Information
- Publication Type: PhD-Thesis
- Workgroup(s)/Project(s):
- Date: March 2006
- Date (Start): May 2002
- Date (End): March 2006
- First Supervisor: Eduard Gröller
- Keywords: Vessel Visualization, 3D Modeling, Segmentation, 3D Reconstruction
Abstract
A model is a simplified representation of an object. The modeling stage could be described as shaping individual objects that are later used in the scene. For many years scientists are trying to create an appropriate model of the blood vessels. It looks quite intuitive to believe that a blood vessel can be modeled as a tubular object, and this is true, but the problems appear when you want to create an accurate model that can deal with the wide variability of shapes of diseased blood vessels. From the medical point of view it is quite important to identify, not just the center of the vessel lumen but also the center of the vessel, particularly in the presences of some anomalies, which is the case diseased blood vessels.An accurate estimation of vessel parameters is a prerequisite for automated visualization and analysis of healthy and diseased blood vessels. We believe that a model-based technique is the most suitable one for parameterizing blood vessels. The main focus of this work is to present a new strategy to parameterize diseased blood vessels of the lower extremity arteries.
The first part presents an evaluation of different methods for approximating the centerline of the vessel in a phantom simulating the peripheral arteries. Six algorithms were used to determine the centerline of a synthetic peripheral arterial vessel. They are based on: ray casting using thresholds and a maximum gradient-like stop criterion, pixel-motion estimation between successive images called block matching, center of gravity and shape based segmentation. The Randomized Hough Transform and ellipse fitting have been used as shape based segmentation techniques. Since in the synthetic data set the centerline is known, an estimation of the error can be calculated in order to determine the accuracy achieved by a given method.
The second part describes an estimation of the dimensions of lower extremity arteries, imaged by computed tomography. The vessel is modeled using an elliptical or cylindrical structure with specific dimensions, orientation and CT attenuation values. The model separates two homogeneous regions: Its inner side represents a region of density for vessels, and its outer side a region for background. Taking into account the point spread function of a CT scanner, which is modeled using a Gaussian kernel, in order to smooth the vessel boundary in the model. An optimization process is used to find the best model that fits with the data input. The method provides center location, diameter and orientation of the vessel as well as blood and background mean density values.
The third part presents the result of a clinical evaluation of our methods, as a prerequisite step for being used in clinical environment. To perform this evaluation, twenty cases from available patient data were selected and classified as 'mildly diseased' and 'severely diseased' datasets. Manual identification was used as our reference standard. We compared the model fitting method against a standard method, which is currently used in the clinical environment. In general, the mean distance error for every method was within the inter-operator variability. However, the non-linear model fitting technique based on a cylindrical model shows always a better center approximation in most of the cases, 'mildly diseased' as well as 'severely diseased' cases. Clinically, the non-linear model fitting technique is more robust and presented a better estimation in most of the cases. Nevertheless, the radiologists and clinical experts have the last word with respect to the use of this technique in clinical environment.
Additional Files and Images
Weblinks
No further information available.BibTeX
@phdthesis{Cruz-thesis, title = "3D Modelling and Reconstruction of Peripheral Vascular Structure", author = "Alexandra La Cruz", year = "2006", abstract = "A model is a simplified representation of an object. The modeling stage could be described as shaping individual objects that are later used in the scene. For many years scientists are trying to create an appropriate model of the blood vessels. It looks quite intuitive to believe that a blood vessel can be modeled as a tubular object, and this is true, but the problems appear when you want to create an accurate model that can deal with the wide variability of shapes of diseased blood vessels. From the medical point of view it is quite important to identify, not just the center of the vessel lumen but also the center of the vessel, particularly in the presences of some anomalies, which is the case diseased blood vessels. An accurate estimation of vessel parameters is a prerequisite for automated visualization and analysis of healthy and diseased blood vessels. We believe that a model-based technique is the most suitable one for parameterizing blood vessels. The main focus of this work is to present a new strategy to parameterize diseased blood vessels of the lower extremity arteries. The first part presents an evaluation of different methods for approximating the centerline of the vessel in a phantom simulating the peripheral arteries. Six algorithms were used to determine the centerline of a synthetic peripheral arterial vessel. They are based on: ray casting using thresholds and a maximum gradient-like stop criterion, pixel-motion estimation between successive images called block matching, center of gravity and shape based segmentation. The Randomized Hough Transform and ellipse fitting have been used as shape based segmentation techniques. Since in the synthetic data set the centerline is known, an estimation of the error can be calculated in order to determine the accuracy achieved by a given method. The second part describes an estimation of the dimensions of lower extremity arteries, imaged by computed tomography. The vessel is modeled using an elliptical or cylindrical structure with specific dimensions, orientation and CT attenuation values. The model separates two homogeneous regions: Its inner side represents a region of density for vessels, and its outer side a region for background. Taking into account the point spread function of a CT scanner, which is modeled using a Gaussian kernel, in order to smooth the vessel boundary in the model. An optimization process is used to find the best model that fits with the data input. The method provides center location, diameter and orientation of the vessel as well as blood and background mean density values. The third part presents the result of a clinical evaluation of our methods, as a prerequisite step for being used in clinical environment. To perform this evaluation, twenty cases from available patient data were selected and classified as 'mildly diseased' and 'severely diseased' datasets. Manual identification was used as our reference standard. We compared the model fitting method against a standard method, which is currently used in the clinical environment. In general, the mean distance error for every method was within the inter-operator variability. However, the non-linear model fitting technique based on a cylindrical model shows always a better center approximation in most of the cases, 'mildly diseased' as well as 'severely diseased' cases. Clinically, the non-linear model fitting technique is more robust and presented a better estimation in most of the cases. Nevertheless, the radiologists and clinical experts have the last word with respect to the use of this technique in clinical environment.", month = mar, address = "Favoritenstrasse 9-11/E193-02, A-1040 Vienna, Austria", school = "Institute of Computer Graphics and Algorithms, Vienna University of Technology ", keywords = "Vessel Visualization, 3D Modeling, Segmentation, 3D Reconstruction", URL = "https://www.cg.tuwien.ac.at/research/publications/2006/Cruz-thesis/", }