Related Experiment Videos
Segmentation of coronary arteriograms by iterative ternary classification.
1Department of Electrical Engineering, College of Engineering, University of Rhode Island, Kingston 02881.
IEEE Transactions on Bio-Medical Engineering
|August 1, 1990
Summary
This study introduces an iterative ternary classification algorithm for segmenting coronary arteries in angiograms. The novel algorithm mimics human vision, achieving superior performance across various image qualities.
Area of Science:
- Medical Imaging
- Computer Vision
- Biomedical Engineering
Background:
- Coronary angiograms are crucial for diagnosing heart conditions.
- Accurate segmentation of arterial structures is essential for analysis.
- Existing segmentation algorithms have limitations in performance and adaptability.
Purpose of the Study:
- To develop and evaluate a novel segmentation algorithm for coronary arterial structures.
- To mimic human visual interpretation for improved segmentation accuracy.
- To compare the algorithm's performance against existing methods.
Main Methods:
- An iterative ternary classification and learning algorithm was developed.
- Two gray-scale thresholds were computed to classify pixels into artery, background, or undecided.
- Threshold adaptation used a learning algorithm based on line and consistency measurements.
Main Results:
- The iterative ternary classifier demonstrated superior performance compared to relaxation and scattering-based algorithms.
- Quantitative analysis using computer-generated images validated the algorithm's effectiveness.
- The algorithm showed robust performance across a broad range of coronary angiogram image qualities.
Conclusions:
- The developed iterative ternary classification algorithm offers an effective method for coronary artery segmentation.
- The algorithm's approach, inspired by human vision, enhances segmentation accuracy.
- Visualization and user interaction are valuable tools in developing and refining medical imaging algorithms.