Related Experiment Video
Updated: Oct 13, 2025
10:44
Three-Dimensional Phase Resolved Functional Lung Magnetic Resonance Imaging
Published on: June 21, 2024
692
Automatic upper airway segmentation in static and dynamic MRI via anatomy-guided convolutional neural networks
Lipeng Xie1,2, Jayaram K Udupa2, Yubing Tong2
1School of Information and Communication Engineering, University of Electronic Science and Technology of China, Chengdu, China.
Medical Physics
|November 13, 2021
Summary
This study introduces a deep learning system for accurate upper airway segmentation in MRI scans. The method achieves high precision, comparable to human experts, for both static and dynamic imaging.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Anatomy
Background:
- Accurate upper airway segmentation in MRI is crucial for anatomical and functional studies.
- Challenges include intensity/shape variability and diverse imaging protocols.
- Existing methods struggle with automatic segmentation accuracy.
Purpose of the Study:
- To develop and validate a comprehensive deep learning system for upper airway segmentation on various MRI data.
- To address the challenges of automatic segmentation in static and dynamic MRI.
- To achieve high accuracy and efficiency in segmenting the upper airway.
Main Methods:
- Utilized static and dynamic MRI datasets from 160 subjects (over 20,000 slices).
- Developed a unified framework using a generalized region-of-interest (GROI) strategy and 2D U-Nets.
- Employed a novel loss function to minimize false positives/negatives and conducted an inter-reader study.
Main Results:
- Achieved high mean Dice coefficients across different MRI types (0.84-0.89).
- Demonstrated excellent agreement between automated and manual delineations.
- Segmentation performance was statistically indistinguishable from human inter-reader variability.
Conclusions:
- The deep learning system enables accurate and efficient upper airway segmentation from static and dynamic MR images.
- The approach shows potential for broader applications in dynamic MRI, such as lung or heart segmentation.
