Artificial intelligence in plastic and reconstructive surgery: clinical applications, translational readiness and barriers to real-world implementation

Abstract

Artificial intelligence (AI) is expanding rapidly in plastic and reconstructive surgery, but the central question has shifted from whether an algorithm can perform a task to whether it can generalize, improve clinical decisions, and remain safe after deployment. This critical review examines contemporary AI applications in burn assessment, facial and craniofacial surgery, microsurgical free-flap monitoring, breast reconstruction, and patient communication, with emphasis on translational readiness rather than cataloguing technologies. A structured PubMed/MEDLINE search through September 6, 2026 was supplemented by reference chaining from recent systematic reviews and reporting guidance. Image-based systems have reached high performance for selected bounded tasks, including automated free-flap monitoring (area under the curve 0.960), burn-depth assessment (95% accuracy in a prospective clinical cohort), and some externally validated breast-reconstruction outcome models (area under the curve up to 0.97). In contrast, prediction of multifactorial complications remains less discriminative, and general-purpose large language models have shown clinically important limitations in factual accuracy and postoperative triage. The principal barrier is therefore not algorithm development but clinical translation: external validation, prospective impact studies, subgroup performance, human-AI workflow design, data governance, and post-deployment monitoring remain inconsistently addressed. Plastic surgery is well positioned to benefit from AI because it is image-rich and outcome-sensitive, but implementation should prioritize narrowly defined tasks with measurable clinical value and preserve surgeon oversight, patient-reported outcomes, and transparent lifecycle governance.