Background: The incorporation of artificial intelligence (AI) in healthcare is revolutionizing
real-time monitoring during surgical procedures, with the potential to improve patient safety, refine decision-making, and diminish problems. Real-time monitoring is the ongoing assessment of vital signs, hemodynamic parameters, and surgical circumstances to facilitate intraoperative decision-making. This systematic review seeks to examine the utilization of AI in real-time surgical monitoring, emphasizing its effects on clinical outcomes, efficiency, and safety. The main aim is to compile existing research and evaluate the efficacy of AI in delivering actionable insights during surgical procedures. Materials and Methods: The review adhered to PRISMA guidelines. Databases such as PubMed, Scopus, Web of Science, and IEEE Xplore were carefully queried for articles published without time limitation until ۲۰۲۴. Search phrases encompassed "artificial intelligence," "real-time monitoring," "surgery," and "patient status". Studies were eligible if they concentrated on AI-based instruments for intraoperative monitoring, documented clinical results, and underwent peer review. The exclusion criteria comprised non-English publications, animal research, and studies devoid of clinical validation. The Cochrane Risk of Bias Tool was employed to evaluate bias in randomized research, while ROBINS-I was utilized for non-randomized studies. Data were synthesized both qualitatively and quantitatively, with effect sizes computed for primary outcomes and illustrated using forest plots. Results: ۳۲ studies, comprising ۲,۴۶۰ surgical cases, fulfilled the inclusion criteria. Significant data demonstrated that AI models markedly enhanced the identification of crucial intraoperative occurrences, including hypoxia (effect size: ۱.۲۸, ۹۵% CI: ۱.۱۰-۱.۴۶) and hemodynamic instability (effect size: ۱.۲۲, ۹۵% CI: ۱.۰۵-۱.۳۹). The trials indicated an average reduction of ۳۵% in response times to adverse events.
Artificial intelligence systems shown significant accuracy (area under the curve, AUC: ۰.۸۹-۰.۹۷) in forecasting bad events. The risk of bias was predominantly low to moderate in the majority of investigations. Conclusion: This comprehensive analysis emphasizes the potential of AI in improving
real-time monitoring of patient conditions during surgery, offering substantial advantages in forecasting and addressing adverse outcomes. The results indicate AI's capacity to enhance surgical safety and outcomes, yet obstacles persist in incorporating these technologies into clinical workflows. Future research should focus on verifying AI systems across varied surgical contexts and addressing ethical problems connected to their use.
