Predicting Difficult Tracheal Intubation Using Multi-Angle Photographic Analysis with Convolutional Neural Networks and EfficientNet

Abstract

Background: Difficult intubation is an important clinical problem faced by anesthesiologists and is one of the most important causes of anesthesia-related morbidity. According to various sources, the frequency of encountering a difficult airway is stated as 1–4%. Aim: We thought that difficult tracheal intubation could be predicted by photographic analysis using artificial intelligence. Methods: Sixteen photographs were taken in the preoperative period in the sitting and lying positions anteriorly, laterally, with the mouth open, with the mouth closed, with the neck straight, and with the neck extended. Intubations performed without intervention for the first time were considered easy. Intubations with external tracheal intervention and with more than one attempt were evaluated as medium. Intubations requiring more than three attempts; intubation with stylets, fiberoptic bronchoscopes, or video laryngoscopes; or cases in which patients could not be intubated and provided airway with a laryngeal mask were considered difficult. Results: In our study, the CNN (convolutional neural network) model performed well overall, with the best results generally obtained using batch sizes of 32 and 128 and learning rates ranging from 0.1 to 0.001. Conclusions: The prominent aspects of our study are that it can be conducted with an easily accessible mobile phone, can be performed at the bedside, and is successful in predicting difficult intubation. The sensitivity of methods currently used to assess difficult airways is generally low, and the likelihood of clinicians successfully identifying this condition using available information varies widely; thus far, there is no gold standard for prediction. We believe that our study will bring a different perspective to estimating the difficulty of intubation, which occupies a very important place in anesthesia practice.

Ähnliche Arbeiten