Artificial intelligence in airway management for oral and maxillofacial surgery: a narrative review

Introduction

The global prevalence of oral and oropharyngeal diseases is rising, leading to increased demand for oral and maxillofacial (OMF) surgery (1,2). OMF anesthesia presents unique challenges due to the complexity of airway management inherent to this surgical specialty (3). Airway management in OMF anesthesia is complicated by three primary factors. First, patients often present with a high likelihood of difficult airways due to anatomical variations or tumor obstruction. Second, the shared airway between the surgery and anesthesia increases the risk of airway management. Finally, surgical manipulation of the airway heightens the potential for injury and complicated extubation decisions.

Major complications from inadequate airway management during OMF anesthesia include death, brain injury, emergency surgical airway placement, and unplanned intensive care unit (ICU) admission (4). These risks underscore the need for more refined and precise strategies that span preoperative assessment, intraoperative monitoring, and safe extubation (5,6). While tools, such as bronchoscopes and visualization devices, have enhanced safety (7-9), artificial intelligence (AI) is emerging as a transformative technology in OMF anesthesia (10), demonstrating superior accuracy predicting difficult airways compared to traditional assessment methods (11,12).

AI is a branch of computer science that enables machines to mimic human intelligence in tasks such as learning, reasoning, and decision-making (13). In clinical anesthesia, AI is increasingly applied to ultrasound-guided regional anesthesia (14), postoperative pain control, airway prediction (15,16), depth of anesthesia monitoring (17), and intraoperative drug management (13,18). In OMF anesthesia, AI technologies have demonstrated potential to improved patient safety, optimizing clinical decision-making, and enhance procedural efficiency (19-21). Airway management in OMF anesthesia is a continuous and dynamic process spanning the preoperative, intraoperative, and post-extubation phases. Although current AI research remains largely stage-specific, integrating these components into a unified, adaptive decision-support system represents an important future direction.

Rationale and knowledge gap

Despite growing interest in AI-assisted perioperative care, no systematic or comprehensive review currently focuses specifically on AI applications in airway management for OMF anesthesia. For example, Xia (10) reviewed AI in OMF broadly but not offer focused analysis on airway management. This lack of dedicated evaluation represents a significant gap in literature.

Objective

This review aims to assess the development and current progress of AI applications in OMF airway management (Figure 1). It is structured into three sections: (I) an overview of relevant AI technologies; (II) a synthesis of their clinical applications in OMF airway management; and (III) discussion of future directions and research priorities.

Figure 1 AI applications in OMF airway management. AI, artificial intelligence; DA, difficult airway; OMF, oral and maxillofacial.

Maintaining adequate oxygenation during airway procedure is critical for all patients undergoing OMF surgery. A robust stepwise airway strategy supported by emerging AI technologies is essential to anticipate and mitigate risks. This review adopts an evidence-based approach to summarize current techniques and their potential for integration into clinical practice. We present this article in accordance with the Narrative Review reporting checklist (available at https://joma.amegroups.com/article/view/10.21037/joma-2025-28/rc).

Methods

A literature search was conducted using PubMed, Web of Science and Google Scholar in April 2025. The search included various combinations of keywords such as: “anesthesia”, “airway”, “OMF”, and “artificial intelligence”. In addition, a manual search was performed to identify additional potentially relevant articles. Studies were eligible if published in English between January 1, 2010 and March 31, 2025. Exclusion criteria included non-peer reviewed articles, case reports, comments, and conference summaries. Duplicate records were removed using Endnote (version 20), followed by manual screening of titles and abstracts by M.Y.L. Table 1 shows the summary of search strategy. Based on these criteria, 161 articles were initially retrieved. After applying inclusion and exclusion filters, 37 articles were included in this review.

AI technology

Overview of AI technology

AI includes various subfields such as machine learning (ML) and deep learning (DL). ML focuses on training algorithms to recognize patterns in structured data, while DL uses multi-layered artificial neural networks to extract complex features from unstructured inputs such as images and sound.

ML has been widely used in clinical decision support, including tasks such as optimal ventilator mode selection (22), real-time decision of patient-ventilator asynchrony (23), and prediction of difficult intubation using structured clinical data pipelines (24). It has also been employed to differentiate kinematic movement between expert and novice intubators (25).

DL techniques are especially effective for high-dimensional tasks, including acoustic respiratory monitoring (26), interpretation of ventilatory waveforms and respiratory sounds (27), extraction of orofacial landmarks (28), and segmentation of anatomical structures in the oral cavity using video laryngoscope (VL) images. DL models have also been used to detect flow starvation during square-flow-assisted ventilation and to enhance VL-based structural segmentation (29).

Choi et al. (30) utilized a region-based convolutional neural network (R-CNN) for a DL-based image recognition system that performs time-sequence analysis of tracheal intubation. Similarly, Wu et al. (31) developed a model that labeled six anatomical and procedural targets during intubation: the lip, epiglottis, laryngopharynx, glottic opening, tube tip, and the black line on the endotracheal tube. Obeso et al. (32) designed a ML approach specifically for pediatric patients to detect patient-ventilator desynchrony (PVD) by transforming ventilator waveform data into spectrograms and analyzing them using convolutional neural networks (CNNs).

In the medical field, AI applications can be categorized into three types of learning based on the presence or absence of labeled data: supervised learning, semi-supervised learning, and unsupervised learning.

