Related Experiment Video
Updated: Jul 18, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.8K
Deep Learning Models for Automatic Upper Airway Segmentation and Minimum Cross-Sectional Area Localisation in
Guang Chu1, Rongzhao Zhang2, Yingqing He2
1Orthodontics, Division of Paediatric Dentistry and Orthodontics, Faculty of Dentistry, The University of Hong Kong, Hong Kong SAR, China.
Bioengineering (Basel, Switzerland)
|August 26, 2023
Summary
This study developed an AI system for automatic upper airway segmentation and minimum cross-sectional area (CSAmin) localization in 2D radiographic images, proving efficient and accurate.
Area of Science:
- Radiology
- Artificial Intelligence
- Medical Imaging Analysis
Background:
- Accurate upper airway segmentation and minimum cross-sectional area (CSAmin) localization are crucial for diagnosing airway diseases.
- Manual analysis of 2D radiographic airway images is time-consuming and prone to variability.
Purpose of the Study:
- To develop and validate convolutional neural network (CNN) algorithms for automated upper airway segmentation.
- To enable automatic localization of CSAmin in 2D radiographic airway images.
Main Methods:
- Trained four AI models (UNet18, UNet36, DeepLab50, DeepLab101) on 161 2D airway images from cone-beam computed tomography (CBCT).
- Evaluated segmentation performance using precision, recall, Intersection over Union, Dice similarity coefficient, and size difference.
- Assessed CSAmin localization accuracy by comparing AI-derived heights with manually determined values from 3D CBCT data.
Main Results:
- All four AI segmentation models achieved precision exceeding 90.0% with no significant accuracy differences.
- AI demonstrated high consistency (0.944) with manual CSAmin localization.
- AI processing significantly reduced the time for airway segmentation and CSAmin localization compared to manual methods.
Conclusions:
- A fully automatic AI-driven system for upper airway segmentation and CSAmin localization using 2D radiographic images was successfully developed and validated.
- The AI system offers an efficient and accurate alternative to manual analysis for clinical applications.
