Home

Introduction

This small project creates a link between the statismo library and the elastix registration software.

Elastix is a powerful image registration software. It is based on the registration framework of the Insight Toolkit ITK. It extends ITK in many ways and allows for an easy interface based on parameter files. Statismo is a library for PCA-based statistical shape models.

With statismo-elastix, statistical shape model-based registration can be performed in elastix using shape models from statismo. For medical image analysis, these shape models typically represent the possible shape variations of an organ. By using statismo-elastix, the registration result is constrained by this model, so only valid organ shapes are allowed as registration results. This can make the registration much more robust.

The Statistical Shape Model

The shape model is not supplied or calculated by staismo-elastix. It needs to be created by the user. It needs to be a statismo model using a representer of the type itk::VectorImageRepresenter.

This means that the shape variation is represented not only on the surface of the shape but in a rectangular 2D or 3D image around it. An obvious way to build such a model is to use elastix to register a reference shape to a set of example shapes and use the resulting deformation fields as the input data sets for the VectorImageRepresenter model.

Initialization

For the registration to work, the model needs to be rigidly aligned to the target image. This means the organ in the image needs to be in the same coordinate system an position as the model. Typically, this is not the case to begin with, so the user needs to rigidly register the target to the model, e.g. by using elastix with a rigid transform or by resorting to some simple landmark-based method.