AI-Enabled Multimodal Clinical Decision Support for Solid Tumor Surveillance: A Structured Narrative Review

Post-curative surveillance of solid tumors increasingly combines circulating tumor DNA (ctDNA), imaging, and clinical information, yet these data are rarely integrated into validated clinical decision-support (CDS) pathways. Artificial intelligence (AI) may provide a framework for dynamic, multimodal risk estimation. To synthesize evidence and methodological approaches supporting AI-enabled integration of longitudinal ctDNA, imaging/radiomics, and clinical variables for post-curative solid tumor surveillance. This structured narrative review examined clinical surveillance evidence, longitudinal ctDNA modeling strategies, multimodal data-fusion approaches, and decision frameworks for discordant ctDNA–imaging findings. The synthesis emphasized recurrence-risk discrimination, molecular lead time, dynamic prediction, multimodal integration, and requirements for prospective CDS evaluation. Longitudinal ctDNA consistently improves recurrence-risk stratification and can identify molecular relapse before radiologic recurrence across several solid tumors. Joint longitudinal–survival models, Bayesian updating, and machine-learning approaches enable time-updated risk estimation while accounting for serial measurements and uncertainty. Multimodal models combining ctDNA with imaging/radiomics and clinical variables demonstrate incremental prognostic discrimination compared with single-modality approaches. However, discordant ctDNA–imaging states remain a major translational challenge, and predictive performance does not by itself establish clinical utility. AI-enabled multimodal surveillance could transform post-curative monitoring from static assessment toward continuously updated, interpretable risk estimation. Prospective external validation, standardized data acquisition, explicit management pathways for discordance, and clinical-utility trials are required before such systems can support routine surveillance decisions.