Related Experiment Video
Updated: Jul 10, 2026
07:31
Image Acquisition using Portable Sonography for Emergency Airway Management
Published on: September 28, 2022
Machine learning-based prediction of difficult laryngoscopy in infants with Pierre Robin sequence using quantitative
Danling Hu1, Weiwei Cai1, Anwen Zheng1
1Department of Anesthesiology, Children's Hospital of Nanjing Medical University, Nanjing, China.
Frontiers in Neurology
|July 9, 2026
Summary
Quantitative 3D-CT scans identify key airway predictors for difficult laryngoscopies in infants with Pierre Robin sequence (PRS). Machine learning models, particularly Extra Trees, show promise in predicting these challenging exposures.
Area of Science:
- Medical Imaging
- Pediatric Airway Management
- Computational Biology
Background:
- Infants with Pierre Robin sequence (PRS) often face difficult laryngoscopic exposure due to craniofacial abnormalities.
- Quantitative three-dimensional computed tomography (3D-CT) offers objective airway assessment for PRS patients.
- Predicting laryngoscopic difficulty is crucial for optimizing surgical planning and patient outcomes.
Purpose of the Study:
- To identify key 3D-CT parameters associated with difficult laryngoscopic exposure in infants with PRS.
- To develop and externally validate machine learning (ML) models for predicting laryngoscopic difficulty.
- To improve preoperative airway assessment in infants with PRS.
Main Methods:
- Retrospective analysis of 214 infants with PRS undergoing mandibular distraction osteogenesis.
- Classification of patients into easy (Cormack-Lehane grades I-II) and difficult (grades III-IV) laryngoscopic exposure groups.
- Development and validation of seven ML models using quantitative 3D-CT parameters, including Extra Trees and XGBoost.
Main Results:
- Tongue length, tongue base-posterior pharyngeal wall distance, sagittal oropharyngeal area, and tongue base-epiglottic angle were identified as independent predictors.
- The Extra Trees model demonstrated superior generalizability in the temporal validation cohort (AUC=0.876, accuracy=0.812, F1-score=0.805).
- The Extra Trees model showed excellent calibration and substantial net clinical benefit across a wide range of threshold probabilities.
Conclusions:
- Quantitative 3D-CT parameters of tongue morphology and oropharyngeal dimensions are significant predictors of difficult laryngoscopic exposure in PRS infants.
- The Extra Trees ML model shows potential for accurate and generalizable prediction of laryngoscopic difficulty in this population.
- These findings can enhance preoperative airway assessment and surgical planning for infants with PRS.