Supervised learning requires manually labeled data. For example, Zhu et al. (33) evaluated the performance of an AI-driven intelligent laryngoscopy monitoring assistant (ILMA) for anatomical landmark identification in laryngoscopy images and videos. The system utilized CNN architecture combining Inception-ResNet-v2 with Squeeze-and-Excitation Networks (SENet). ILMA achieved a total accuracy of 0.976 in identifying 20 anatomical landmarks. Similarly, Anitha et al. (29) applied supervised DL techniques to detect flow starvation during square-flow-assisted ventilation.

Semi-supervised learning combines a small amount of labeled data with a large volume of unlabeled data. For instance, Ren et al. (34) proposed an end-to-end semi-supervised landmark prediction framework using global-to-local cross pseudo supervision for assessing airway difficulty.

Unsupervised learning involves uncovering hidden patterns or structures in data without the use of pre-labeled examples. This approach is valuable for exploratory analysis and discovering new insights from complex physiological or imaging data.

It is worth noting that in the medical field, there is often a trade-off between model performance and interpretability. This trade-off represents a critical barrier to clinical adoption, as clinicians need to understand how a model makes its decisions to ensure trust, safety, and accountability. Strategies based on explainable AI (XAI) methods, such as SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-agnostic Explanations), may help improve transparency by highlighting key contributing features while maintaining high predictive performance (35-37).

Common AI pipeline for airway management

A common AI pipeline for airway management involves several key stages. It begins with data collection, which includes acquiring high-quality inputs from imaging modalities, acoustic respiratory monitoring, and electronic patient records. This is followed by data preprocessing, where raw data are cleaned, normalized, and prepared for analysis. For example, through image segmentation or feature extraction. Next, model development entails training ML or DL algorithms to perform specific clinical tasks such as predicting difficult intubation or selecting optimal ventilator modes. Once trained, these models undergo validation, where their performance is evaluated using metrics such as accuracy, sensitivity, and specificity. Finally, the hope is that the successful models will be deployed in clinical settings for real-time monitoring and decision support to improve the safety and efficiency of airway management, and ongoing interdisciplinary efforts aim to translate these models into clinical tools.

AI in preoperative assessment

Visualization tool-based airway estimation

Thorough preoperative airway assessment is essential for planning safe and effective airway management in OMF anesthesia. Recent advancements in AI have enhanced the accuracy and efficiency of these assessments, particularly through visualization tools. A study showed that A-LRYNGO’s AI-based glottis guidance system demonstrated a high accuracy of 93.1% in identifying the glottic opening among airway visualization (Choi et al., 2022) (38).

Kim et al. (39) developed ML models to predict VL views using the percentage of glottic opening (POGO) score. Their approach incorporated predictors such as the modified Mallampati classification, thyromental height and distance, sternomental distance, mouth opening, neck circumference, and age (as a key predictor). Model performance was evaluated using root mean squared error (RMSE), mean squared error (MSE), mean absolute error (MAE), and the coefficient of determination (R²). The overall RMSE ranged between 20.4 and 21.9, suggesting a moderate deviation between predicted and actual POGO scores. Given that POGO scores range from 0% (no glottic opening visible) to 100% (full glottic exposure), this level of deviation represents a moderate prediction error, suggesting that while the model demonstrates reasonable predictive ability, further refinement is needed for clinical decision-making accuracy. These findings underscore the potential utility of regression-based ML models in preoperative airway assessment, offering a quantitative and reproducible approach to identifying difficult airways based on pre-intubation clinical variables.

In addition, Matava et al. (40) applied CNNs for real-time classification and labeling of vocal cord and tracheal anatomy using video laryngoscopy and bronchoscopy. This approach allows for automated identification of airway structures, potentially assisting clinicians in making rapid, accurate decisions during preoperative airway assessment. Model performance was evaluated using confidence scores, sensitivity, specificity, and frame-per-second (FPS) processing rates. Following transfer learning, model performance improved substantially, with the confidence score for vocal cord detection increasing to 0.96 in the ResNet model and 0.93 in the Inception model. These findings demonstrate the feasibility and effectiveness of applying AI-based visualization tools to support real-time airway structure recognition.

Nasal airway evaluation

Accurate evaluation of the nasal and pharyngeal airways is particularly relevant in OMF procedures requiring nasotracheal intubation (41). Cone-beam computed tomography (CBCT) has become an essential imaging modality in preoperative planning for such cases.

Recent studies have demonstrated the utility of AI in automatically segmenting the pharyngeal airway and sinonasal cavities from CBCT scans. Leonardi et al. (42) and Sin et al. (43) developed DL algorithms capable of accurately delineating these structures, which are critical for evaluating airway patency and predicting potential difficulties with nasotracheal intubation. Similarly, Shujaat et al. (44) used CNNs to automate segmentation of the pharyngeal airway space (PAS), providing objective measurements that enhance clinical decision-making.

In addition, Garcia et al. (45) proposed a depth map-based approach using endoscopy to estimate nasal airway cross-sectional areas, offering a less invasive yet promising method for preoperative nasal airway evaluation. Jin et al. (46) further demonstrated the feasibility of automatic identification of three-dimensional nasal and pharyngeal airway subregions using a vision transformer model.

Together, these AI-driven tools offer efficient, reproducible, and quantitative assessments of nasal and pharyngeal airways, supporting more precise planning for airway management in OMF anesthesia.

Clinical decision support

AI offers valuable support in this context by integrating patient-specific data into clinical decision support systems. These tools can recommend personalized airway management strategies, such as whether to pursue awake intubation, use supraglottic airway devices, or prepare for surgical airway access.

