Development and validation of an artificial intelligence algorithm for detecting vocal cords in video laryngoscopy

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Emergency Medicine

Background:

• Airway management is critical in life-threatening emergencies.
• Video laryngoscopy (VL) is a key tool for endotracheal intubation (ETI) in emergency departments.
• Artificial intelligence (AI) shows promise in identifying anatomical structures from medical images.

Purpose of the Study:

• To develop and validate an AI algorithm for detecting vocal cords in VL images.
• To assess the algorithm's performance in emergent intubation scenarios.
• To enhance the safety and efficiency of ETI procedures through AI assistance.

Main Methods:

• A retrospective study utilizing VL images from emergency department intubations.
• Development of an AI algorithm using the YOLOv4 model on a training dataset.
• Validation of the algorithm on separate validation and test datasets, including subgroup analysis.

Main Results:

• The AI algorithm achieved an F1 score of 0.906, sensitivity of 0.963, and specificity of 0.842 on the validation set.
• Performance on the test set yielded an F1 score of 0.808, sensitivity of 0.823, and specificity of 0.804.
• The algorithm demonstrated high accuracy in detecting vocal cords from VL images.

Conclusions:

• An AI algorithm capable of detecting vocal cords from VL images was successfully developed and validated.
• The algorithm exhibits high performance, offering a valuable tool for clinical use.
• This AI application can improve the safety of endotracheal intubation by ensuring accurate vocal cord identification.