Stent Graft Change Detection After Endovascular Abdominal Aortic Aneurysm Repair.

Full text

Page 1

Stent Graft Change Detection after Endovascular
Abdominal Aortic Aneurysm Repair.
Josu Maiora1 , Guillermo García2, Arantxa Tapia2, Iván Macía3, Jon Haitz
Legarreta3, Céline Paloc3, Manuel Graña4 and Mariano de Blas5,


1 Electronics and Telecomunications Department, 2 Engineering Systems and Automatic
Department, Technical University School, University of the Basque Country, Donostia-San
Sebastián, Spain
{j.maiora, g.garcia, arantxa.tapia}@ehu.es

3 Biomedical Applications Department, VICOMTech, Donostia-San Sebastián, Spain
{imacia, jhlegarreta, cpaloc}@vicomtech.org

4 Computational Intelligence Group, Computer Science Faculty, University of the Basque
Country, Donostia-SanSebastián, Spain
manuel.grana@ehu.es

5 Interventional Radiology Service, Donostia Hospital, Donostia-SanSebastián, Spain
mariano.deblasbravo@osakidetza.net

Abstract. The use of the endovascular prostheses in Abdominal Aortic
Aneurysm (EVAR) has proven to be an effective technique to reduce the
pressure and rupture risk of aneurysms. Nevertheless, in a long-term
perspective different complications, such as prosthesis displacement or leaks
inside the aneurysm sac (endoleaks), could appear, causing a pressure elevation
and increasing the danger of rupture consequently. At present Computed
Tomography Angiography (CTA) is the most commonly used examination for
imaging surveillance of stent graft monitoring. However, endoleak
complications cannot always be detected by visual inspection on CTA scans.
The purpose of this work was to study the application of image registration
techniques to the detection of changes in the stent graft. Previously we segment
the lumen using semi-automatic methods.
Keywords: image processing, aneurysm, endovascular, registration.
1. Introduction
Abdominal Aortic Aneurysm (AAA) [1] [2] [3] is a focal dilation in some point of the
section of the abdominal aorta. Several treatments exist today; one option is a
minimally invasive surgical procedure called Endovascular Aneurysm Repair
(EVAR), in which an endovascular prosthesis (endovascular graft) is inserted to

Page 2

exclude the aneurysm from blood circulation [4].

The EVAR (Fig.1) requires a postoperative follow-up to ensure that stent is stable
(no leakage, i.e.: blood flow within the aneurysmatic sac) and that the bag was
properly excluded from circulation. A reduction in the size of the aneurysm shows
that exclusion has been effective [5] [6], while an expansion and/or the presence of
a leakage are indicators that the treatment has not been effective, continuing the riskof
blood vessel rupture at the aneurysm location.



Fig. 1. Endovascular graft

An analysis of the FDA (Food and Drug Administration) estimated that the mortality
rate after EVAR is 0.4% annually, compared to 0.18% per annum after open surgery.
Therefore, follow-up after EVAR has become an important part of patient care.
The most widely used technique for EVAR monitoring is to obtain Computerized
Tomography (CT) images of the abdominal region with intravenous contrast agent.
Such images are scans of the patient's abdominal area and they are available in the
clinical routine as a set of 2D images that are difficult to interpret and analyze. The
aim of our work is to make an automatic analysis of the AAA with visual and
quantitative information for monitoring and tracking of patients who underwent
EVAR, using digital image processing techniques, which are currently in an advanced
state of the art to deal with this problem.
There are several approaches to perform a more accurate EVAR monitoring, one of
them is the analysis of the density of the material inside the aneurysmatic sac
(thrombus). It detects the various changes experienced in the interior of the aneurysm
by analyzing changes in gray levels or image texture. Quantitative assessment of
temporal intensity profiles for analysis of trends in EVAR behavior and classification
of its states is performed in [8].
The approach that we follow in this paper is to estimate the rigid motion of the
stent relative to the spinal cord [9] as well as its deformation [10]. These data can
identify patterns for a high probability of dangerous progression of the aneurysm. The
long-term goal is to make a prediction about future complications and disease

Page 3

progression. Because the processes are complex and data from medical devices are
difficult to assess visually, a computer-assisted analysis is required. In the current
state of the art, the morphological changes and migration of the stent and the
aneurysm after EVAR are not studied in a systematic manner to deal with the
complexity of these movements.
The segmentation methods used in the state of the art publications are not easily
reproducible. Here we use semiautomatic methods that have a proven intrarater and
interrater reliability and validity. Segmentation of the spinal canal is based on User
Guided Level Set Segmentation method. The same method has been used for the
segmentation of the lumen (whose controur is limited by the stent graft, and thus
indicates any transformation in it). Aneurysm outer wall or thrmobus segmentation
has proven to be more difficult. A novel technique for thrombus segmentation, based
on region growing algorithms, centerline extraction and radial functions, has been
started to be evaluated.
Furthermore, registration methods used by state of the art works are based in point
sets registration methods that always suffer a loss of information, therefore in our
pipeline, registration is performed over binary images with much less information
loss.