For trauma patients with oral-maxillofacial or tracheal injury, AI-based prediction models have shown promise in assessing the need for emergency front-of-neck airway procedures (41,47,48). Furthermore, preoperative imaging data, such as computed axial tomography, can be analyzed using AI to identify features associated with difficult airways in head and neck surgery (49). These advancements demonstrate the potential of AI to enhance clinical decision-making by integrating multimodal data for individualized airway planning in OMF anesthesia.

AI in difficult airway assessment

Due to the unique anatomical and surgical considerations in OMF surgeries, airway scenarios tend to be more complex and variable. Despite routine preoperative airway assessments, unanticipated difficult airways continue to occur (50). Historically, clinicians have relied on accessed difficult airway information to deal with airway management challenges (51). Such as, assessing the physical status of the airway in patients with no apparent anatomical airway abnormalities (52), and evaluating the role of transnasal flexible endoscopic laryngoscopy (TFEL) (53).

With the emergence of advanced technologies, there is a growing development of clinically applicable predictive models aimed at reducing the unanticipated difficult airway. In light of this, we provide an overview of advanced technologies and prediction models, and validated predictors that have demonstrated utility in the evaluation of difficult airways (Table 2).

Non-image feature analysis

ML classifiers have been increasingly used to predict difficult airways based on non-image data, such as electronic health record data. Using supervised learning algorithms, researchers label cases as difficult or non-difficult, for example, defining difficult laryngoscopy as a Cormack-Lehane grade III or IV (11). Through feature engineering, relevant variables are extracted and evaluated using performance metrics such as the area under the receiver operating characteristic curve (AUROC). This workflow is highly feasible and interpretable, leveraging clinically relevant variables derived from expert knowledge.

Wang et al. (12) compared ML models for difficult airway during OMF anesthesia, the best area under the curve (AUC) for difficult tracheal intubation (DTI) is 0.956, and for difficult laryngoscopy is 0.903. Such high performance may be influenced by the single-center cohort, limited sample size, and characteristics of the studied Chinese patient population, as noted by the authors. This study showed that comparing to non- difficult laryngoscopy and non-DTI patients, those with difficult laryngoscopy or DTI were significantly older and exhibited smaller thyromental distance (TMD), inter-incisor distance (IID), and head angle, reduced temporomandibular joint (TMJ) mobility, and greater tongue thickness (TT), along with higher upper lip bite test (ULBT), the modified Mallampati test (MMT), and the grade of maximum forward movement of the lower incisor beyond the upper incisor (SLUX) scores (all P<0.001); however, difficult laryngoscopy patients showed significant differences in gender, height, and weight, no such differences were observed in the DTI group.

Sezari et al. (54) identified 24 important features from 32 collected variables, including both discrete (e.g., loose front teeth, Mallampati score, snoring) and continuous values. Srivilaithon and Thanasarnpaiboon (11) highlighted nine key predictors such as male sex, trauma, limited mouth opening, and poor neck mobility. Zhou et al. (55) found age, sex, weight, height, and body mass index (BMI) to be the top five predictive features. Yamanaka et al. (56) compared with the modified LEMON (look, evaluate, Mallampati, obstruction, neck mobility) criteria with ML model for the difficult airway prediction. The most contributing predictor was age, followed by any criterion met in the modified LEMON criteria and hyoid mental distance ≥3 fingers. Kim et al. (57) emphasized neck circumference and thyromental height as significant indicators. Eberhart et al. (69) recognized presence of upper front teeth, a history of difficult intubation, any Mallampati status different from ‘1’ and equal to ‘4’ and mouth opening less than 4 cm are independent risk factors.

These findings highlight the importance of incorporating multidimensional anatomical and demographic features into AI models for accurate prediction of difficult airway. However, its reliance on predefined features may limit the discovery of novel predictive indicators.

Image analysis

Image-based analysis combined with AI technologies has rapidly advanced in difficult airway assessment to facilitate the endotracheal intubation (ETI) (70,71). These approaches involve feeding large volumes of images, especially facial images and VL images, into DL to extract relevant features or detect anatomical structures. Model performance is typically evaluated using metrics such as accuracy or AUC, and visual tools like heatmaps may assist in feature localization. Unlike traditional ML, this process emphasizes algorithmic performance over interpretability.

Hayasaka et al. (58) were the first to use DL on facial images to classify intubation difficulty. Tavolara et al. (59) trained models using frontal celebrity images to distinguish between easy and difficult intubations. Wang et al. (60) developed a semi-supervised DL model based on six clinical predictors extracted from facial photos, such as TMD and mouth opening. Kim et al. (61) enhanced prediction accuracy by training EfficientNet-B5 on multiple facial views, including frontal, lateral, and open-mouth images. Wu et al. (62) proposed a dual-path multi-view fusion network to integrate information across multiple angles. Additionally, Yoo et al. (63) applied DL to video bronchoscopy for anatomical interpretation and intubation guidance. Kim et al. used You Only Look Once version 4 (72) to detect vocal cords from VL images (73). Cho et al. (64) uses lateral cervical spine radiographs around the hyoid bone, pharynx and cervical spine to predict Cormack-Lehane grade 3 or 4 direct laryngoscopy views.

While DL models excel at extracting clinically relevant image features and show promise for real-time support, their limited interpretability compared to traditional feature-based models like logistic regression (LR) or decision trees remains a major barrier. This “black box” nature can hinder clinical trust and adoption, particularly in high-stakes scenarios. Striking a balance between predictive performance and explainability is essential, and future research should prioritize interpretable AI methods.

