Contrast-enhanced liver MR image sequences acquired at multiple times before and after contrast administration
have been shown to be critically important for the diagnosis and monitoring of liver tumors and may be used
for the quantification of liver inflammation and fibrosis. However, over multiple acquisitions, the liver moves
and deforms due to patient and respiratory motion. In order to analyze contrast agent uptake one first needs
to correct for liver motion. In this paper we present a method for the motion correction of dynamic contrastenhanced
liver MR images. For this purpose we use a modified version of the Local Correlation Coefficient
(LCC) Demons non-rigid registration method. Since the liver is nearly incompressible its displacement field has
small divergence. For this reason we add a divergence term to the energy that is minimized in the LCC Demons
method. We applied the method to four sequences of contrast-enhanced liver MR images. Each sequence had a
pre-contrast scan and seven post-contrast scans. For each post-contrast scan we corrected for the liver motion
relative to the pre-contrast scan. Quantitative evaluation showed that the proposed method improved the liver
alignment relative to the non-corrected and translation-corrected scans and visual inspection showed no visible
misalignment of the motion corrected contrast-enhanced scans and pre-contrast scan.
Deep brain structures are frequently used as targets in neurosurgical procedures. However, the boundaries of these
structures are often not visible in clinically used MR and CT images. Techniques based on anatomical atlases and
indirect targeting are used to infer the location of these targets intraoperatively. Initial errors of such approaches may be
up to a few millimeters, which is not negligible. E.g. subthalamic nucleus is approximately 4x6 mm in the axial plane
and the diameter of globus pallidus internus is approximately 8 mm, both of which are used as targets in deep brain
stimulation surgery. To increase the initial localization accuracy of deep brain structures we have developed an atlas-based
segmentation method that can be used for the surgery planning. The atlas is a high resolution MR head scan of a
healthy volunteer with nine deep brain structures manually segmented. The quality of the atlas image allowed for the
segmentation of the deep brain structures, which is not possible from the clinical MR head scans of patients. The subject
image is non-rigidly registered to the atlas image using thin plate splines to represent the transformation and normalized
mutual information as a similarity measure. The obtained transformation is used to map the segmented structures from
the atlas to the subject image. We tested the approach on five subjects. The quality of the atlas-based segmentation was
evaluated by visual inspection of the third and lateral ventricles, putamena, and caudate nuclei, which are visible in the
subject MR images. The agreement of these structures for the five tested subjects was approximately 1 to 2 mm.
Proc. SPIE. 6918, Medical Imaging 2008: Visualization, Image-Guided Procedures, and Modeling
KEYWORDS: Visual process modeling, Optical spheres, Magnetic resonance imaging, Image segmentation, Heart, 3D modeling, Algorithm development, 3D image processing, Cardiovascular magnetic resonance imaging, Anisotropy
This paper presents a novel method for the generation of a four-chamber surface model from segmented cardiac
MRI. The method has been tested on 3D short-axis cardiac magnetic resonance images with strongly anisotropic
voxels in the long-axis direction. It provides a smooth triangulated surface mesh that closely follows the endocardium
and epicardium. The surface triangles are close-to-regular and their number can be preset. The input
to the method is the segmentation of each of the four cardiac chambers. The same algorithm is independently
used to generate the surface mesh of the epicardium and of the endocardia of the four cardiac chambers. For
each chamber, a sphere that includes the chamber is centered at its barycenter. A triangulated surface mesh
with almost perfectly regular triangles is constructed on the sphere. Then, the Laplace equation is solved over
the region bounded by the segmented chamber surface and the sphere. Finally, each vertex from the triangulated
mesh on the sphere is mapped from the sphere to the chamber surface by following the gradient flow of
the solution of the Laplace equation. The proposed method was compared to the marching cubes algorithm.
The proposed method provides a smooth mesh of the heart chambers despite the strong voxel anisotropy of the
3D images. This is not the case for the marching cubes algorithm, unless the mesh is significantly smoothed.
However, the smoothing of the mesh shrinks it, which makes it a less accurate representation of the chamber
surface. The second advantage is that the mesh triangles are more regular for the proposed method than for the
marching cubes algorithm. Finally, the proposed method allows for a finer control of the number of triangles
than the marching cubes algorithm.
Deep brain stimulation (DBS) surgery is a treatment for patients suffering from Parkinson's disease and other
movement disorders. The success of the procedure depends on the implantation accuracy of the DBS electrode
array. Surgical planning and navigation are done based on the pre-operative patient scans, assuming that brain
tissues do not move from the time of the pre-operative image acquisition to the time of the surgery. We performed
brain shift analysis on nine patients that underwent DBS surgery using a 3D non-rigid registration algorithm. The
registration algorithm automatically aligns the pre-operative and the post-operative 3D MRI scans and provides
the shift vectors over the entire brain. The images were first aligned rigidly and then non-rigidly registered with
an algorithm based on thin plate splines and maximization of the normalized mutual information. Brain shift of
up to 8 mm was recorded in the nine subjects, which is significant given that the size of the targets in the DBS
surgery is a few millimeters.
