Abstract
Threedimensional image reconstruction from projections corresponds to a set of techniques which give information on the inner structure of the studied object. These techniques are mainly used in medical imaging or in non destructive evaluation. Image reconstruction is an ill-posed problem. So the inversion has to be regularized. This thesis deals with the introduction of a priori information within the reconstruction algorithm. The knowledge is introduced through an object model. The proposed scheme is applied to the medical domain for cone beam geometry. We address two specific problems. First, we study the reconstruction of high contrast objects. This can be applied to bony morphology (bone/soft tissue) or to angiography (vascular structures opacified by injection of contrast agent). With noisy projections, the filtering steps of standard methods tend to smooth the natural transitions of the investigated object. In order to regularize the reconstruction but to keep contrast, we introduce a model of classes which involves the Markov random fields theory. We develop a reconstruction scheme: analytic reconstruction-reprojection. Then, we address the case of an object changing during the acquisition. This can be applied to angiography when the contrast agent is moving through the vascular tree. The problem is then stated as
More>>
Citation Formats
Delageniere-Guillot, S.
Use of an object model in three dimensional image reconstruction. Application in medical imaging; Utilisation d`un modele d`objet en reconstruction d`image tridimensionnelle. Application a l`imagerie medicale.
France: N. p.,
1993.
Web.
Delageniere-Guillot, S.
Use of an object model in three dimensional image reconstruction. Application in medical imaging; Utilisation d`un modele d`objet en reconstruction d`image tridimensionnelle. Application a l`imagerie medicale.
France.
Delageniere-Guillot, S.
1993.
"Use of an object model in three dimensional image reconstruction. Application in medical imaging; Utilisation d`un modele d`objet en reconstruction d`image tridimensionnelle. Application a l`imagerie medicale."
France.
@misc{etde_10136781,
title = {Use of an object model in three dimensional image reconstruction. Application in medical imaging; Utilisation d`un modele d`objet en reconstruction d`image tridimensionnelle. Application a l`imagerie medicale}
author = {Delageniere-Guillot, S}
abstractNote = {Threedimensional image reconstruction from projections corresponds to a set of techniques which give information on the inner structure of the studied object. These techniques are mainly used in medical imaging or in non destructive evaluation. Image reconstruction is an ill-posed problem. So the inversion has to be regularized. This thesis deals with the introduction of a priori information within the reconstruction algorithm. The knowledge is introduced through an object model. The proposed scheme is applied to the medical domain for cone beam geometry. We address two specific problems. First, we study the reconstruction of high contrast objects. This can be applied to bony morphology (bone/soft tissue) or to angiography (vascular structures opacified by injection of contrast agent). With noisy projections, the filtering steps of standard methods tend to smooth the natural transitions of the investigated object. In order to regularize the reconstruction but to keep contrast, we introduce a model of classes which involves the Markov random fields theory. We develop a reconstruction scheme: analytic reconstruction-reprojection. Then, we address the case of an object changing during the acquisition. This can be applied to angiography when the contrast agent is moving through the vascular tree. The problem is then stated as a dynamic reconstruction. We define an evolution AR model and we use an algebraic reconstruction method. We represent the object at a particular moment as an intermediary state between the state of the object at the beginning and at the end of the acquisition. We test both methods on simulated and real data, and we prove how the use of an a priori model can improve the results. (author).}
place = {France}
year = {1993}
month = {Feb}
}
title = {Use of an object model in three dimensional image reconstruction. Application in medical imaging; Utilisation d`un modele d`objet en reconstruction d`image tridimensionnelle. Application a l`imagerie medicale}
author = {Delageniere-Guillot, S}
abstractNote = {Threedimensional image reconstruction from projections corresponds to a set of techniques which give information on the inner structure of the studied object. These techniques are mainly used in medical imaging or in non destructive evaluation. Image reconstruction is an ill-posed problem. So the inversion has to be regularized. This thesis deals with the introduction of a priori information within the reconstruction algorithm. The knowledge is introduced through an object model. The proposed scheme is applied to the medical domain for cone beam geometry. We address two specific problems. First, we study the reconstruction of high contrast objects. This can be applied to bony morphology (bone/soft tissue) or to angiography (vascular structures opacified by injection of contrast agent). With noisy projections, the filtering steps of standard methods tend to smooth the natural transitions of the investigated object. In order to regularize the reconstruction but to keep contrast, we introduce a model of classes which involves the Markov random fields theory. We develop a reconstruction scheme: analytic reconstruction-reprojection. Then, we address the case of an object changing during the acquisition. This can be applied to angiography when the contrast agent is moving through the vascular tree. The problem is then stated as a dynamic reconstruction. We define an evolution AR model and we use an algebraic reconstruction method. We represent the object at a particular moment as an intermediary state between the state of the object at the beginning and at the end of the acquisition. We test both methods on simulated and real data, and we prove how the use of an a priori model can improve the results. (author).}
place = {France}
year = {1993}
month = {Feb}
}