Recent studies have explored the integration of image and non-image data to improve the accuracy of difficult airway prediction. Xia et al. (65) combined facial images with baseline and clinical features using a neural network to predict difficult VL, demonstrating the potential of multimodal integration. Liu et al. (66) showed that Naïve Bayes model outperformed DL approaches, with comparable accuracy across different radiological predictors, and a hybrid approach combining image and non-image data may offer a practical and flexible solution for clinical airway prediction. A comparison of image-based and non-image-based AI approaches in difficult airway assessment for OMF anesthesia is summarized in Table 3.

Voice analysis

Recent studies have explored voice analysis as a non-invasive method for airway management. Steffensen et al. (67) utilized a full-size patient simulator equipped with two small electret condenser microphones and a piezoelectric buzzer to capture acoustic signals. ML was then employed to deliver real-time performance feedback. Intubations were categorized into six distinct groups based on combinations of tube placement depth (correct, deep, and shallow) and cuff inflation status (inflated or deflated). While the model demonstrated high sensitivity for detecting intubation depth [0.97 (0.91–0.99)], it showed limitations in accurately distinguishing cuff inflation status [0.85 (0.70–0.93)]. Rodiera et al. (68) applied voice analysis for preoperative difficult airway prediction by recording vowels in neutral, flexion, and extension positions. Their model successfully distinguished between easy and difficult airways based on acoustic features, highlighting the potential of voice-based AI tools as complementary methods in difficult airway assessment.

As summarized in Table 2, traditional ML algorithms such as LR, random forest, and XGBoost achieved robust performance in non-image datasets, with AUROC values typically between 0.79 and 0.96 and accuracy approaching 0.9. DL architectures, particularly CNN-based models (e.g., ResNet, DenseNet, EfficientNet), demonstrated superior capability in image and acoustic analysis, reaching AUCs up to 0.96 and accuracies above 0.9. Overall, DL methods show clear advantages in handling complex imaging and voice data, whereas conventional ML models remain effective and interpretable for structured clinical information.

AI in mechanical ventilation

The shared airway between the surgical and anesthesia teams introduces significant complexity to airway management in OMF surgery. Understanding the physiological basis of these challenges is essential. Computational fluid dynamics analyses have shown that even subtle changes in head and neck positioning can significantly modify upper airway airflow dynamics, offering an explanation for the variable airway conditions encountered during shared-airway procedures (74). The interventions can directly impact airway stability, increasing the risk of complications such as hypoxia, hypotension, or bleeding. These challenges highlight the critical importance of early warning systems and decision support for airway management, particularly during mechanical ventilation.

Predictive alarm systems can proactively warn of impending airway collapse or hemodynamic instability, providing clinicians with valuable time to intervene. AI-assisted alert systems can analyze ventilator parameters to detect subtle changes that may indicate developing issues. For example, Hezarjaribi et al. (75) demonstrated a ML system could identify early signs of airway deterioration, including respiratory resistance and compliance, that may not be immediately visible to clinicians during mechanical ventilation. Kim et al. (76) developed an AI-based breathing sound analysis system for patients with tracheostomy tubes. The model, utilizing advanced architectures such as MobileNet and Inception_v3, demonstrates sensitivity and specificity above 94%, indicating reliable detection of abnormal breathing sounds.

In addition, AI can enhance decision-making during ventilation. Soundoulounaki et al. (77) employed a neural network-based classifier to detect weak inspiratory efforts by analyzing ventilator waveforms. This approach ensures timely detection of ventilator dyssynchrony, a critical issue that can compromise patient safety. Sottile et al. (78) further validated multiple ML algorithms for identifying ventilator dyssynchrony, providing a solid foundation for AI-driven intraoperative decision support.

These AI-enabled monitoring and decision support systems hold significant potential to improve safety and outcomes in OMF surgery by providing clinicians with early warnings, as well as advanced respiratory monitoring capabilities that are often unavailable through traditional methods such as manual assessment and capnography.

AI in emergence and extubation

In OMF surgery, the decision to extubate at the end of the procedure is critical. The timing of extubation directly impacts postoperative recovery, and in some cases, it can be a life-threatening decision (79). Prolonged mechanical ventilation is associated with a higher risk of postoperative complications, including chest infection (80), postoperative cardiovascular and respiratory complications (81) and oral-mucosal pressure injury (82). The advent of AI offers promising solutions to support clinical decision-making in this critical phase. For example, Lee et al. (83) developed a reinforcement learning-based AI model for controlling ventilation during emergence from general anesthesia. This model can dynamically adjust ventilation parameters to facilitate a smoother transition to spontaneous breathing.

Otaguro et al. (84) successfully predicted extubation success, while Zhao et al. (85) and Huang et al. (86) predicted extubation failure in patients with difficult airways following head, neck, and maxillofacial surgeries. A total of seven clinically relevant features were selected for modeling, including surgical complexity, American Society of Anesthesiologists (ASA) physical status classification, whether tracheal reconstruction was performed, intraoperative blood loss, postoperative delirium, postoperative anemia, and postoperative hypokalemia. The support vector machine (SVM) and LR models demonstrated the best predictive performance, with AUROCs of 0.74 (0.55–0.93) and 0.71 (0.59–0.82). These findings highlight the potential of ML methods to assist in identifying patients at high risk of extubation failure after complex head and neck surgeries. The inclusion of both intraoperative and postoperative variables reflects the multifactorial nature of extubation outcomes, offering a more comprehensive risk assessment compared to traditional criteria.

