clothes and recording kinematics by only using a smartphone. Joint angles, averaged over 20 cycles, were analysed using SPM1D and RMSE. The markerless method produced kinematic waveforms closely matching the marker-based results, with RMSEs of 5.6° (hip), 3.5° (ankle), and 2.9° (knee), despite some significant differences identified by SPM1D. Clothing had minimal impact, with RMSEs under 2.8° for all joints. These findings highlight the potential of the OpenPifPaf-based markerless method as an accessible, simple, and reliable tool for assessing running kinematics, even in natural attire, for research and clinical applications.

Medical Entity Linking (MEL) refers to the task of automatically detecting and codifying medical concepts, which is an important preprocessing step for exploiting unstructured medical reports. SapBERT is an efficient MEL method for the English language. Unfortunately, its cross-language counterpart underperforms in French due to the lack of training data. To address this limitation, this paper explores the possibility of training SapBERT on machine-generated French translations of the UMLS Metathesaurus. Evaluation on the QuaeroFrenchMed benchmark demonstrates that this approach outperforms Cross-lingual SapBERT for SNOMED-CT MEL on the QuaeroFrenchMed benchmark.

Background: Critical-size bone defects (CSBDs) pose significant challenges in clinical orthopaedics and traumatology. Developing reliable preclinical models that accurately simulate human conditions is crucial for translational research. This study addresses the need for a reliable preclinical model by evaluating the design and efficacy of a custom-made 3D-printed intramedullary nail (IMN) specifically for CSBDs in minipigs. The study aims to answer the following questions: Can a custom-made 3D-printed IMN be designed for femoral osteosynthesis in minipigs? Does the use of the custom-made IMN result in consistent and reproducible surgical procedure, particularly in the creation and fixation of CSBDs? Can the custom-made IMN effectively treat and promote bone consolidation of CSBDs?Hypothesis: The custom-made 3D-printed IMN can be designed to effectively create, fix and treat CSBDs in minipigs, resulting in consistent surgical outcomes.Materials and Methods: The IMN was designed based on CT scans of minipig femurs, considering factors such as femoral curvature, length, and medullary canal diameters. It was 3D-printed in titanium and evaluated through both in vitro and in vivo testing. Female Aachen minipigs underwent bilateral femoral surgeries to create and fix CSBDs using the custom-made IMN. Post-operative follow-up included X-rays and CT scans every 2 weeks, with manual examination of explanted femurs to assess consolidation and mechanical stability after 3 months.Results: The custom-made IMN effectively fitted the minipig femoral anatomy and facilitated reproducible surgical outcomes. Symmetric double osteotomies were successfully performed, and allografts showed minimal morphological discrepancies. However, proximal fixation faced challenges, leading to non-union in several cases, while most distal osteotomy sites achieved stable consolidation.Discussion: The custom-made 3D-printed IMN demonstrated potential in modelling and treating CSBDs in minipigs. While the design effectively supported distal bone healing, issues with proximal fixation highlight the need for further refinements. Potential improvements include better screw placement, additional mechanical support, and adaptations such as a reduction clamp or a cephalic screw to enhance stability and distribute forces more effectively.

Background: Emergency departments face increasing pressures from staff shortages, patient surges, and administrative burdens. While large language models (LLMs) show promise in clinical support, their deployment in emergency medicine presents technical and regulatory challenges. Previous studies often relied on simplistic evaluations using public datasets, overlooking real-world complexities and data privacy concerns. Methods: At a tertiary emergency department, we retrieved 79 consecutive cases during a peak 24-hour period constituting a siloed dataset. We evaluated six pipelines combining open- and closed-source embedding models (text-embedding- ada-002 and MXBAI) with foundational models (GPT-4, Llama3, and Qwen2), grounded through retrieval-augmented generation with emergency medicine textbooks. The models’ top-five diagnostic predictions on early clinical data were compared against reference diagnoses established through expert consensus based on complete clinical data. Outcomes included diagnostic inclusion rate, ranking performance, and citation sourcing capabilities. Results: All pipelines showed comparable diagnostic inclusion rates (62.03-72.15%) without significant differences in pairwise comparisons. Case characteristics, rather than model combinations, significantly influenced predictive diagnostic performance. Cases with specific diagnoses were significantly more diagnosed versus unspecific ones (85.53% vs. 31.41%, p<0.001), as did surgical versus medical cases (79.49% vs. 56.25%, p<0.001). Open-source foundational models demonstrated superior sourcing capabilities compared to GPT-4-based combinations (OR: 33.92 to ∞, p<1.4e-12), with MBXAI/Qwen2 achieving perfect sourcing. Conclusion: Open and closed-source LLMs showed promising and comparable predictive diagnostic performance in a real-world emergency setting when evaluated on siloed data. Case characteristics emerged as the primary determinant of performance, suggesting that current limitations reflect AI alignment fundamental challenges in medical reasoning rather than model-specific constraints. Open-source models’ demonstrated superior sourcing capabilities—a critical advantage for interpretability. Continued research exploring larger-scale, multi-centric efforts, including real-time applications and human-computer interactions, as well as real- world clinical benchmarking and sourcing verification, will be key to delineating the full potential of grounded LLM-driven diagnostic assistance in emergency medicine.