Lisa Le¹, Ofelia Loani Elvir-Lazo^2*, Robert Wong²

¹Kirk Kerkorian School of Medicine at UNLV, Las Vegas, NV, USA 

²Department of Anesthesiology, Cedars-Sinai Medical Center, Los Angeles, CA, USA

^*Corresponding Author: Ofelia Loani Elvir-Lazo, Department of Anesthesiology, Cedars-Sinai Medical Center, Los Angeles, CA, USA

Received: 28 August 2025; Accepted: 09 September 2025; Published: 23 October 2025

Article Information

Citation: Lisa Le, Ofelia Loani Elvir-Lazo, Robert Wong. A Scoping Review of Artificial Intelligence in Perioperative Anesthesia: Current Applications, Challenges, and Roadmap for the Future. Journal of Surgery and Research. 8 (2025): 507-511.

DOI: 10.26502/jsr.10020477

Abstract

Keywords

Artificial Intelligence, Anesthesiology, Perioperative Care, Machine Learning, Deep Learning, Precision Anesthesia, Clinical Decision Support

Article Details

1. Introduction

The integration of Artificial Intelligence (AI) into medicine holds immense potential to enhance diagnostic accuracy, personalize treatments, and streamline clinical workflows [1]. Anesthesiology, a data-intensive specialty, is uniquely positioned to benefit from these advancements [1]. The shift toward precision anesthesia marks a departure from population-based norms, enabling individualized care based on genetic, physiologic, and historical patient data [2]. Foundational research has highlighted key domains, including depth of anesthesia (DoA) monitoring, automated anesthetic control, risk prediction, and ultrasound guidance [3].

Systematic reviews have further categorized AI's clinical applicability into DoA monitoring, image-guided techniques, and adverse event prediction with artificial neural networks (ANNs) frequently employed [4]. These studies have established the potential for AI to facilitate a more predictive and preventive approach to anesthesia care [5]. Since 2018, bibliometric analyses have revealed growing interest in hypotension prediction and anesthetic depth monitoring [6]. This scoping review builds on that foundation by synthesizing the major innovations and applications reported in the 2025 literature.

2. Methods

A comprehensive literature review was conducted to identify current and emerging applications of AI in the clinical practice of anesthesiology. Systematic searches were performed across key databases, including PubMed and Google Scholar, and enhanced using AI-powered tools such as Research Rabbit and Google Gemini to uncover interconnected and relevant publications. The search strategy combined keywords including "AI," "anesthesia," "perioperative care," "algorithms," and "devices." The initial search was supplemented by a citation tracking and snowballing technique to identify interconnected and relevant publications, ensuring a thorough synthesis of the available evidence.

This scoping review incorporated 24 peer-reviewed articles, divided into two cohorts: a foundational group of 8 articles published prior to 2025 to establish context, and a primary analysis group of 16 articles published in 2025. Data from each article was systematically extracted, focusing on the specific AI algorithms and devices discussed and their application within the perioperative stages of anesthesia. Findings were categorized into three phases of care: preoperative, intraoperative, and postoperative, to enable a structured analysis of AI’s evolving role in anesthetic practice. Table 1 provides a detailed summary of these AI applications categorized by perioperative phase.

3. AI in preoperative care: Proactive risk mitigation

The preoperative phase is crucial for identifying patient risks and planning for a safe anesthetic course. AI applications have demonstrated significant sophistication in enhancing these assessments. A primary focus has been on airway management, a cornerstone of anesthetic safety. While traditional airway assessment methods often lack precision, AI provides a more robust approach (Table 1). Systematic reviews of current literature show that AI models, particularly those using Convolutional Neural Networks (CNNs) and other deep learning algorithms, excel at analyzing patient images (e.g., facial photographs, ultrasound) to accurately predict difficult airways [7,8]. These tools can automatically classify Mallampati scores, measure thyromental distance, and identify other subtle anatomical predictors of difficult intubation from images, providing clinicians with a noninvasive, rapid, and objective assessment tool [7]. The predictive power of these analyses extends to forecasting the likelihood of difficult ventilation and intubation.

Beyond airway assessment, AI models are being used for comprehensive risk stratification (Table 1). By analyzing electronic health records (EHRs), AI can predict the likelihood of various perioperative complications, such as postoperative nausea and vomiting (PONV), delirium, or acute kidney injury [9]. Specific AI tools like the POTTER (Postoperative Transfer to ICU Risk) calculator have demonstrated superior accuracy in predicting postoperative complications and mortality compared to traditional methods. Furthermore, real-time systems like MySurgeryRisk can be integrated directly into EHRs to support bedside clinical decision making, signaling a shift toward data-driven, precision surgical care [2]. AI is also beginning to assist in anesthesia planning itself, with models suggesting optimal drug choices and initial dosing based on a patient’s specific physiological profile and surgical context [10].

4. AI in intraoperative care: Enhancing real-time decision making

The intraoperative phase is the most dynamic period of anesthesia care, requiring constant vigilance. AI is emerging as a powerful copilot for the anesthesiologist, enhancing monitoring, automating drug delivery, and providing real-time decision support.