AI can also evaluate post-extubation conditions, where subtle laryngeal injuries are often under-recognized or misjudged by clinicians (87,88). Pandian et al. (89) used ML for the post-extubation assessment of laryngeal symptoms, while Wang et al. (90) focused on early prediction of non-invasive ventilation failure post-extubation. For patients at risk of difficult weaning, Xu et al. (91) developed a ML model for identifying high-risk cases, enabling proactive management strategies. Collectively, these AI-driven approaches enhance the safety and precision of extubation decisions, optimizing outcomes for OMF surgical patients.

Conclusions

This review highlights the breadth of AI applications in airway management for OMF anesthesia. As the complexity of these procedures increases, interest in AI offers powerful tools for improving safety and precision. However, challenges remain, particularly in data quality, standardization, model interpretability, and clinical validation. Addressing these gaps will be essential for AI to transition from a promising adjunct to an integrate component of airway safety in OMF anesthesia.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the Narrative Review reporting checklist. Available at https://joma.amegroups.com/article/view/10.21037/joma-2025-28/rc

Peer Review File: Available at https://joma.amegroups.com/article/view/10.21037/joma-2025-28/prf

Funding: This work was supported by Jilin Provincial Natural Science Foundation (No. YDZJ202401458ZYTS).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://joma.amegroups.com/article/view/10.21037/joma-2025-28/coif). J.P.W. serves as the unpaid Deputy Editorial-in-Chief of Journal of Oral and Maxillofacial Anesthesia from March 2025 to February 2027. The other authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Warnakulasuriya S. Global epidemiology of oral and oropharyngeal cancer. Oral Oncol 2009;45:309-16. [Crossref] [PubMed]
2. Peacock ZS, Aghaloo T, Bouloux GF, et al. J Oral Maxillofac Surg 2014;72:241-53. [Crossref] [PubMed]
3. King BJ, Levine A. Controversies in Anesthesia for Oral and Maxillofacial Surgery. Oral Maxillofac Surg Clin North Am 2017;29:515-23. [Crossref] [PubMed]
4. Cook TM, Woodall N, Frerk C, et al. Major complications of airway management in the UK: results of the Fourth National Audit Project of the Royal College of Anaesthetists and the Difficult Airway Society. Part 1: anaesthesia. Br J Anaesth 2011;106:617-31. [Crossref] [PubMed]
5. Torrini F, Gendreau S, Morel J, et al. Prediction of extubation outcome in critically ill patients: a systematic review and meta-analysis. Crit Care 2021;25:391. [Crossref] [PubMed]
6. Marchis IF, Negrut MF, Blebea CM, et al. Trends in Preoperative Airway Assessment. Diagnostics (Basel) 2024;14:610. [Crossref] [PubMed]
7. Gaszyński T, Gómez-Ríos MÁ, Serrano-Moraza A, et al. New Devices, Innovative Technologies, and Non-Standard Techniques for Airway Management: A Narrative Review. Healthcare (Basel) 2023;11:2468. [Crossref] [PubMed]
8. Żurek M, Jasak K, Niemczyk K, et al. Artificial Intelligence in Laryngeal Endoscopy: Systematic Review and Meta-Analysis. J Clin Med 2022;11:2752. [Crossref] [PubMed]
9. Liu Y, He Y, Wang X, et al. Advances in airway management in recent 10 years from 2013 to 2023. Anesthesiology and Perioperative Science 2023;1:27.
10. Xia M. Application of Artificial Intelligence in Oral and Maxillofacial Anesthesia. In: Jiang H, Xia M, editors. Anesthesia for Oral and Maxillofacial Surgery. Singapore: Springer Nature Singapore; 2023:327-38.
11. Srivilaithon W, Thanasarnpaiboon P. Performance of machine learning models in predicting difficult laryngoscopy in the emergency department: a single-centre retrospective study comparing with conventional regression method. BMC Emerg Med 2025;25:28. [Crossref] [PubMed]
12. Wang B, Li X, Xu J, et al. Comparison of Machine Learning Models for Difficult Airway. Journal of Anesthesia and Translational Medicine 2023;2:21-8.
13. Xie BH, Li TT, Ma FT, et al. Artificial intelligence in anesthesiology: a bibliometric analysis. Perioper Med (Lond) 2024;13:121. [Crossref] [PubMed]
14. Lonsdale H, Gray GM, Ahumada LM, et al. Machine Vision and Image Analysis in Anesthesia: Narrative Review and Future Prospects. Anesth Analg 2023;137:830-40. [Crossref] [PubMed]
15. Wang J, Xia M, Jiang H. Advances in studies on imaging and artificial intelligence technology-assisted difficult airway assessment. Front Oral Maxillofac Med 2021;3:8.
16. Wilk M, Pikiewicz W, Florczak K, et al. Use of Artificial Intelligence in Difficult Airway Assessment: The Current State of Knowledge. J Clin Med 2025;14:1602. [Crossref] [PubMed]
17. Rosero EB, Corbett J, Mau T, et al. Intraoperative Airway Management Considerations for Adult Patients Presenting With Tracheostomy: A Narrative Review. Anesth Analg 2021;132:1003-11. [Crossref] [PubMed]
