Artificial intelligence for perioperative precision in surgical oncology

Multimodal deep learning is advancing rapidly in surgical oncology, yet fragmented clinical data continue to constrain comprehensive understanding of tumor biology. Here we review progress in artificial intelligence (AI) for precision surgical oncology, with emphasis on its contributions to surgical quality across the preoperative, intraoperative, and postoperative continuum. Preoperatively, radiogenomic models link imaging phenotypes with molecular features to support noninvasive biopsy, neoadjuvant treatment selection, and organ-preservation strategies. Intraoperatively, real-time augmented reality, optical biopsy, and deformable image registration may improve visualization of tumor margins and functional boundaries. Postoperatively, pathomic and multi-omic frameworks may support biology-informed risk stratification and adaptive adjuvant therapy. Critical barriers include limited interpretability, unstable performance under distribution shift, and insufficient prospective evidence. Current evidence supports biologically informed AI as a promising decision-support layer; however, prospective multicenter trials, cost-effectiveness analyses, and governance safeguards are required before routine clinical adoption.