Applications of Artificial Intelligence in Anesthesia and Perioperative Medicine: A Systematic Review of the Literature

Abstract: Introduction: Artificial intelligence (AI) is rapidly expanding across many fields of medicine, including anesthesia and perioperative medicine. Through machine learning and deep learning techniques, it has become possible to analyze large amounts of physiological, hemodynamic, ventilatory, and electroencephalographic data in order to improve intraoperative monitoring, anticipate complications, and optimize clinical decision-making. Objective: To analyze the main applications of AI in anesthesia and perioperative medicine and to evaluate its potential contribution to clinical practice based on the available literature. Methods: A systematic review of the literature was conducted using PubMed, Scopus, and Embase for the period 2010–2025. The search strategy included the following keywords: “artificial intelligence,” “machine learning,” “deep learning,” “anesthesia,” “anesthesiology,” “perioperative,” “hypotension,” “EEG,” “ventilation,” and “postoperative complications.” Original studies focusing on clinical or perioperative applications of AI in anesthesia were included. Results: The main areas of application identified were prediction of intraoperative hypotension, monitoring of anesthetic depth, automated administration of anesthetic agents, ventilatory optimization, and prediction of postoperative complications. In the study by Hatib et al., an algorithm developed from 1,334 patients, 545,959 minutes of recordings, and 25,461 hypotensive episodes achieved an area under the curve (AUC) of 0.95 at 15 minutes and 0.97 at 5 minutes before the event. The randomized HYPE trial demonstrated a reduction in the time-weighted average (TWA) of hypotension from 0.44 mmHg to 0.10 mmHg using a predictive alert system. For postoperative delirium, Bishara et al., reported an AUC of 0.851 with XGBoost in a cohort of 24,885 patients. For postoperative pulmonary complications, Li et al., reported an AUC ranging from 0.878 to 0.881, outperforming the ARISCAT score. Closed-loop propofol delivery systems also provided better maintenance of the target depth of anesthesia than manual control. Conclusion: AI represents a promising tool in anesthesia and perioperative medicine. It may improve early detection of adverse events, personalize patient management, and enhance anesthetic safety. However, most currently available studies remain retrospective or focused mainly on technical performance. Prospective multicenter validation studies are still needed before broad implementation in routine practice.