Fig 2.: Pipeline of the process stent graft change analysis

We compute the registration of the spinal canal of the last image to the fixation
image. This registration allows aligning the two images to the same reference system.
After that we compute the rigid registration of the lumen, and therefore, we can
Segmentation of the Spinal Canal
Segmentation of the lumen
Registration of the Spinal Canal
Migration of the
Stent Graft
Rigid Registration
of the lumen

Deformation of the
Stent Graft
Transformation of the
lumen
3D Images

Page 4

compute and visualize the migration of the stent graft, and at the end of the pipeline
we can compute and visualize its elastic deformation.
The work reported in this paper is integrated in a medical image processing
platform under development that consists in a module for accessing and managing
information. We intend to implement a data base with a user-friendly interface in the
hospital information system, through which health professionals have access to all
information obtained from the patient's clinical history, from the implantation of
EVAR and preoperative and post-operative evaluations (monitoring) of the aneurysm.
It allows the introduction, access, and modification of information.
2. Methods
Images obtained by Computerized Tomography (CT) are visualized as 3D volumes
with appropriate software tools. We have used ITK based applications (a popular
library of image analysis algorithms funded under the Visible Human Project by the
U.S. National Library of Medicine) [11] and VTK [12]. First, metaimages were
created from computerized tomography slices in DICOM format. This process keeps
the resolution and spacing of the original images. The files obtained in this way will
be used as inputs for subsequent 3D processing pipeline.
Then, a segmentation process of anatomical structures (spinal canal, lumen) is
carried out. We have used a User-Guided Level Set Segmentation (UGLSS) [13]
based on the well-known 3D active contour segmentation method called Region
Competition [14] to get both the segmented image of the spinal canal and the lumen.
The UGLSS algorithm proceeds as follows: first, the image is resampled into a
volume with isotropic spacing (1,1,1). Then, a Region of Interest (ROI) that contains
a segmented structure, the spinal canal in the first case, the stent graft (lumen) in the
second case, is selected. During the preprocessing, probability maps are computed, by
applying a smooth lower and upper threshold. This ensures that voxels inside the
spinal canal (and lumen) have a positive value and the outside negative.
We place a seed to initialize the evolving contour into the spinal canal (and lumen)
and we establish the parameters that control the propagation velocity and curvature
velocity. An evolving contour is a closed surface C(t,u,v) parameterized by variables
u, v and by the time variable t. The contour evolves according to the following partial
differential equation (PDE):



We compute the external force F by estimating the probability that a voxel belongs
to the structure of interest and the probability that it belongs to the background at each
voxel in the input image:

Page 5

Binary images are obtained after segmentation process. Once the structures of
interest are segmented, registration is performed. The registration is the process of
finding a spatial transform that maps points of an image with its corresponding point
in another image. In this case it will be intra-subject and mono-modal, as it examines
the same patient on different dates. A sequence of rigid, affine and deformable (B-
splines) registrations is performed. The first study is considered the fix image and the
others are registered respect to it. A linear interpolator, Mutual Information metric
and Regular Step Gradient Descent Optimizer are used.
Rigid, affine and deformable registration of the lumen allows for visual
comparison of the evolution of the stent-graft.
As the patient is not exactly placed in the same position in the different studies, the
rigid registration of the spinal canal is required to have the same reference system to
compare the images. In the registration process we get the transformation parameters
that we will use to translate the stent graft of the moving images to the same reference
system that we have in the fixed image.
First the two images corresponding to the patient spinal canal are roughly aligned
by using a transform initialization and then the two images are registered using a rigid
transformation. The parameters of the transformation are then applied to the lumen.
When these operations are performed, the migration of the stent graft from one study
to the next is visualized.
To compare the deformation of the stent graft, the two stent grafts are registered
using the rigid transformation obtained from the spinal canal.
The rigid transformation is used to initialize a registration with an affine transform
of the stent graft. The transform resulting from the affine registration is used as the
bulk transform of a B-spline deformable transform. The deformable registration is
computed, and finally the resulting transform is used to correct the stent graft moving
image.
3. Results
We have tested the approach with patients which have been treated with stent-graft
devices. The CT image stacks consists of slices with 512 x 512 pixels per slice and a
number of 70 to 100 slices, with a thickness of 3 mm and a x-y spacing of 0.684 mm
The time elapsed between different studies varies between 6 and 12 months.
The segmented images are visualized together with a CT slice to have a referenced
view (Fig. 3).
After we register the spinal canal, we visualize the migration of the stent graft from
one point in time to the next (Fig. 4).
We can compare the stent-graft of two different studies after a rigid registration
(Fig. 5). Deformation is visualized.

