Chondrogenic differentiated mesenchymal stromal cells (MSCs) are a promising cell source for articular cartilage repair. This study was undertaken to determine the effectiveness of two three-dimensional (3D) culture systems for chondrogenic MSC differentiation in comparison to primary chondrocytes and to assess the effect of Interleukin (IL)-10 and Tumor Necrosis Factor (TNF)α on chondrogenesis by MSCs in 3D high-density (H-D) culture. MSCs were isolated from femur spongiosa, characterized using a set of typical markers and introduced in scaffold-free H-D cultures or non-woven polyglycolic acid (PGA) scaffolds for chondrogenic differentiation. H-D cultures were stimulated with recombinant IL-10, TNFα, TNFα + IL-10 or remained untreated. Gene and protein expression of type II collagen, aggrecan, sox9 and TNFα were examined. MSCs expressed typical cell surface markers and revealed multipotency. Chondrogenic differentiated cells expressed cartilage-specific markers in both culture systems but to a lower extent when compared with articular chondrocytes. Chondrogenesis was more pronounced in PGA compared with H-D culture. IL-10 and/or TNFα did not impair the chondrogenic differentiation of MSCs. Moreover, in most of the investigated samples, despite not reaching significance level, IL-10 had a stimulatory effect on the type II collagen, aggrecan and TNFα expression when compared with the respective controls.

Background: Large language models are increasingly applied in health care for documentation, patient education, and clinical decision support. However, their factual reliability can be compromised by hallucinations and a lack of source traceability. Retrieval-augmented generation (RAG) enhances response accuracy by combining generative models with document retrieval mechanisms. While promising in medical contexts, RAG-based systems remain underexplored in orthopedic and trauma surgery patient education, particularly in non-English settings. Objective: This study aimed to develop and evaluate a RAG-based chatbot that provides German-language, evidence-based information on common orthopedic conditions. We assessed the system’s performance in terms of response accuracy, contextual precision, and alignment with retrieved sources. In addition, we examined user satisfaction, usability, and perceived trustworthiness. Methods: The chatbot integrated OpenAI’s GPT language model with a Qdrant vector database for semantic search. Its corpus consisted of 899 curated German-language documents, including national orthopedic guidelines and patient education content from the Orthinform platform of the German Society of Orthopedics and Trauma Surgery. After preprocessing, the data were segmented into 18,197 retrievable chunks. Evaluation occurred in two phases: (1) human validation by 30 participants (orthopedic specialists, medical students, and nonmedical users), who rated 12 standardized chatbot responses using a 5-point Likert scale, and (2) automated evaluation of 100 synthetic queries using the Retrieval-Augmented Generation Assessment Scale, measuring answer relevancy, contextual precision, and faithfulness. A permanent disclaimer indicated that the chatbot provides general information only and is not intended for diagnosis or treatment decisions. Results: Human ratings indicated high perceived quality for accuracy (mean 4.55, SD 0.45), helpfulness (mean 4.61, SD 0.57), ease of use (), and clarity (mean 4.77, ), while trust scored slightly lower (mean 4.23, ). Retrieval-Augmented Generation Assessment Scale evaluation confirmed strong technical performance for answer relevancy (mean 0.864, SD 0.223), contextual precision (mean 0.891, SD 0.201), and faithfulness (mean 0.853, SD 0.171). Performance was highest for knee and back-related topics and lower for hip-related queries (eg, gluteal tendinopathy), which showed elevated error rates in differential diagnosis. Conclusions: The chatbot demonstrated strong performance in delivering orthopedic patient education through an RAG framework. Its deployment on the national Orthinform platform has led to more than 9500 real-world user interactions, supporting its relevance and acceptance. Future improvements should focus on expanding domain coverage, enhancing retrieval precision, and integrating multimodal content and advanced RAG techniques to improve robustness and safety in patient-facing apps.

