Bone and soft tissue tumors (BSTs) are rare, heterogeneous neoplasms with overlapping imaging features that complicate diagnosis and management. Conventional interpretation of modalities such as magnetic resonance imaging (MRI), computed tomography (CT), and positron emission tomography (PET) is limited by interobserver variability and qualitative assessment. This narrative review summarizes advances in artificial intelligence (AI) and machine learning (ML) for diagnostic imaging of BSTs, with emphasis on their clinical applications, challenges, and translational potential. A literature search was conducted in PubMed, Scopus, and Google Scholar for studies published between 2010 and 2025 using keywords including “AI,” “machine learning,” “deep learning,” “radiomics,” “bone tumors,” and “soft tissue sarcoma.” Peer-reviewed English-language articles were included; conference abstracts and studies unrelated to imaging were excluded. AI and ML applications in musculoskeletal oncology include automated lesion detection and segmentation, improved benign–malignant differentiation through radiomics, noninvasive tumor subtyping and heterogeneity mapping, prediction of treatment response, recurrence monitoring, and survival outcome modeling. Multimodal frameworks integrating MRI, CT, PET, and genomic or clinical data show promise for personalized oncology care. As a narrative review, this work may not capture all relevant studies. Heterogeneity in study design and small sample sizes limit generalizability, and meta-analysis was not feasible. AI and ML are transforming musculoskeletal oncology imaging by enhancing diagnostic accuracy and workflow efficiency. However, clinical translation remains constrained by limited datasets, a lack of standardization, annotation bias, and regulatory challenges. Future progress will depend on large multicenter collaborations, standardized imaging protocols, and harmonized regulatory frameworks to ensure safe and equitable integration into oncology practice.
Gupta et al. (Thu,) studied this question.