18. Ephros H, Shah S, Herrod RJ. A Review of Sedation Agents. Oral Maxillofac Surg Clin North Am 2022;34:21-34. [Crossref] [PubMed]
19. Khan MJ, Karmakar A. Emerging Robotic Innovations and Artificial Intelligence in Endotracheal Intubation and Airway Management: Current State of the Art. Cureus 2023;15:e42625. [Crossref] [PubMed]
20. Zang Q, Cui H, Guo X, et al. Clinical value of video-assisted single-lumen endotracheal intubation and application of artificial intelligence in it. Am J Transl Res 2022;14:7643-52.
21. Koul A, Bawa RK, Kumar Y. Artificial Intelligence Techniques to Predict the Airway Disorders Illness: A Systematic Review. Arch Comput Methods Eng 2023;30:831-64. [Crossref] [PubMed]
22. T A. Machine learning algorithm for ventilator mode selection, pressure and volume control. PLoS One 2024;19:e0299653. [Crossref] [PubMed]
23. van Diepen A, Bakkes THGF, De Bie AJR, et al. A model-based approach to generating annotated pressure support waveforms. J Clin Monit Comput 2022;36:1739-52. [Crossref] [PubMed]
24. García-García F, Lee DJ, Mendoza-Garcés FJ, et al. Reliable prediction of difficult airway for tracheal intubation from patient preoperative photographs by machine learning methods. Comput Methods Programs Biomed 2024;248:108118. [Crossref] [PubMed]
25. Carlson JN, Cho S, Ohu IPN, et al. Preliminary Experience With Inertial Movement Technology to Characterize Endotracheal Intubation Kinematics. Simul Healthc 2020;15:160-6. [Crossref] [PubMed]
26. Shimizu Y, Ohshimo S, Saeki N, et al. New acoustic monitoring system quantifying aspiration risk during monitored anaesthesia care. Sci Rep 2023;13:20196. [Crossref] [PubMed]
27. Shimizu Y, Saeki N, Ohshimo S, et al. Usefulness of new acoustic respiratory sound monitoring with artificial intelligence for upper airway assessment in obese patients during monitored anesthesia care. J Med Invest 2023;70:430-5. [Crossref] [PubMed]
28. García-García F, Lee DJ, Mendoza-Garcés FJ, et al. Automated location of orofacial landmarks to characterize airway morphology in anaesthesia via deep convolutional neural networks. Comput Methods Programs Biomed 2023;232:107428. [Crossref] [PubMed]
29. de Haro C, Santos-Pulpón V, Telías I, et al. Flow starvation during square-flow assisted ventilation detected by supervised deep learning techniques. Crit Care 2024;28:75. [Crossref] [PubMed]
30. Choi SJ, Kim DK, Kim BS, et al. Mask R-CNN based multiclass segmentation model for endotracheal intubation using video laryngoscope. Digit Health 2023;9:20552076231211547. [Crossref] [PubMed]
31. Wu YH, Huang KY, Tseng AC. Development of an Artificial Intelligence-Based Image Recognition System for Time-Sequence Analysis of Tracheal Intubation. Anesth Analg 2024;139:357-65. [Crossref] [PubMed]
32. Obeso I, Yoon B, Ledbetter D, et al. A Novel Application of Spectrograms with Machine Learning Can Detect Patient Ventilator Dyssynchrony. Biomed Signal Process Control 2023;86:105251. [Crossref] [PubMed]
33. Zhu JQ, Wang ML, Li Y, et al. Convolutional neural network based anatomical site identification for laryngoscopy quality control: A multicenter study. Am J Otolaryngol 2023;44:103695. [Crossref] [PubMed]
34. Ren Z, Dinh VS, Wong PM, et al. G2LCPS: End-to-end semi-supervised landmark prediction with global-to-local cross pseudo supervision for airway difficulty assessment. Comput Biol Med 2024;183:109246. [Crossref] [PubMed]
35. Ribeiro MT, Singh S, Guestrin C, editors. "Why should i trust you?": Explaining the predictions of any classifier. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining; 2016.
36. Lundberg SM, Lee SI. A unified approach to interpreting model predictions. Advances in Neural Information Processing Systems 2017;30.
37. Gunning D, Aha D. DARPA’s explainable artificial intelligence (XAI) program. AI Magazine 2019;40:44-58.
38. Choi J, Lee Y, Kang GH, et al. Educational suitability of new channel-type video-laryngoscope with AI-based glottis guidance system for novices wearing personal-protective-equipment. Medicine (Baltimore) 2022;101:e28890. [Crossref] [PubMed]
39. Kim JH, Han SW, Hwang SM, et al. Machine Learning Predictions and Identifying Key Predictors for Safer Intubation: A Study on Video Laryngoscopy Views. J Pers Med 2024;14:902. [Crossref] [PubMed]
40. Matava C, Pankiv E, Raisbeck S, et al. A Convolutional Neural Network for Real Time Classification, Identification, and Labelling of Vocal Cord and Tracheal Using Laryngoscopy and Bronchoscopy Video. J Med Syst 2020;44:44. [Crossref] [PubMed]
41. Jain U, McCunn M, Smith CE, et al. Management of the Traumatized Airway. Anesthesiology 2016;124:199-206. [Crossref] [PubMed]
42. Leonardi R, Lo Giudice A, Farronato M, et al. Fully automatic segmentation of sinonasal cavity and pharyngeal airway based on convolutional neural networks. Am J Orthod Dentofacial Orthop 2021;159:824-835.e1. [Crossref] [PubMed]
43. Sin Ç, Akkaya N, Aksoy S, et al. A deep learning algorithm proposal to automatic pharyngeal airway detection and segmentation on CBCT images. Orthod Craniofac Res 2021;24:117-23. [Crossref] [PubMed]
