Искусственный интеллект в диагностике трудных дыхательных путей: обзор литературы. Artificial intelligence for difficult airways diagnosis: review.

INTRODUCTION. The development of artificial intelligence has opened up the possibility of its application in the practice of an anesthesiologist in the direction associated with the most life-threatening complications - the prediction of difficult airways. This article is about the principles of artificial intelligence and the experience of its modern application to determine the risk of difficult airways.
OBJECTIVES. To review the literature to determine the role of artificial intelligence in the diagnosis of difficult airways.
MATERIALS AND METHODS. A review of the literature on the international Pubmed database, the Russian-language Elibrary. The search words for English-language databases were: Artificial intelligence, deep learning, difficult airways; for Russian-language: искусственный интеллект, глубокое машинное обучение, трудные дыхательные пути. The criteria for inclusion of articles were: systematic reviews, meta-analysis, randomized clinical trials, review articles. Exclusion criteria: clinical case, dissertation, abstract, thesis, application of artificial intelligence methods in pediatric practice.
RESULTS. 9 articles were received for analysis. The main methods of searching for predictors of difficult airways are based on the use of photographs of patients, the use of anthropometry and physical examination data, methods using thermal imager heat maps using gradient-weighted class activation mapping. In all the analyzed works, the effectiveness of predicting difficult airways using artificial intelligence was noted, with the exception of the Siriussawakul et al. study and related to the anthropometric characteristics of patients.
CONCLUSION. Diagnostic methods based on the artificial intelligence in the practice of the anesthesiologist make it easier to work and improve the detection of patients with difficult airways. However, there are still a number of unresolved issues regarding the legal and ethical components of the application of these methods in clinical practice.

{'INTRODUCTION': '引言。人工智能的发展为麻醉实践中应对最危及生命并发症——即困难气道预测——的应用提供了可能性。本文旨在探讨人工智能的基本原理及其在现代应用中预测困难气道的风险方面的经验。', 'OBJECTIVES': '目标。通过对文献的回顾，确定人工智能在困难气道诊断中的作用。', 'MATERIALS_AND_METHODS': '材料与方法。对国际PubMed数据库以及俄语Elibrary上的文献进行了回顾。英文数据库的搜索词为：人工智能、深度学习、困难气道；俄语搜索词为：искусственный интеллект、глубокое машинное обучение、трудные дыхательные пути。纳入标准为：系统综述、荟萃分析、随机临床试验、综述文章。排除标准：临床案例、学位论文、摘要、论文、人工智能方法在儿科实践中的应用。', 'RESULTS': '结果。共收到9篇文章进行分析。寻找困难气道预测指标的主要方法基于患者照片的使用、人体测量学和体检数据的运用，以及使用热成像仪的热图和基于梯度加权的类激活映射方法。在所有分析的工作中，都注意到了使用人工智能预测困难气道的有效性，除了Siriussawakul等人关于患者人体测量学特征的研究。', 'CONCLUSION': '结论。基于人工智能的诊疗方法在麻醉实践中简化了工作流程，并提高了对困难气道患者的检测能力。然而，关于这些方法在临床实践中的应用的法律法规和伦理问题仍存在许多未解决的问题。'}