An ultrasound-based artificial intelligence framework for difficult airway prediction: A two-model, three-step...

Area of Science:

• Anesthesiology
• Medical Imaging
• Artificial Intelligence

Background:

• Difficult airway management presents significant clinical challenges.
• Ultrasound parameters and deep learning show potential for difficult airway assessment.
• An AI framework for difficult airway prediction is needed.

Purpose of the Study:

• To develop and internally validate an ultrasound-based AI framework for predicting difficult airways.
• To construct a "two-model, three-step" hierarchical strategy for risk stratification.
• To evaluate the performance of AI models in difficult airway prediction.

Main Methods:

• A cohort of 903 patients undergoing elective general anesthesia was studied.
• Two convolutional neural network models, CL-AI and VIDIAC-AI, were developed using neck ultrasound images.
• Model performance was assessed using cross-validation and an independent test set.

Main Results:

• Difficult laryngoscopy occurred in 20.9% (direct) and 5.5% (video) of patients.
• The CL-AI model achieved an AUC of 0.86 and accuracy of 0.84.
• The VIDIAC-AI model achieved an AUC of 0.82 and accuracy of 0.81.

Conclusions:

• An ultrasound-based AI framework was developed for difficult airway risk stratification.
• The "Two-Model, Three-Step" framework serves as a clinical decision-support tool.
• Further validation in large multicenter cohorts is required.