Artificial intelligence in risk stratification and therapeutic decision-making for multiple myeloma: Recent advances and future directions

Multiple myeloma (MM) remains a clinically heterogeneous plasma-cell malignancy in which outcomes vary markedly across patients despite therapeutic progress. Conventional staging systems, such as the International Staging System (ISS) and Revised International Staging System (R-ISS), provide only limited prognostic value in the era of multi-omics and real-world data integration. In recent years, artificial intelligence (AI) and machine-learning approaches have emerged as promising tools for individualized risk assessment and therapeutic optimization in MM. This review synthesizes current evidence on AI-driven models applied to MM risk stratification and treatment decision-making. Representative algorithms, including the Consensus-based Risk-Stratification System (CRSS), modified risk staging (MRS), graph convolutional network-based risk-stratification system (GCRS), the Gaussian mixture model using expectation-maximization (GMM-EM), and transformer-based Simultaneous Cancer Outcome Prediction Estimator (SCOPE) platform, have shown modestly improved or comparable prognostic discrimination relative to conventional staging by integrating clinical, laboratory, cytogenetic, and longitudinal variables, although the absolute gains are often small and not formally tested. In therapeutic contexts, AI-enabled image analysis (e.g., 3D ResNet-34, convolutional neural network-based radiomics), clinical-data modeling (e.g., XGBoost infection prediction, Morphogo-based morphology recognition), and synthetic patient-data approaches (e.g., SynthIPD) support individualized treatment strategy and toxicity monitoring, reflecting a broader shift from static, diagnosis-time stratification toward dynamic monitoring. Despite these advances, challenges remain regarding data heterogeneity, algorithm interpretability, limited external validation across ethnic and clinical subgroups, and cost-effectiveness. Emerging paradigms such as multimodal data fusion, explainable AI, and federated learning frameworks are expected to enhance the transparency, reproducibility, and clinical applicability of AI-based precision medicine in MM. Collectively, current findings suggest that responsible and interpretable AI integration can meaningfully support clinical decision-making and patient-specific management in MM.