In medical imaging a frequent task has become
the registration of images from a subject taken with different imaging modalities,
where the term modalities here refers to imaging techniques such as Computed Tomography (CT), Magnetic Resonance Tomography (MRT) and Positron Emission Tomography (PET).
The challenge in inter-modality registration lies in the fact that e.g in CT images’bright’ regions are not necessarily bright regions in MRT images of the same subject.
an afﬁne registration, i.e. it determines an optimal transformation with respect to translation, rotation, anisotrope scaling, and shearing.
Closely related to registration is the task of image fusion, i.e. the simultaneous visualization of two
registered image datasets.
————Basic Manual Registration
play with software (i.e. amira) for a better alignment of the CT and MRT data, but it’s still not perfect…
an automatic registration is via optimization of a quality function.
For registration of datasets from different imaging modalities, in amira, the Normalized Mutual Information is the best suited quality function. In short, it favors an alignment which ’maps similar gray values to similar
gray values’. A hierarchical strategy is applied, starting at a coarse resampling of the datasets, and
proceeding to ﬁner resolutions later on.
————Registration Using Landmarks
You should be able to load ﬁles, interact with the 3D viewer, and be
familiar with the 2-viewer layout and the viewer toggles.
We will transform two 3D objects into each other by ﬁrst setting landmarks on their surfaces and then
deﬁning a mapping between the landmark sets. As a result we shall see a rigid transformation and a
warping which deforms one of the objects to match it with the other. The steps are:
1. Displaying data sets in two viewers.
2. Creating a landmark set.
3. Alignment via a rigid transformation.
4. Warping two image volumes.