Artificial intelligence in perioperative pain management: A systematic review

Background: Inadequate assessment and management of perioperative pain remains a clinical challenge associated with adverse outcomes and increased costs. Artificial intelligence (AI) offers potential solutions. A synthesis of its specific application in perioperative pain is lacking. This systematic review evaluated their utility, limitations, and implementation challenges.Methods: This systematic review followed the PRISMA 2020 guidelines. PubMed, Embase, Cochrane CENTRAL, and Web of Science were searched from January 1, 2020, to February 25, 2025. Studies evaluating AI algorithms for predicting, detecting, or managing perioperative pain were included. Studies published before 2020, animal research, chronic pain-only studies, protocols, and reviews were excluded. Two independent reviewers performed study selection and data extraction. Quality and reporting were assessed using TRIPOD, CONSORT-AI, and RoB 2. Findings were synthesized narratively, categorized by application type.Results: Twenty-nine studies met the inclusion criteria. AI applications were categorized into: (1) Prediction of pain-related adverse events (n=12), showing globally moderate performance (AUC 0.65–0.85; (2) Monitoring and pharmacological treatment (n=9), where NOL-guided analgesia reduced postoperative pain scores and AI showed feasibility for predicting remifentanil dose adjustments and detecting postoperative pain via facial analysis; (3) Optimization of locoregional analgesia techniques (n=8), where AI-assisted ultrasound demonstrated improved anatomical structure identification.Conclusions: AI demonstrates significant potential in perioperative pain management. However, the current evidence is limited by methodological heterogeneity, inconsistent reporting, lack of external validation, and insufficient assessment of clinical impact. Prospective studies, standardization, and careful consideration of ethical and implementation challenges are required before widespread adoption.