Statistical Shape Models are a popular method for segmenting three-dimensional medical images. To
obtain the required landmark correspondences, various automatic approaches have been proposed. In
this work, we present an improved version of minimizing the description length (MDL) of the model. To
initialize the algorithm, we describe a method to distribute landmarks on the training shapes using a conformal
parameterization function. Then, we introduce a novel procedure to modify landmark positions
locally without disturbing established correspondences. We employ a gradient descent optimization to
minimize the MDL cost function, speeding up automatic model building by several orders of magnitude
when compared to the original MDL approach. The necessary gradient information is estimated from
a singular value decomposition, a more accurate technique to calculate the PCA than the commonly
used eigendecomposition of the covariance matrix. In this work, we first present a basic version where
spatial locations are used in the MDL cost function; next, we introduce an extended version where any
combination of features can be used as a metric. As an example application, we present results based on
local curvature measurements. Finally, we present results for synthetic and real-world datasets demonstrating
the efficiency of our procedures and give details about the implementation using the Insight
Toolkit (ITK).

Implementing the Automatic Generation of 3D Statistical Shape Models with ITK

This paper describes the Insight Toolkit (ITK) Conformal Flattening filter: itkConformalFlatteningFilter.
This ITK filter is an implementation of a paper by Sigurd Angenent, et al., “On the Laplace-Beltrami
Operator and Brain Surface Flattening”. This filter performs an angle preserving map of any genus
zero (i.e. no handles) triangulated mesh to the sphere or, alternatively, to the plane. In other words,
any given triangle in the resulting sphere will have the same angles as the corresponding triangle in the
original mesh, though their areas may differ. In this paper, we describe our code and provide the user
with enough details to reproduce the results which we present in this paper. This filter has a variety of
applications including the flattening of brain surfaces, which was the initial motivation for this work.

Conformal Flattening ITK Filter

The current conversion from an ellipse spatial object to an image
using SpatialObjectToImageFilter filter could cost minutes because
the filter traverses all pixels in the output image. This delay is
quite noticeable during applications such as human-machine
interface. This document describes a contribution to the Insight
Toolkit ITK www.itk.org, which implements what we call an
itk::EllipseBoundaryToImageFilter. This is a derived from
SpatialObjectToImageFilter. It marks the output image by setting
the pixels along the boundary of the input ellipse to the
InsideValue and the OutsideValue elsewhere. This filter is fast
because it traverses the boundary only. This paper is accompanied
with the source code.

itkEllipseBoundaryToImageFilter

This paper describes the use of open source software in the development and testing of an image-guided robot system for small animal research, presented at MICCAI 2006. This system relied on a significant amount of open source software, including 3D Slicer,
VTK, our own cisst software, the NetLib numerical methods, Python, and wxPython
(which uses wxWidgets). In addition, several open source development tools were used, including CVS, CMake, and Swig. The paper will be accompanied by the source code and raw data that were used to obtain the results presented at MICCAI.

Open Source Software in the Development and Testing of an Image-Guided Robot System

Medical image analysis is an important problem relating to
the study of various diseases. Since their inception to MICCAI in
2001, ''deformable organisms'' have emerged as a fruitful
methodology with examples ranging from 2D corpus callosum
segmentation to 3D vasculature and spinal cord segmentation.
Essentially we have developed an artificial life framework that
complements classical deformable models (snakes and deformable
meshes) with high-level, anatomically-driven control mechanisms.
This paper describes the integration of deformable organisms
into the Insight Toolkit (ITK) \url{www.itk.org}. Our code attempts
to bridge the ITK framework and coding style with deformable organism
design methodologies. In the interest of open science, as the
framework develops it will serve as a basis for the community to
develop new deformable organisms as well as experiment with those
recently published by our group.

I-DO: A Deformable Organisms framework for ITK

This article presents an approach for decreasing the computational time for converting a sparse SpatialObject to an Image. The method is applicable for SpatialObjects which occupy less volume than the total output Image (which occurs often, especially when using TubeSpatialObjects to represent vessels). A new filter MaskedSpatialObjectToImageFilter is introduced which dramatically decreases the execution time by creating a mask of relevant pixels. The mask is computed by calling the conventional SpatialObjectToImageFilter at a lower resolution, which is significantly faster for large images. The mask is then up-sampled to control which areas are further processed.

Using a Mask to Decrease Computation Time for SpatialObject to Image Conversions

VoxBo (www.voxbo.org) is a distributed processing framework, developed at the Center of Functional Neuroimaging at the University of Penssylvania, with the purpose of processing and analyzing brain functional MRI data. 
VoxBo is written in C++, and it is escentially a fair translation of some SPM functionality, originally written in Matlab. Along with the software, VoxBo defines two new image formats:
1) .cub, a 3D image format, used for anatomical data and single EPI frames. It is important to notice that this format defines a proper origin field, lacking, for instance, in Analyze.
2) .tes, a 4D image format, designed to contain the temporal EPI information. It is worth noticing that this format organizes the voxels in such a way to optimize longitudinal (temporal) voxel access.
Furthermore, the tes format is compressed through encoding of the functional voxels alone, i.e., background voxels are not stored for each frame, but only for the first one.

In this work we present the necessary IO classes that allow ImageFileReader and ImageFileWriter to properly identify, read and write .cub files only. We will provide .tes support in a subsequent work.
This effort is carried out with the purpose of letting VoxBo implementors to take advantage of ITK's power and overcome the limitations currently afflicting VoxBo.

Support for "cub" image format

This documents describes a set of classes to integrate digital topology in the
Insight Toolkit, with an application to homotopic thinning. The skeletonization
filter implemented using our set of classes is twice as fast as BinaryThinningImageFilter
and can be used on images of any dimension.

Digital Topology

We consider 2-D and 3-D digital binary images characterized by their well-composedness. Well-composed images exhibit important topological and geometrical properties not shared by their ill-composed counterparts. 
These properties have important implications for various algorithms used by the ITK community such as thinning algorithms and Marching Cubes. We introduce two image filters which repair images that are ill-composed such that the output images are well-composed.

Well-Composed Image Filters for Repairing 2-D and 3-D Binary Images

There is nowadays an increasing need in interaction between discrete surfaces (2-manifolds) and images. More specifically, segmentation and registration of n-dimensional images are taking advantage of a priori geometrical information, most often provided as discrete 2-manifolds.
Most of the publicly available libraries are oriented either toward mesh processing or image processing.
Through a careful study we will show that none of those libraries are complete enough to fullfill image, surface and joint image-surface interaction.
We propose to implement in ITK library a powerful 2-manifold data structure. The choice of ITK was driven by the fact that it provides the best base framework along with a strong n-dimensional image kernel. Based on a Quad-Edge data structure, it has been specifically tailored not only to represent
orientable 2-manifolds (surfaces of real objects) but also ease further processing.
We illustrate the integration of the design into ITK as a native object, enhancing existing algorithms.
We also illustrate the power of the new design for further surface processing.