Automatic endotracheal tube position confirmation system based on image classification--a preliminary assessment

Area of Science:

• Medical Devices
• Artificial Intelligence
• Anesthesiology

Background:

• Endotracheal intubation is critical for ventilation but carries risks of misplacement.
• Accurate confirmation of endotracheal tube (ETT) position is essential to prevent potentially fatal complications.
• Current methods for ETT confirmation can be subjective or invasive.

Purpose of the Study:

• To introduce and evaluate a novel system for automated endotracheal tube position confirmation.
• To develop an algorithm for classifying video images of anatomical structures during intubation.
• To improve the safety and accuracy of endotracheal intubation procedures.

Main Methods:

• A system integrating a CMOS sensor on a stylet with a digital signal processor (DSP) was developed.
• Video signals were processed using a confirmation algorithm based on image classification (Gaussian Mixture Models).
• The system identified anatomical structures like the trachea and esophagus to determine ETT placement accuracy.

Main Results:

• The system achieved high accuracy in classifying endotracheal tube positions.
• On a dataset of cow intubations, 95.6% accuracy was observed (1530/1600 images).
• On human subjects, the system achieved 98.0% accuracy (351/358 images), with extremely rare fatal misclassifications.

Conclusions:

• The novel CMOS sensor-based system offers a highly accurate and reliable method for endotracheal tube confirmation.
• The AI-driven image classification approach demonstrates significant potential for enhancing patient safety during intubation.
• This technology could reduce complications arising from incorrect endotracheal tube placement in clinical practice.