Accurate classification of bone tumors as benign, malignant, or intermediate is crucial for patient treatment decisions. Misclassification may result in overtreatment of benign cases or delayed intervention for aggressive tumors, significantly impacting patient prognosis. However, current methods rely heavily on single-modality imaging analysis, making it difficult to handle variable lesion locations and complex cancer types. To address these limitations, we propose a novel multimodal deep learning framework that integrates clinical images, pathological slices, and blood biomarkers for automated bone tumor detection and three-class classification. The framework operates in two stages: first, a YOLOv5-based detection model localizes tumor regions on clinical images. Next, a classification model utilizes ResNet to extract deep features from both the clinical images and pathological slices, while abnormal blood biomarkers are transformed into descriptive text by a large language model and subsequently encoded into semantic features using BioBERT. Finally, features from all three modalities are integrated via a fusion module to capture complementary information and enable accurate tumor classification. The evaluation was performed using two distinct datasets: a clinical imaging dataset for bone tumor detection, and a separate multi-modal cohort comprising clinical imaging, pathology, and blood biomarkers for tumor classification. The detection model demonstrated strong localization capabilities, achieving a test mAP@0.5 of 0.7925. For the classification task, ablation studies validated the complementary contribution of each modality. Notably, our multimodal fusion approach outperformed unimodal baselines, attaining a macro-average precision of 0.9056, F1-score of 0.8736, and AUC of 0.9759 in tumor classification—outperforming existing models.

The integration of multimodal medical imaging with artificial intelligence (AI) is potentially catalysing a paradigm shift in orthopaedic diagnosis and treatment, moving beyond experience-based practices toward intelligent, data-driven precision medicine. This narrative review synthesizes recent key evidence across imaging modalities and AI frameworks, and highlights the translational gap that persists between algorithmic development and real-world clinical implementation. By combining complementary information from X-ray, CT, MRI, PET, ultrasound, and biochemical data, multimodal AI overcomes the inherent limitations of single-modality approaches, enabling more comprehensive structural, functional, and metabolic assessments. Recent advances demonstrate broad applications, including accurate fracture detection and classification, differentiation of benign and malignant bone tumours, quantitative assessment of osteoarthritis, risk prediction for osteoporosis, and intelligent preoperative planning and intraoperative navigation. Moreover, multimodal AI facilitates efficacy prediction and personalised treatment decision-making, positioning future systems as AI-assisted decision-support tools that support surgeons in surgical strategy, implant design, and long-term follow-up. Nevertheless, significant challenges remain, particularly in data heterogeneity, model generalisation, interpretability, and clinical integration. Progress in constructing standardised multimodal databases, developing self-supervised and multi-task learning strategies, and strengthening ethical–regulatory frameworks will be essential for clinical translation. Ultimately, multimodal AI holds immense potential to transition from laboratory validation to routine practice, delivering safer, more efficient, and precise diagnostic and therapeutic solutions for orthopaedic patients.

Streptomyces species are ubiquitous soil actinomycetes and a major source of antibiotics, but invasive human infection with spinal involvement is exceeding

Percutaneous vertebroplasty-associated spinal infection is uncommon but may cause progressive vertebral destruction and can mimic tuberculous spondylitis on imaging. The detection of low-virulence skin commensals such as Cutibacterium species in deep specimens also creates a diagnostic challenge because contamination and true infection must be distinguished. We report a 78-year-old man who developed recurrent progressive low back pain shortly after percutaneous vertebroplasty for an L2 compression fracture. Baseline computed tomography (CT) and contrast-enhanced magnetic resonance imaging (MRI) showed destructive spondylodiscitis centered at L2/3, with adjacent vertebral body and disc-space involvement, paravertebral inflammatory extension, and bilateral psoas abscess-like collections, initially suggesting tuberculous spondylitis. However, molecular testing for the Mycobacterium tuberculosis complex and targeted sequencing for mycobacterial drug-resistance genes were negative, and histopathology demonstrated acute and chronic inflammatory necrotic changes without typical tuberculous granulomas. Metagenomic next-generation sequencing of aspirated fluid identified a high abundance of Cutibacterium modestum , which was subsequently detected again in lesion tissue and peripheral blood by targeted high-throughput sequencing. Given the repeated detection in deep-site specimens, exclusion of tuberculosis and other competing diagnoses, compatible clinical and imaging findings, and improvement after targeted antimicrobial therapy and percutaneous stabilization, C. modestum was considered the most probable causative pathogen. These findings suggest that pathogen attribution in post-vertebroplasty spinal infection should rely on integrated clinical, microbiological, pathological, and therapeutic evidence rather than on imaging findings alone. Integrated clinico-microbiological assessment may improve pathogen attribution for rare low-virulence organisms in postoperative spinal infection.