INTRODUCTION
Meningiomas are the most common primary intracranial tumors and are frequently monitored over extended periods. Volumetric assessment typically requires manual segmentation, which is time-consuming and associated with interrater variability. This study aimed to develop and validate a deep learning-based model for the automated segmentation of meningiomas and associated peritumoral edema on preoperative magnetic resonance imaging (MRI).
METHODS
We trained a standard nnU-Net deep learning model on contrast-enhanced T1-weighted and FLAIR MRI scans from 100 patients treated at the University Hospital of Zurich. The model was then externally validated on 88 cases from the meningioma SEG-Class dataset from the Cancer Imaging Archive. Segmentation performance was assessed using the Dice similarity coefficient, Jaccard index, and 95th percentile Hausdorff distance.
RESULTS
The model achieved mean Dice scores of 0.87 ± 0.23 for meningioma segmentation and 0.63 ± 0.38 for peritumoral edema in internal cross-validation. On the external validation set, the model achieved scores of 0.86 ± 0.17 for meningioma segmentation and 0.31 ± 0.35 for edema.
CONCLUSION
The deep learning model demonstrated high accuracy in segmenting meningiomas and modest performance for peritumoral edema. These results support the potential utility of automated segmentation tools in clinical workflows. Future work should focus on validating model performance across larger multi-center datasets.
D. de Wilde, Olivier Zanier, A. Alakmeh et al.· Neuroradiology· 0 citations
ABSTRACT Background Preclinical spine research is limited by heterogeneous experimental reporting, fragmented documentation, and barriers to reproducibility and translational alignment. Large language models (LLMs) and related artificial intelligence (AI) technologies may support semantic interpretation, structured data extraction, and reasoning over biomedical text, but their role in experimental spine science remains unclear. This focused scoping review and expert perspective mapped current AI/LLM applications in spine research, quantified the preclinical evidence gap, and identified responsible integration opportunities. Methods A structured search of PubMed, Embase, and Web of Science was performed for studies published from January 2020 to January 2026 evaluating LLM, AI, chatbot, or advanced natural language processing applications in spine‐related contexts. The search intentionally captured both LLM‐specific and broader AI/chatbot applications to map the translational landscape. For this preclinical‐focused analysis, the corpus was re‐examined for experimental and translational use cases, supplemented by expert synthesis of methodologically relevant adjacent biomedical literature. Results Of 792 records identified, 166 unique studies met inclusion criteria. Publication activity increased markedly over time. The literature was dominated by conversational assessment/patient‐reported outcome measure applications (82/166; 49.4%), patient education/information quality studies (53/166; 31.9%), and other LLM/AI applications (19/166; 11.4%). Preclinical/basic science applications were rare (3/166; 1.8%) and used classical machine learning, deep learning, or broader AI frameworks rather than generative LLMs. The most credible near‐term opportunities include schema‐constrained data extraction, protocol completeness checking, ontology‐aligned data structuring, and evidence‐grounded workflow support under human supervision. Conclusion In preclinical spine research, LLMs are best positioned as human‐supervised workflow instruments for structuring fragmented experimental knowledge. Spine‐specific validation and robust governance are essential for responsible translational use.
Siegmund Lang, Stefan Motov, J. Krueckel et al.· JOR Spine· 0 citations