Automatic segmentation of carotid B-mode images using fuzzy classification
Rui Rocha1, Jorge Silva, Aurélio Campilho
1INEB-Instituto de Engenharia Biomédica, Rua Dr. Roberto Frias, s/n, 4200-465 Porto, Portugal. rhr@isep.ipp.pt
Medical & Biological Engineering & Computing
|March 15, 2012
Summary
A novel method automatically segments the common carotid artery in B-mode images using edge detection and fuzzy classification. This approach achieves high accuracy, comparable to manual tracings, for intima-media thickness measurements, especially for the far wall.
Area of Science:
- Medical Imaging
- Ultrasound Technology
- Biomedical Engineering
Background:
- Accurate segmentation of the common carotid artery is crucial for assessing cardiovascular health.
- Intima-media thickness (IMT) is a key indicator of atherosclerosis.
- Existing segmentation methods often require manual intervention or lack real-time processing capabilities.
Purpose of the Study:
- To develop and evaluate a new, fully automatic method for segmenting the common carotid artery in B-mode ultrasound images.
- To assess the accuracy and efficiency of the proposed method for measuring intima-media thickness (IMT).
- To compare the performance of the automatic method against manual tracings by experts.
Main Methods:
- Utilized an instantaneous coefficient of variation edge detector, fuzzy edge classification, and dynamic programming.
- Incorporated discriminating features of intima and adventitia boundaries, including edge strength, intensity gradient orientation, and contextual information.
- The method avoids low-pass filtering through fuzzy edge classification and is designed for real-time processing without user interaction.
Main Results:
- The automatic segmentation method successfully detected both near and far wall boundaries, even in arteries with plaques.
- Automatic detection of the far wall demonstrated accuracy comparable to manual expert tracings for IMT measurement.
- Error coefficients of variation for IMT were [5.6, 6.6%] for automatic far wall detection versus [6.7, 7.1%] for manual, and [11.2, 13.0%] for automatic near wall detection versus [5.9, 9.0%] for manual.
Conclusions:
- The developed automatic segmentation method provides accurate IMT measurements, particularly for the far wall, rivaling manual detection accuracy.
- The method's real-time processing capability (mean 2.1s/image) and high accuracy encourage its application in clinical practice.
- This automated approach offers a promising tool for efficient and reliable carotid artery assessment.
