The simulation of magnetic resonance (MR) images plays an important role in the validation of image analysis algorithms such as image segmentation, due to lack of sufficient ground truth in real MR images. Previous work on MRI simulation has focused on explicitly modeling the MR image formation process. However, because of the overwhelming complexity of MR acquisition these simulations must involve simplifications and approximations that can result in visually unrealistic simulated images. In this work, we describe an example-based simulation framework, which uses an “atlas” consisting of an MR image and its anatomical models derived from the hard segmentation. The relationships between the MR image intensities and its anatomical models are learned using a patch-based regression that implicitly models the physics of the MR image formation. Given the anatomical models of a new brain, a new MR image can be simulated using the learned regression. This approach has been extended to also simulate intensity inhomogeneity artifacts based on the statistical model of training data. Results show that the example based MRI simulation method is capable of simulating different image contrasts and is robust to different choices of atlas. The simulated images resemble real MR images more than simulations produced by a physics-based model.
This paper aims at analyzing gender differences in the 3D shapes of lateral ventricles, which will provide reference for
the analysis of brain abnormalities related to neurological disorders. Previous studies mostly focused on volume analysis,
and the main challenge in shape analysis is the required step of establishing shape correspondence among individual
shapes. We developed a simple and efficient method based on anatomical landmarks. 14 females and 10 males with
matching ages participated in this study. 3D ventricle models were segmented from MR images by a semiautomatic
method. Six anatomically meaningful landmarks were identified by detecting the maximum curvature point in a small
neighborhood of a manually clicked point on the 3D model. Thin-plate spline was used to transform a randomly selected
template shape to each of the rest shape instances, and the point correspondence was established according to Euclidean
distance and surface normal. All shapes were spatially aligned by Generalized Procrustes Analysis. Hotelling T<sup>2</sup> twosample
metric was used to compare the ventricle shapes between males and females, and False Discovery Rate
estimation was used to correct for the multiple comparison. The results revealed significant differences in the anterior
horn of the right ventricle.
Statistical shape analysis of brain structures has gained increasing interest from neuroimaging community because it
can precisely locate shape differences between healthy and pathological structures. The most difficult and crucial
problem is establishing shape correspondence among individual 3D shapes. This paper proposes a new algorithm for
3D shape correspondence. A set of landmarks are sampled on a template shape, and initial correspondence is
established between the template and the target shape based on the similarity of locations and normal directions. The
landmarks on the target are then refined by iterative thin plate spline. The algorithm is simple and fast, and no
spherical mapping is needed. We apply our method to the statistical shape analysis of the corpus callosum (CC) in
phenylketonuria (PKU), and significant local shape differences between the patients and the controls are found in the
most anterior and posterior aspects of the corpus callosum.
A number of studies have documented that autism has a neurobiological basis, but the anatomical extent of these
neurobiological abnormalities is largely unknown. In this study, we aimed at analyzing highly localized shape
abnormalities of the corpus callosum in a homogeneous group of autism children. Thirty patients with essential autism
and twenty-four controls participated in this study. 2D contours of the corpus callosum were extracted from MR images
by a semiautomatic segmentation method, and the 3D model was constructed by stacking the contours. The resulting 3D
model had two openings at the ends, thus a new conformal parameterization for high genus surfaces was applied in our
shape analysis work, which mapped each surface onto a planar domain. Surface matching among different individual
meshes was achieved by re-triangulating each mesh according to a template surface. Statistical shape analysis was used
to compare the 3D shapes point by point between patients with autism and their controls. The results revealed significant
abnormalities in the anterior most and anterior body in essential autism group.