4.1 Depth of anesthesia (DoA) and hemodynamic monitoring

AI continues to refine DoA monitoring (Table 1), moving beyond traditional electroencephalogram (EEG) based indices. A study by Wang et al. [11] introduced a novel hybrid AI model combining Long Short Term Memory (LSTM) networks, Transformer architectures, and Kolmogorov Arnold Networks (KAN). This sophisticated model predicts anesthesia depth not just from EEG signals but also from the drug infusion history itself, capturing complex temporal dependencies and nonlinear relationships. Concurrently, AI systems incorporating LSTM algorithms are being used to manage anesthetic dosage in real-time during complex procedures like cancer surgery, demonstrating the ability to maintain greater hemodynamic stability compared to manual control [12].

4.2 Automated anesthesia and closed-loop systems

The field of automated anesthesia and closed-loop systems (Table 1) has been the cornerstone of AI’s integration into intraoperative monitoring. Comprehensive reviews in 2025 detail the advances in closed-loop systems that use AI to automatically titrate hypnotic and analgesic drugs [13]. These systems utilize physiological modeling and reinforcement learning to adapt to the patient's state, aiming to maintain a steady DoA and stable hemodynamics. Models like Support Vector Machines, Decision Trees, and Gradient Boosting are being evaluated for their accuracy in predicting real-time anesthetic requirements to power these automated systems [10].

Table 1: Key applications of artificial intelligence (AI) across the perioperative period in anesthesiology.

4.3 Surgical and anesthetic synergy

Furthermore, AI is fostering a new synergy between surgical and anesthetic teams. AI-powered computer vision can track surgical progress, anticipate high-stimulation events (Table 1), and alert the anesthesiologist to prepare for hemodynamic responses [14]. This integration creates a more cohesive and anticipatory intraoperative environment, allowing anesthesiologists to better predict potential hemodynamic changes, specifically during novel or unfamiliar surgical procedures.

5. AI in postoperative care: Optimizing recovery

The impact of AI is increasingly extending into the postoperative period, focusing on pain management, complication prediction, and facilitating a smoother recovery. These elements are essential aspects of postoperative care to ensure patient safety. AI is being used to predict and manage postoperative pain more effectively (Table 1). Models can forecast which patients are likely to experience severe pain, enabling preemptive multimodal analgesia [15]. Some systems leverage Natural Language Processing (NLP) to analyze patient-reported outcomes and nurses' notes from EHRs to get a richer understanding of a patient's recovery trajectory [16]. Unlike structured vitals or lab results, these narratives often capture nuanced symptoms such as breakthrough pain, anxiety, or functional limitations that are missed by standard monitoring [16]. By integrating all of the data and insights, AI provides a more holistic and patient-centered understanding of recovery trajectories.

Furthermore, AI algorithms can be used to predict the risk of postoperative complications such as respiratory depression, infection, or readmission [9] (Table 1). By identifying at-risk patients early, clinicians can implement enhanced monitoring and targeted interventions. The development of wearable sensors that feed data into AI platforms allows for continuous remote monitoring after discharge, enabling early detection of deterioration [16] (Table 1). This continuous oversight is particularly important during the vulnerable transition from hospital to home, reducing preventable readmissions and improving long-term outcomes.

6. Discussion and Future directions

The current literature demonstrates a notable maturation of AI in anesthesiology, transitioning from theoretical models to validated clinical support tools. The "three pillars" of AI research, prediction, monitoring, and intervention, are now well established across all phases of perioperative care [17]. However, significant hurdles remain. The Anesthesia Research Council highlights barriers, including the need for high-quality, large-scale datasets, the "black box" nature of some complex models, and the challenges of clinical implementation and regulatory approval [18]. To ensure the responsible integration of AI, essential safeguards include robust data governance, transparency, clinician education, and bias mitigation [23].

Future research must focus on model interpretability, ensuring that clinicians can understand and trust AI-driven recommendations. The development of federated learning approaches may help address data privacy concerns, allowing models to be trained across institutions without sharing raw patient data [13]. This collaborative framework is particularly promising for fields with sensitive data, like anesthesia, as it enables the development of more robust, generalizable AI models. Ultimately, the goal is not to replace the anesthesiologist but to augment their capabilities, reducing cognitive load and allowing them to focus on complex decision-making and patient care [5].

7. Conclusion

Current research highlights the growing integration of AI in anesthesiology, spanning preoperative risk assessment, intraoperative management, and postoperative monitoring. These technologies enhance safety, precision, and efficiency, supporting a shift toward data-driven, personalized care. Despite ongoing challenges in governance and implementation, AI is increasingly seen as a powerful tool to augment rather than replace clinical expertise, empowering anesthesiologists to improve outcomes across the perioperative spectrum.

Conflicts of interest

The authors certify that there is no conflict of interest with any financial organization regarding the material discussed in the manuscript.