Most of the suggested image registration methods are based on the optimization of an objective function. Drawbacks of this approach are the problem of local minima and the need to initialize the transformation close to the true solution. This paper presents a method for N-dimensional rigid and similarity image registration that is not optimization-based and consequently it doesn't involve local minima and initialization. Instead of obtaining the transformation parameters implicitly through an iterative optimization process, they are obtained explicitly. The proposed method has advantages over existing explicit methods. The explicit expressions for transformation parameters involve image integrals and no image derivatives, which makes the method robust to noise. It is shown that the method has a few desired properties including symmetry and transitivity, and that it is invariant to initial alignment of the images. The method has been tested on simulated and real brain 2D and 3D MR image pairs and the achieved average registration error was one voxel.
Joint entropy, mutual information, and normalized mutual information are widely used image similarity measures in multimodality image registration and other problems that involve comparing images with arbitrary intensity relationships. While these image similarity measures have been successfully used in various applications, their mathematical properties have not been studied thoroughly. This paper analyzes several properties of practical interest of the three image similarity measures. It is shown that mutual information, despite its popularity, and joint entropy have a few undesirable properties. On the other hand, normalized mutual information does not suffer from these problems. The properties are proven mathematically, which renders the conclusions independent of image type, noise, and artifacts. The conclusions are in line with the results of previous experimental studies, in which normalized mutual information outperformed other information theoretic image similarity measures.
This paper presents a method for automated deformation recovery of the left and right ventricular wall from a time sequence of anatomical images of the heart. The deformation is recovered within the heart wall, i.e. it is not limited only to the epicardium and endocardium. Most of the suggested methods either ignore or approximately model incompressibility of the heart wall. This physical property of the cardiac muscle is mathematically guaranteed to be satisfied by the proposed method. A scheme for decomposition of a complex incompressible geometric transformation into simpler components and its application to cardiac deformation recovery is presented. A general case as well as an application specific solution is discussed. Furthermore, the manipulation of the constructed incompressible transformations, including the computation of the inverse transformation, is computationally inexpensive. The presented method is mathematically guaranteed to generate incompressible transformations which are experimentally shown to be a very good approximation of actual cardiac deformations. The transformation representation has a relatively small number of parameters which leads to a fast deformation recovery. The approach was tested on six sequences of two-dimensional short-axis cardiac MR images. The cardiac deformation was
recovered with an average error of 1.1 pixel. The method is directly
extendable to three dimensions and to the entire heart.
The goal of this work is to develop reliable constitutive models of the mechanical behavior of the in-vivo human brain tissue for applications in neurosurgery. We propose to define the mechanical properties of the brain tissue in-vivo, by taking the global MR or CT images of a brain response to ventriculostomy - the relief of the elevated intracranial pressure. 3D image analysis translates these images into displacement fields, which by using inverse analysis allow for the constitutive models of the brain tissue to be developed. We term this approach Image Guided Constitutive Modeling (IGCM). The presented paper demonstrates performance of the IGCM in the controlled environment: on the silicone brain phantoms closely simulating the in-vivo brain geometry, mechanical properties and boundary conditions. The phantom of the left hemisphere of human brain was cast using silicon gel. An inflatable rubber membrane was placed inside the phantom to model the lateral ventricle. The experiments were carried out in a specially designed setup in a CT scanner with submillimeter isotropic voxels. The non-communicative hydrocephalus and ventriculostomy were simulated by consequently inflating and deflating the internal rubber membrane. The obtained images were analyzed to derive displacement fields, meshed, and incorporated into ABAQUS. The subsequent Inverse Finite Element Analysis (based on Levenberg-Marquardt algorithm) allowed for optimization of the parameters of the Mooney-Rivlin non-linear elastic model for the phantom material. The calculated mechanical properties were consistent with those obtained from the element tests, providing justification for the future application of the IGCM to in-vivo brain tissue.
This paper presents an automated algorithm for extraction of Subdural Electrode Grid (SEG) from post-implant MRI scans for epilepsy surgery. Post-implant MRI scans are corrupted by the image artifacts caused by implanted electrodes. The artifacts appear as dark spherical voids and given that the cerebrospinal fluid is also dark in T1-weigthed MRI scans, it is a difficult and time-consuming task to manually locate SEG position relative to brain structures of interest. The proposed algorithm reliably and accurately extracts SEG from post-implant MRI scan, i.e. finds its shape and position relative to brain structures of interest. The algorithm was validated against manually determined electrode locations, and the average error was 1.6mm for the three tested subjects.
A method for constructing transitive nonrigid image registration
algorithms is presented. Although transitivity is a desirable property
of nonrigid image registration algorithms, the algorithms available in
the literature are not transitive. The proposed approach can be
applied to any nonrigid image registration algorithm and it
generalizes to any-dimensional case. The transitivity property is
achieved exactly up to the error of numerical implementation, which
can be arbitrary small. To the best of our knowledge, this is the
first time that transitivity of image registration algorithms has been
An existing 2D nonrigid image registration algorithm was made
transitive using the presented method. The algorithm was tested on two
sequences of cardiac short axis MR images. The maximal transitivity
error (defined in the paper) for several triples of images randomly
selected from the two sequences of cardiac images was on the order of
a millionth of a pixel.