Related Experiment Video
Updated: Jul 17, 2025
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
Two-stage contextual transformer-based convolutional neural network for airway extraction from CT images
Yanan Wu1, Shuiqing Zhao2, Shouliang Qi3
1College of Medicine and Biological Information Engineering, Northeastern University, Shenyang, China; Key Laboratory of Intelligent Computing in Medical Image, Ministry of Education, Northeastern University, Shenyang, China; Department of Electronic Engineering, The Chinese University of Hong Kong, Hong Kong, China.
This study introduces a new AI framework for segmenting lung airways in CT scans, improving accuracy for chronic obstructive pulmonary disease (COPD) assessment and navigation bronchoscopy planning.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Pulmonary Medicine
Background:
- Accurate airway segmentation in computed tomography (CT) is crucial for diagnosing and managing chronic obstructive pulmonary disease (COPD) and planning navigation bronchoscopy.
- Current segmentation methods struggle with small airway branches and limited data, hindering clinical application.
Purpose of the Study:
- To develop an advanced AI framework for precise airway segmentation in CT images, focusing on both overall airways and small branches.
- To enhance the quantitative assessment of COPD and improve the planning of minimally invasive procedures like navigation bronchoscopy.
Main Methods:
- A novel two-stage framework utilizing a modified 3D U-Net with an integrated 3D contextual transformer block for enhanced feature extraction.
- The framework captures contextual and long-range information, employing a two-stage training process to segment overall airways and then focus on small intrapulmonary branches.
Main Results:
- The proposed method significantly improves airway segmentation performance, extracting more airway branches and longer airway tree lengths compared to existing approaches.
- Demonstrated state-of-the-art results on both in-house and public datasets, validated through extensive quantitative and qualitative analyses.
Conclusions:
- The novel AI framework offers a significant advancement in airway segmentation from CT images, particularly for small branches.
- This improved segmentation capability holds promise for more accurate COPD assessment and enhanced navigation bronchoscopy planning.