Page 6

Fig. 3. Segmented spinal canal and lumen with a 3D CT image
faded in.



Fig. 4. Stent grafts in different points in time after placing them in the
same reference system. Migration is visualized.

Page 7

Fig. 5. Stent grafts in different points in time after a rigid registration.
Deformation is visualized.

4. Conclusions
We have registered the spinal canal of different studies to place the patient in a single
reference system. After segmentation of the lumen in different studies, we have
applied them the transformation we get during the spinal canal registration. After this
process, migration of the stent graft is visualized. We have observed that vertical
movement of the stent occurs.
The registration process is carried out over binary images improving on the works
that perform registration over point sets, which always involve a greater loss of
information.
To visualize the deformation of the stent graft from one study to the next, a rigid
registration is performed. Kinking or buckling phenomenon happen.
As future work, registering images from different studies from a given patient can
provide us quantified values of the migration and deformation of the stent graft. The
correlation between the values of deformations and the actual evolution of the patient
will be stated with the medical team. This could lead to a model that would predict the
evolution of other patients and provide support for the decision. Support Vector
Machines (SVM) will be used to determine if the evolution of the EVAR is favorable
or unfavorable.
This will be part of a more complex database where multiple information about the
patients and monitored aneurysm will be available.

Page 8

Acknowledgments. This work is sponsored and partially financed by the Basque
Government through the INTEK Program. We also thank the companies Bilbomatica
and eMedica for their participation in this work.
References

1. J.Cronenwett, W.Krupski, R.Rutherford.: Abdominal aortic and iliac aneurysm.Saunders,
Philadelphia: Vascular Surgery; p. 1246-80 (2000)
2. Rodin, M. B., Daviglus, M. L., Wong, G. C., Liu, K., Garside, D. B., Greenland, P. And
Stamler, J.: Middle Age Cardiovascular Risk Factors Risk Factors and Abdominal Aortic
Aneurysms in Older Age. Hypertension, American Heart Association. (2003)
3. U.S. Preventive Services Task Force.: Screening for Abdominal Aortic Aneurysm:
Recommendation Statement. Annals of Internal Medicine. Vol 142. No. 3.198-202(2005)
4. J.Parodi, J.Palmaz, and H. Barone.: Transfemoral intraluminal graft implantation for
abdominal aortic aneurysms” Ann Vasc Surg. 5(6), pp. 491-499, (1991)
5. Ellozy, S.H., Carroccio, A., Lookestein, R.A., Jakobs T.S., Addis, M.D., Teodorescu, V.J.,
Marin, M.L.: Abdominal aortic aneurysm sac shrinkage after endovascular aneurysm repair:
correlation with chronic sac pressure measurement. J. Vasc. Surg., 43(1): 2-7(2006)
6. Sonesson, B., Dias, N., Malina, M., Olofsson, P., Griffin, D., Lindbland, B., Ivancev, K. :
Intra-aneurysm pressure measurements in successfully excluded abdominal aortic aneurysm
after endovascular repair. J. Vasc. Surg.; 37(4): 733-8. (2003)
7. Dominik S. Meier, Charles R.G. Guttmann.: Time-series analysis of MRI intensity patterns
in multiple sclerosis. NeuroImage 20 1193–1209 (2003)
8. Mattes, J. Steingruber, I. Netzer, M. Fritscher, K. Kopf, H. Jaschke, W. Schubert, R.: Spatio-
temporal changes and migration of stent grafts after endovascular aortic aneurysm repair.
International Congress Series 1281 393–397(2005)
9. Mattes, J. Steingruber, I. Netzer, M. Fritscher, K. Kopf, H. Jaschke, W. Schubert, R.:
Quantification of the migration and deformation of abdominal aortic aneurysm stent grafts
[6144-31]. Proc. of SPIE. VOL 6144; PART 1, pages 61440V (2006)
10.Ibanez, L., Schroeder, W.: The ITK Software Guide. Kitware, Inc. ISBN 1-930934-10-6,
http://www.itk.org/ItkSoftwareGuide.pdf (2003)
11. VTK: The visualization toolkit. http://www.vtk.org.
12. Paul A. Yushkevich, Joseph Piven, Heather Cody Hazlett, Rachel Gimpel Smith, Sean Ho,
James C. Gee, and Guido Gerig.: User-guided 3D active contour segmentation of anatomical
structures: Significantly improved efficiency and reliability. Neuroimage, Article In Press,
(2006)
13. Zhu, S.C., Yuille, A.: Region competition: unifying snakes, region growing, and Bayes/mdl
for multiband image segmentation. IEEE Trans. Pattern Anal. Mach. Intell. 18 (9), 884–900
(ISSN 0162-8828) (1996)