Pain is a challenging, multifaceted symptom reported by most pediatric patients. This systematic review aims to explore the progress and effectiveness of applying artificial intelligence (AI) technology in pediatric pain management. A comprehensive search of PubMed, Embase, Web of Science, Cochrane, Scopus, IEEE Xplore, ACM Library, and ClinicalTrials.gov was conducted. The search combined pain-related terms ("Pain management", "Pain assessment", "Pain measurement", "Pain relief", "Pain control", "Analgesics", "Pediatric pain"), age-related terms ("Children", "pediatrics", "Neonate", "Infant"), and AI core sub-domain terms ("Artificial intelligence", "Machine learning", "Deep learning", "Convolutional neural network", "Support vector machine", "Random forest", "Long short-term memory") using strict Boolean operators (AND/OR/NOT) published up to March 20, 2026. AI models, their validation, and performance were summarized. The analysis of 71 studies revealed distinct AI application patterns in pediatric pain management. Fifty-nine studies focused on pain assessment using deep learning (post-2020: 86.7%) and classical machine learning (pre-2015: 83.3%) through facial expression analysis (40.8%), multimodal fusion (25.4%), or physiological signals (9.9%). Most employed observational designs (57.7%) with small (<50 participants, 42.2%) to medium (50-200, 33.8%) samples, reflecting clinical data challenges. Multimodal approaches significantly outperformed unimodal methods (AUC difference: +0.13, p<0.01). The remaining 12 studies (16.9%) explored pain management, primarily using robot-assisted interventions with cognitive-behavioral strategies like guided breathing and gamification during procedural pain. However, a substantial proportion of the included studies (47.5% of assessment studies and 11 of 12 intervention studies) showed high risk of bias. This review offers substantiation that AI technology has been employed to enhance the efficiency of pain detection and assessment, thereby assisting healthcare professionals and patients in more adeptly managing acute pain. Future progress in precise pain assessment and management will depend on integrating deep learning with multimodal data and large clinical databases, alongside efforts to establish standardized datasets and validate models in real-world settings.
Wang et al. (Thu,) studied this question.