44. Shujaat S, Jazil O, Willems H, et al. Automatic segmentation of the pharyngeal airway space with convolutional neural network. J Dent 2021;111:103705. [Crossref] [PubMed]
45. Garcia GJM, Catalano D, Shum A, et al. Estimation of Nasal Airway Cross-sectional Area From Endoscopy Using Depth Maps: A Proof-of-Concept Study. Otolaryngol Head Neck Surg 2024;170:1581-9. [Crossref] [PubMed]
46. Jin S, Han H, Huang Z, et al. Automatic three-dimensional nasal and pharyngeal airway subregions identification via Vision Transformer. J Dent 2023;136:104595. [Crossref] [PubMed]
47. Okada Y, Hashimoto K, Ishii W, et al. Development and validation of a model to predict the need for emergency front-of-neck airway procedures in trauma patients. Anaesthesia 2020;75:591-8. [Crossref] [PubMed]
48. Id Said B, Eskander A, Kapadia A, et al. A narrative review of oncologic emergencies in patients with head and neck cancers: initial management and the role of radiation therapy. Ann Palliat Med 2024;13:719-28. [Crossref] [PubMed]
49. Gutiérrez JC, Merino S, de la Calle P, et al. Correlation of preoperative findings in computed axial tomography with the presence of difficult airway in patients undergoing head and neck otorhinolaryngological surgery. Rev Esp Anestesiol Reanim (Engl Ed) 2018;65:252-7. [Crossref] [PubMed]
50. Frerk C, Mitchell VS, McNarry AF, et al. Difficult Airway Society 2015 guidelines for management of unanticipated difficult intubation in adults. Br J Anaesth 2015;115:827-48. [Crossref] [PubMed]
51. Sajayan A, Nair A, McNarry AF, et al. Analysis of a national difficult airway database. Anaesthesia 2022;77:1081-8. [Crossref] [PubMed]
52. Roth D, Pace NL, Lee A, et al. Airway physical examination tests for detection of difficult airway management in apparently normal adult patients. Cochrane Database Syst Rev 2018;5:CD008874. [Crossref] [PubMed]
53. Gemma M, Buratti L, Di Santo D, et al. Pre-operative transnasal endoscopy as a predictor of difficult airway: A prospective cohort study. Eur J Anaesthesiol 2020;37:98-104. [Crossref] [PubMed]
54. Sezari P, Kohzadi Z, Dabbagh A, et al. Unravelling intubation challenges: a machine learning approach incorporating multiple predictive parameters. BMC Anesthesiol 2024;24:453. [Crossref] [PubMed]
55. Zhou CM, Wang Y, Xue Q, et al. Predicting difficult airway intubation in thyroid surgery using multiple machine learning and deep learning algorithms. Front Public Health 2022;10:937471. [Crossref] [PubMed]
56. Yamanaka S, Goto T, Morikawa K, et al. Machine Learning Approaches for Predicting Difficult Airway and First-Pass Success in the Emergency Department: Multicenter Prospective Observational Study. Interact J Med Res 2022;11:e28366. [Crossref] [PubMed]
57. Kim JH, Kim H, Jang JS, et al. Development and validation of a difficult laryngoscopy prediction model using machine learning of neck circumference and thyromental height. BMC Anesthesiol 2021;21:125. [Crossref] [PubMed]
58. Hayasaka T, Kawano K, Kurihara K, et al. Creation of an artificial intelligence model for intubation difficulty classification by deep learning (convolutional neural network) using face images: an observational study. J Intensive Care 2021;9:38. [Crossref] [PubMed]
59. Tavolara TE, Gurcan MN, Segal S, et al. Identification of difficult to intubate patients from frontal face images using an ensemble of deep learning models. Comput Biol Med 2021;136:104737. [Crossref] [PubMed]
60. Wang G, Li C, Tang F, et al. A fully-automatic semi-supervised deep learning model for difficult airway assessment. Heliyon 2023;9:e15629. [Crossref] [PubMed]
61. Kim JH, Jung HS, Lee SE, et al. Improving difficult direct laryngoscopy prediction using deep learning and minimal image analysis: a single-center prospective study. Sci Rep 2024;14:14209. [Crossref] [PubMed]
62. Wu J, Yao Y, Zhang G, et al. Difficult Airway Assessment Based on Multi-View Metric Learning. Bioengineering (Basel) 2024;11:703. [Crossref] [PubMed]
63. Yoo JY, Kang SY, Park JS, et al. Deep learning for anatomical interpretation of video bronchoscopy images. Sci Rep 2021;11:23765. [Crossref] [PubMed]
64. Cho HY, Lee K, Kong HJ, et al. Deep-learning model associating lateral cervical radiographic features with Cormack-Lehane grade 3 or 4 glottic view. Anaesthesia 2023;78:64-72. [Crossref] [PubMed]
65. Xia M, Jin C, Zheng Y, et al. Deep learning-based facial analysis for predicting difficult videolaryngoscopy: a feasibility study. Anaesthesia 2024;79:399-409. [Crossref] [PubMed]
66. Liu X, Flanagan C, Fang J, et al. Comparative analysis of popular predictors for difficult laryngoscopy using hybrid intelligent detection methods. Heliyon 2022;8:e11761. [Crossref] [PubMed]
67. Steffensen TL, Bartnes B, Fuglstad ML, et al. Playing the pipes: acoustic sensing and machine learning for performance feedback during endotracheal intubation simulation. Front Robot AI 2023;10:1218174. [Crossref] [PubMed]
