Related Experiment Video
Updated: Sep 5, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.9K
A Deep Attention-based U-Net for Airways Segmentation in Computed Tomography Images.
Anita Khanna1, Narendra Digambar Londhe1, Shubhrata Gupta1
1Electrical Engineering Department, National Institute of Technology, Raipur 492010, India.
Current Medical Imaging
|July 5, 2022
Summary
This study introduces an attention-based U-Net model for improved airway segmentation in pulmonary disease diagnosis. The method enhances feature extraction for more accurate and efficient lung airway identification.
Area of Science:
- Medical Imaging
- Computer Vision
- Artificial Intelligence
Background:
- Airway segmentation is crucial for diagnosing pulmonary diseases like COPD and bronchiectasis.
- Manual segmentation is challenging due to complex airway structures and varying intensities.
- Deeper airways are particularly difficult to segment accurately.
Purpose of the Study:
- To develop an automated airway segmentation method using deep learning.
- To improve the accuracy and efficiency of segmenting complex lung airways.
Main Methods:
- A convolutional neural network based on the U-Net architecture was proposed.
- An attention block technique was integrated to enhance feature extraction for airways.
- The model was trained and validated on the VESSEL12 and EXACT09 datasets.
Main Results:
- The model achieved high Dice Similarity Coefficient (DSC) scores: 95.21% on EXACT09 and 95.80% on VESSEL12.
- Combined dataset performance yielded a DSC score of 94.1%.
- K-fold cross-validation confirmed the model's generalizability and competitive performance.
Conclusions:
- The attention mechanism effectively highlights relevant airway features and reduces irrelevant ones.
- This approach improves segmentation performance and efficiency.
- The proposed model offers a satisfactory and competitive solution for airway segmentation.
