Automated Assessment System for Neonatal Endotracheal Intubation Using Dilated Convolutional Neural Network

Area of Science:

• Medical Simulation
• Artificial Intelligence in Healthcare
• Pediatric Resuscitation Training

Background:

• Neonatal endotracheal intubation (ETI) is a critical resuscitation skill requiring extensive practice.
• Current ETI training relies on physical manikins and subjective expert assessment, limiting training opportunities.
• Automating ETI assessment is challenging due to complex feature identification and feedback generation.

Purpose of the Study:

• To develop an automated assessment model for neonatal endotracheal intubation (ETI).
• To provide objective scoring and performance feedback to pediatric trainees using AI.
• To overcome limitations in expert instructor availability for ETI training.

Main Methods:

• A dilated Convolutional Neural Network (CNN) was developed for ETI assessment.
• The model processes kinematic multivariate time-series (MTS) data from a manikin-based system.
• Class Activation Mapping (CAM) was used for motion-based feedback visualization.

Main Results:

• The AI model automatically extracts crucial features from MTS data for assessment.
• The system provides an overall score and identifies impactful motions via CAM visualization.
• The model achieved an average classification accuracy of 92.2% using Leave-One-Out-Cross-Validation (LOOCV).

Conclusions:

• The proposed CNN-based model offers an effective solution for automated ETI skill assessment.
• The system provides valuable, objective feedback to trainees, enhancing the learning process.
• This AI approach can significantly improve the scalability and consistency of neonatal resuscitation training.