Purpose: This single-center study aimed to develop a convolutional neural network to segment multiple consecutive axial magnetic resonance imaging (MRI) slices of the lumbar spinal muscles of patients with lower back pain and automatically classify fatty muscle degeneration. Methods: We developed a fully connected deep convolutional neural network (CNN) with a pre-trained U-Net model trained on a dataset of 3,650 axial T2-weighted MRI images from 100 patients with lower back pain. We included all qualities of MRI; the exclusion criteria were fractures, tumors, infection, or spine implants. The training was performed using k-fold cross-validation (k = 10), and performance was evaluated using the dice similarity coefficient (DSC) and cross-sectional area error (CSA error). For clinical correlation, we used a simplified Goutallier classification (SGC) system with three classes. Results: The mean DSC was high for overall muscle (0.91) and muscle tissue segmentation (0.83) but showed deficiencies in fatty tissue segmentation (0.51). The CSA error was small for the overall muscle area of 8.42%, and fatty tissue segmentation showed a high mean CSA error of 40.74%. The SGC classification was correctly predicted in 75% of the patients. Conclusion: Our fully connected CNN segmented overall muscle and muscle tissue with high precision and recall, as well as good DSC values. The mean predicted SGC values of all available patient axial slices showed promising results. With an overall Error of 25%, further development is needed for clinical implementation. Larger datasets and training of other model architectures are required to segment fatty tissue more accurately.

The stand-alone treatment of degenerative cervical spine pathologies is a proven method in clinical practice. However, its impact on subsidence, the resulting changes to the profile of the cervical spine and the possible influence of clinical results compared to treatment with additive plate osteosynthesis remain under discussion until present. This study was designed as a retrospective observational cohort study to test the hypothesis that radiographic subsidence of cervical cages is not associated with adverse clinical outcomes. 33 cervical segments were treated surgically by ACDF with stand-alone cage in 17 patients (11 female, 6 male), mean age 56 years (33–82 years), and re-examined after eight and twenty-six months (mean) by means of radiology and score assessment (Medical Outcomes Study Short Form (MOS-SF 36), Oswestry Neck Disability Index (ONDI), painDETECT questionnaire and the visual analogue scale (VAS)). Subsidence was observed in 50.5% of segments (18/33) and 70.6% of patients (12/17). 36.3% of cases of subsidence (12/33) were observed after eight months during mean time of follow-up 1. After 26 months during mean time of follow-up 2, full radiographic fusion was seen in 100%. MOS-SF 36, ONDI and VAS did not show any significant difference between cases with and without subsidence in the two-sample t-test. Only in one type of scoring (painDETECT questionnaire) did a statistically significant difference in t-Test emerge between the two groups (p = 0.03; α = 0.05). However, preoperative painDETECT score differ significantly between patients with subsidence (13.3 falling to 12.6) and patients without subsidence (7.8 dropped to 6.3). The radiological findings indicated 100% healing after stand-alone treatment with ACDF. Subsidence occurred in 50% of the segments treated. No impact on the clinical results was detected in the medium-term study period.

This study aimed to develop a graph neural network (GNN) for automated three-dimensional (3D) magnetic resonance imaging (MRI) visualization and Pfirrmann grading of intervertebral discs (IVDs), and benchmark it against manual classifications. Lumbar IVD MRI data from 300 patients were retrospectively analyzed. Two clinicians assessed the manual segmentation and grading for inter-rater reliability using Cohen's kappa. The IVDs were then processed and classified using an automated convolutional neural network (CNN)–GNN pipeline, and their performance was evaluated using F1 scores. Manual Pfirrmann grading exhibited moderate agreement (κ = 0.455–0.565) among the clinicians, with higher exact match frequencies at lower lumbar levels. Single-grade discrepancies were prevalent except at L5/S1. Automated segmentation of IVDs using a pretrained U-Net model achieved an F1 score of 0.85, with a precision and recall of 0.83 and 0.88, respectively. Following 3D reconstruction of the automatically segmented IVD into a 3D point-cloud representation of the target intervertebral disc, the GNN model demonstrated moderate performance in Pfirrmann classification. The highest precision (0.81) and F1 score (0.71) were observed at L2/3, whereas the overall metrics indicated moderate performance (precision: 0.46, recall: 0.47, and F1 score: 0.46), with variability across spinal levels. The integration of CNN and GNN offers a new perspective for automating IVD analysis in MRI. Although the current performance highlights the need for further refinement, the moderate accuracy of the model, combined with its 3D visualization capabilities, establishes a promising foundation for more advanced grading systems.