Funding

This research did not receive specific grant funding from public, commercial, or not-for-profit organizations.

References

1. McManus S, Khatib R, Mathur P. Artificial intelligence in anesthesiology. In: Artificial Intelligence in Clinical Practice 11 (2024): 307-313.
2. Dabbagh A, Sabouri AS. Preface: Toward Artificial Intelligence-Driven Precision Anesthesia. Anesthesiology Clinics 43 (2025): 15-17.
3. Hashimoto DA, Witkowski E, Gao L, et al. Artificial Intelligence in Anesthesiology: Current Techniques, Clinical Applications, and Limitations. Anesthesiology 132 (2020): 379-394.
4. Lopes S, Rocha G, Guimarães-Pereira L. Artificial intelligence and its clinical application in Anesthesiology: a systematic review. Journal of Clinical Monitoring and Computing 38 (2024): 247-259.
5. Queiroz VN, Dalle LRP, Ashihara C, et al. Key concepts in artificial intelligence for anesthesiologists: a literature review. Perioperative Anesthesia Reports 3 (2025): e00132024.
6. Xie BH, Li TT, Ma FT, et al. Artificial intelligence in anesthesiology: a bibliometric analysis. Perioperative Medicine 13 (2024): 121.
7. Abosamak MF, Zaki HA, Shaban EE, et al. Artificial intelligence in airway management: A systematic review and meta-analysis. Anaesthesia Critical Care & Pain Medicine 44 (2025): 101589.
8. Guo M, Hou Y, Liu Y, et al. From algorithms to airways: Applying artificial intelligence to enhance airway assessment, management, and training. Trends in Anaesthesia and Critical Care 61 (2025): 101548.
9. Mirza M, Mehtab M, Islam B, et al. Artificial Intelligence for Perioperative Medicine in Anesthesia: Present State and Future Perspectives of AI in Anesthesiology. The Healer Journal of Physiotherapy and Rehabilitation Sciences 5 (2025): 46-59.
10. Patangia D, Panda A, Dash SR, et al. AI Enhanced Anesthesia Dosing: Evaluating ML Models. 2025 International Conference on Emerging Systems and Intelligent Computing (ESIC). 12 (2025): 158-163.
11. Wang L, Weng Y, Yu W. Anesthesia depth prediction from drug infusion history using hybrid AI. BMC Medical Informatics and Decision Making 25 (2025): 158.
12. Shen Z, Wang Y, Hu K, et al. Exploration of Clinical Application of AI System Incorporating LSTM Algorithm for Management of Anesthetic Dose in Cancer Surgery. Journal of Theory and Practice in Clinical Sciences 2 (2025): 17-28.
13. Cai X, Wang X, Zhu Y, et al. Advances in automated anesthesia: a comprehensive review. Anesthesiology and Perioperative Science 3 (2025): 122.
14. Lengston V, Falkner S. AI in the OR Revolutionizing Surgery and Anesthesia. Journal of Surgery & Anesthesia Research 240 (2025): 2-6.
15. Ye J, Bronstein S. Artificial intelligence and digital health in Anesthesiology, in the Digital Doctor 10 (2025): 291-307.
16. Kaiati EHMA. The Positive Impacts of Artificial Intelligence in Anesthesia and Anesthesia Technology. Personalized Medicine Journal 10 (2025): 9-19.
17. Lee Y. Three pillars of artificial intelligence research in anesthesiology: welcoming address to the Korean Journal of Anesthesiology’s new guidelines for machine learning and deep learning research. Korean Journal of Anesthesiology 78 (2025): 181-182.
18. Lonsdale H, Burns ML, Epstein RH, et al. Strengthening Discovery and Application of Artificial Intelligence in Anesthesiology: A Report from the Anesthesia Research Council. Anesthesiology 142 (2025): 599-610.
19. Wang G, Li C, Tang F, et al. A fully automatic semi-supervised deep learning model for difficult airway assessment. Heliyon 9 (2023): e15629.
20. Cascella M, Tracey MC, Petrucci E, et al. Exploring Artificial Intelligence in Anesthesia: A Primer on Ethics, and Clinical Applications. Surgeries 4 (2023): 264-274.
21. Wang D, Guo Y, Yin Q, et al. Analgesia quality index improves the quality of postoperative pain management: a retrospective observational study of 14,747 patients between 2014 and 2021. BMC Anesthesiology 23 (2023): 281.
22. Carneiro RAAG, Pereira LAG. Depth of Anesthesia Monitoring and Artificial Intelligence. Current Anesthesiology Reports 15 (2024): 19.
23. Fleisher LA. Foreword: Artificial Intelligence in Anesthesiology and Perioperative Care: An Inevitability Requiring Appropriate Safeguards and Education. Anesthesiology Clinics 43 (2025): xiii-xiv.
24. Zheng Z, Duan Y, Lin J, et al. Artificial intelligence in anesthesia: insights from the 2024 Nobel Prize in Physics. Anesthesiology and Perioperative Science 3 (2025).