In modern medical image data, manifolds arise at varying scales. At one extreme, complete 3D data sets lie along manifolds parameterized by (for example) patient breathing and heartbeat patterns, or by confounding variables such as parameters or templates used in an image warping algorithm. At the other extreme, measurements taken at each voxel in multi-parametric MR images lie along locally defined manifolds that reflect nonlinear relationships among various MRI measurements on a voxel.

Discovering, visualizing and exploiting the structure of these manifolds supports the ability to select image-derived attributes that are informed by the structure of the underlying manifold. This offers fundamentally new tools for image registration, segmentation, visualization, reconstruction, and classification of data volumes. This workshop aims to bring together researchers in computer science, applied mathematics, statistics and medical imaging to present state of the art developments in this area.

Submissions are encouraged in (but not limited to) the following topics:

• Manifold analysis:
  • Learning natural manifolds (such as those occurring in 4D CT and cardiac MR)
  • Manifold tools for deformable shape registration and image segmentation
  • Diffusion tensor image analysis and visualization
• Theoretical Advances
  • Characterization of the manifold structure in natural data sets
  • Distance metrics appropriate for medical imaging
  • Novel and efficient manifold learning algorithms
• Normative data sets
  • Motion phantoms with known ground truth to support formal validation
  • Baseline/comparative results on realistic data