614437 Nonrigid brain image registration using a statistical deformation model

1
6144-37 Non-rigid brain image registration using
a statistical deformation model

• Deformation field is modeled as a linear
  combination of the principal models of 64
  deformations.
• Viscous Fluid Model to create the deformations.
• 16 images from database and 3 new images.
• MI is higher using the new algorithm than using
  viscous fluid model.

2
6144-38 Nonrigid registration using
regularization that accomodates local tissue
rigidity

• Ignoring the elasticity differences between
  tissue types can result in non-physical results.
• A space variant regularization function which
  constrain the local jacobian of the deformation
  is developed.

3
6144-218 A whole brain morphometric analysis of
changes associated with pre-term birth

• The purpose is to identify structural
  differences in preterm infants.
• 88 preterm infants and 19 term born controls.
• PCA (Principal Component Analysis) and MLDA
  (Maximum uncertainty Linear Discriminant
  Analysis)

4
A good classification result between preterm and
term groups.
5
6144-86 Automatic sub-volume registration by
probabilistic random search

• The purpose is to map one part or one organ to
  the entire atlas data.
• Most of existing intensity based algorithm
  require the manual pre-registration.
• Probabilistic random search allows to find the
  optimal transformation without a close
  initialization.

6
6144-42 A comparison of FFD-based nonrigid
registration and AAMs applied to myocardial
perfusion MRI

• FFD vs. AAMs
• FFD Free-form Deformations
• AAMs Active Appearance Models
• FFD provides similar accuracy as the AAMs in
  terms of point to point errors.
• AAMs provide higher accuracy for point to curve
  error.

7
6144-36 Mjolnir deformable image registration
using feature diffusion

• Mjolnir modification of HAMMER.
• HAMMER Hierarchical Attribute Matching
  Mechanism for Elastic Registration.
• There are significant improvements in accuracy
  with reduction in computation time.

8
6144-33 Multimodal 2D-3D non-rigid registration

• X-ray projections 3D CT data
• Algebraic Reconstruction Technique (ART)
• Curvature registration

6144-44 Registration of 2D cardiac images to
real-time 3D ultrasound volumes for 3D stress
echocardiography
9
6144-04 Oriented active shape models

• Combine ASM with live wire.

6144-21 Improved 3D live-wire segmentation for 3D
CT chest image analysis

• Modify the standard 2D live-wire algorithm.
• Define a new 3D live-wire formulation.

10
6144-01 Image segmentation using local shape and
gray-level appearance models
6144-18 Automatic segmentation of vessels in
breast MR as a false positive elimination
technique for lesion detection and segmentation
using the shape tensor