3. An unsupervised graph-based approach for the representation of coronary arteries in X-ray angiograms
-
Abstract
This paper presents a novel unsupervised framework for the graph-based representation of coronary arteries in X-ray angiograms. The framework consists of the steps of vessel detection, segmentation, and skeleton simplification. Vessel detection is performed by Gaussian matched filters (GMF) trained by the univariate marginal distribution algorithm for a continuous domain. The detection results are evaluated in terms of the area (Az) under the receiver operating characteristic curve. The second step is focused on the binary classification of the response obtained from the GMF method. In the final step, the vessel skeleton simplification is carried out by using the Ramer-Douglas-Peucker algorithm. During the computational experiments, the proposed framework obtained a detection performance of Az = 0.926. The interclass variance method was selected from five state-of-the-art thresholding methods according to its classification accuracy on the segmentation of the detection response (0.923). In the final step, a graph-based structure of the coronary arteries with a data compression ratio of 0.954 is obtained. Based on the experimental results, the proposed method has demonstrated that it is suitable for a variety of applications in computer-aided diagnosis.
Abstract
This paper presents a novel unsupervised framework for the graph-based representation of coronary arteries in X-ray angiograms. The framework consists of the steps of vessel detection, segmentation, and skeleton simplification. Vessel detection is performed by Gaussian matched filters (GMF) trained by the univariate marginal distribution algorithm for a continuous domain. The detection results are evaluated in terms of the area (Az) under the receiver operating characteristic curve. The second step is focused on the binary classification of the response obtained from the GMF method. In the final step, the vessel skeleton simplification is carried out by using the Ramer-Douglas-Peucker algorithm. During the computational experiments, the proposed framework obtained a detection performance of Az = 0.926. The interclass variance method was selected from five state-of-the-art thresholding methods according to its classification accuracy on the segmentation of the detection response (0.923). In the final step, a graph-based structure of the coronary arteries with a data compression ratio of 0.954 is obtained. Based on the experimental results, the proposed method has demonstrated that it is suitable for a variety of applications in computer-aided diagnosis.
Chapters in this book
- Frontmatter I
- Preface V
- Contents XI
- 1. Medical color image enhancement: Problems, challenges & recent techniques 1
- 2. Exploring the scope of intelligent algorithms for various community detection techniques 19
- 3. An unsupervised graph-based approach for the representation of coronary arteries in X-ray angiograms 43
- 4. A study of recent trends in content based image classification 65
- 5. Intelligent monitoring and evaluation of digital geometry figures drawn by students 95
- 6. Rough set and soft set models in image processing 123
- 7. Quantum inspired simulated annealing technique for automatic clustering 145
- 8. Intelligent greedy model for influence maximization in multimedia data networks 167
- Index 183
Chapters in this book
- Frontmatter I
- Preface V
- Contents XI
- 1. Medical color image enhancement: Problems, challenges & recent techniques 1
- 2. Exploring the scope of intelligent algorithms for various community detection techniques 19
- 3. An unsupervised graph-based approach for the representation of coronary arteries in X-ray angiograms 43
- 4. A study of recent trends in content based image classification 65
- 5. Intelligent monitoring and evaluation of digital geometry figures drawn by students 95
- 6. Rough set and soft set models in image processing 123
- 7. Quantum inspired simulated annealing technique for automatic clustering 145
- 8. Intelligent greedy model for influence maximization in multimedia data networks 167
- Index 183