68. Rodiera C, Fortuny H, Valls A, et al. Voice Analysis as a Method for Preoperatively Predicting a Difficult Airway Based on Machine Learning Algorithms: Original Research Report. Health Sci Rep 2024;7:e70246. [Crossref] [PubMed]
69. Eberhart LH, Arndt C, Aust HJ, et al. A simplified risk score to predict difficult intubation: development and prospective evaluation in 3763 patients. Eur J Anaesthesiol 2010;27:935-40. [Crossref] [PubMed]
70. Torborg SR, Kim AYE, Rameau A. New developments in the application of artificial intelligence to laryngology. Curr Opin Otolaryngol Head Neck Surg 2024;32:391-7. [Crossref] [PubMed]
71. Vasconcelos Pereira A, Simões AV, Rego L, et al. New technologies in airway management: A review. Medicine (Baltimore) 2022;101:e32084. [Crossref] [PubMed]
72. Bochkovskiy A, Wang CY, Liao HYM. Yolov4: Optimal speed and accuracy of object detection. arXiv preprint 2020. arXiv:2004.10934.
73. Kim DK, Kim BS, Kim YJ, et al. Development and validation of an artificial intelligence algorithm for detecting vocal cords in video laryngoscopy. Medicine (Baltimore) 2023;102:e36761. [Crossref] [PubMed]
74. Wei W, Huang SW, Chen LH, et al. Airflow behavior changes in upper airway caused by different head and neck positions: Comparison by computational fluid dynamics. J Biomech 2017;52:89-94. [Crossref] [PubMed]
75. Hezarjaribi N, Dutta R, Xing T, et al. Monitoring Lung Mechanics during Mechanical Ventilation using Machine Learning Algorithms. Annu Int Conf IEEE Eng Med Biol Soc 2018;2018:1160-3. [Crossref] [PubMed]
76. Kim H, Koh D, Jung Y, et al. Breathing sounds analysis system for early detection of airway problems in patients with a tracheostomy tube. Sci Rep 2023;13:21029. [Crossref] [PubMed]
77. Soundoulounaki S, Sylligardos E, Akoumianaki E, et al. Neural Network-Enabled Identification of Weak Inspiratory Efforts during Pressure Support Ventilation Using Ventilator Waveforms. J Pers Med 2023;13:347. [Crossref] [PubMed]
78. SottilePDSmithBMossMThe Development, Optimization, and Validation of Four Different Machine Learning Algorithms to Identify Ventilator Dyssynchrony.medRxiv 2023. Preprint. doi:.
79. Difficult Airway Society Extubation Guidelines Group. Difficult Airway Society Guidelines for the management of tracheal extubation. Anaesthesia 2012;67:318-40. [Crossref] [PubMed]
80. Morton RP, Mellow CG, Dorman EB. Chest infection following head and neck surgery: a pilot study. Clin Otolaryngol Allied Sci 1990;15:363-6. [Crossref] [PubMed]
81. Dillon JK, Liu SY, Patel CM, et al. Identifying risk factors for postoperative cardiovascular and respiratory complications after major oral cancer surgery. Head Neck 2011;33:112-6. [Crossref] [PubMed]
82. Choi BK, Kim MS, Kim SH. Risk prediction models for the development of oral-mucosal pressure injuries in intubated patients in intensive care units: A prospective observational study. J Tissue Viability 2020;29:252-7. [Crossref] [PubMed]
83. Lee H, Yoon HK, Kim J, et al. Development and validation of a reinforcement learning model for ventilation control during emergence from general anesthesia. NPJ Digit Med 2023;6:145. [Crossref] [PubMed]
84. Otaguro T, Tanaka H, Igarashi Y, et al. Machine Learning for Prediction of Successful Extubation of Mechanical Ventilated Patients in an Intensive Care Unit: A Retrospective Observational Study. J Nippon Med Sch 2021;88:408-17. [Crossref] [PubMed]
85. Zhao QY, Wang H, Luo JC, et al. Development and Validation of a Machine-Learning Model for Prediction of Extubation Failure in Intensive Care Units. Front Med (Lausanne) 2021;8:676343. [Crossref] [PubMed]
86. Huang H, Wang J, Zhu Y, et al. Development of a Machine-Learning Model for Prediction of Extubation Failure in Patients with Difficult Airways after General Anesthesia of Head, Neck, and Maxillofacial Surgeries. J Clin Med 2023;12:1066. [Crossref] [PubMed]
87. Brodsky MB, Levy MJ, Jedlanek E, et al. Laryngeal Injury and Upper Airway Symptoms After Oral Endotracheal Intubation With Mechanical Ventilation During Critical Care: A Systematic Review. Crit Care Med 2018;46:2010-7. [Crossref] [PubMed]
88. Bishop MJ, Weymuller EA Jr, Fink BR. Laryngeal effects of prolonged intubation. Anesth Analg 1984;63:335-42.
89. Pandian V, Cherukuri SPS, Koneru M, et al. Post-extubation Assessment of Laryngeal Symptoms and Severity (PALSS) in the Intensive Care Unit: Protocol of a Prospective Cohort Study. ORL Head Neck Nurs 2024;42:8-21.
90. Wang H, Zhao QY, Luo JC, et al. Early prediction of noninvasive ventilation failure after extubation: development and validation of a machine-learning model. BMC Pulm Med 2022;22:304. [Crossref] [PubMed]
91. Xu H, Ma Y, Zhuang Y, et al. Machine learning-based risk prediction model construction of difficult weaning in ICU patients with mechanical ventilation. Sci Rep 2024;14:20875. [Crossref] [PubMed]