An Intelligent Augmented Reality Training Framework for Neonatal Endotracheal Intubation

Area of Science:

• Medical Simulation
• Augmented Reality
• Machine Learning

Background:

• Neonatal endotracheal intubation (ETI) is a critical skill requiring extensive practice.
• Current manikin-based training lacks real-time procedural visualization and expert guidance, leading to prolonged learning curves.
• Limited instructor availability further hinders effective ETI training.

Purpose of the Study:

• To develop an intelligent augmented reality (AR) training framework for neonatal endotracheal intubation (ETI).
• To provide trainees with real-time procedural visualization and automated performance assessment.
• To enhance the efficiency and effectiveness of ETI training for medical professionals.

Main Methods:

• Developed an AR framework capturing laryngoscope and manikin motion data.
• Implemented 3D see-through visualization via a head-mounted display (HMD).
• Utilized an attention-based Convolutional Neural Network (CNN) for performance assessment and motion analysis.
• Integrated interpretable, color-coded feedback based on the ETI scoring rubric.

Main Results:

• The AR system provides 3D see-through visualization for real-time procedural guidance.
• The CNN model achieved an 84.6% classification accuracy in assessing ETI performance.
• Automated feedback highlights areas needing improvement through color-coded motion trajectories.

Conclusions:

• The proposed AR training framework effectively visualizes the ETI procedure for enhanced learning.
• Automated performance assessment using CNNs offers objective and interpretable feedback.
• This intelligent training system has the potential to significantly shorten the learning curve for neonatal endotracheal intubation.