Related Experiment Video
Updated: Sep 17, 2025
10:44
Three-Dimensional Phase Resolved Functional Lung Magnetic Resonance Imaging
Published on: June 21, 2024
624
Pulmonary vessel segmentation in computed tomography images: a cascaded approach combining U-Net and
Zhaofeng Xue1, Ying Sun2, Guiyuan Tong1
1Department of Electrical Engineering, Shenyang University of Technology, Shenyang, China.
Quantitative Imaging in Medicine and Surgery
|July 3, 2025
Summary
This study introduces a cascaded algorithm combining U-Net and parameter-adaptive fully connected conditional random fields (PA-FCCRFs) for improved pulmonary vessel segmentation in CT images, enhancing diagnostic accuracy.
Area of Science:
- Medical Imaging
- Computer Vision
- Biomedical Engineering
Background:
- Accurate pulmonary vessel segmentation is crucial for diagnosing lung diseases and tailoring treatment plans.
- Current methods using convolutional neural networks (CNNs) can struggle with long-distance pixel dependencies, leading to segmentation errors.
- This study addresses the need for improved pulmonary vessel segmentation in computed tomography (CT) images.
Purpose of the Study:
- To develop and validate a cascaded algorithm for enhancing the accuracy of pulmonary vessel segmentation in CT images.
- To improve the efficiency and precision of diagnosing pulmonary diseases through advanced imaging analysis.
- To reduce medical resource waste by enabling more targeted patient treatment plans.
Main Methods:
- A cascaded model integrating U-Net for initial segmentation and parameter-adaptive fully connected conditional random fields (PA-FCCRFs) for refinement was proposed.
- U-Net performed preliminary segmentation of pulmonary vessels within the lung region.
- PA-FCCRFs were incorporated to address limitations in modeling long-distance pixel dependencies inherent in CNNs, utilizing Bayesian optimization for parameter tuning.
Main Results:
- The cascaded PA-FCCRFs approach significantly improved segmentation precision from 73.14±10.67 to 90.24±4.63 and F1 score from 82.67±6.86 to 91.85±3.41.
- Hausdorff distance, a measure of segmentation accuracy, was reduced from 35.12±6.04 to 30.86±2.71.
- Validation using AH-Net and V-Net demonstrated substantial accuracy enhancements in CNN-based vascular segmentation after PA-FCCRFs optimization.
Conclusions:
- The proposed cascaded PA-FCCRFs method effectively segments pulmonary vessels, offering a promising tool for clinical applications.
- The algorithm supports more accurate diagnosis of pulmonary diseases.
- This advanced segmentation technique has the potential to optimize treatment strategies and reduce healthcare costs.
Keywords:
Bayesian optimizationPulmonary vessel segmentationconvolutional networks for biomedical image segmentation (U-Net)fully connected conditional random fields (FCCRFs)
