2011 Winter Project Week:Efficient co-registration of multiple MR modalities using the ABC
Illustration
Result of co-registration 1
Result of co-registration 2
Key Investigators
- Utah: Bo Wang, Marcel Prastawa, Guido Gerig
- UCLA: Jack Van Horn, Andrei Irimia, Micah Chambers
Objective
Up to now, the UCLA clinicians haven't done any studies on multiple MR modalities. We want to explore the use of ABC for coregistering different modalities for multiple modalities analysis, joint visualization of multiple co-registered modalities.
Approach, Plan
ABC (Atlas-Based Classification) is a fully automatic segmentation method developed in Utah. It can process arbitrary number of channels/modalities by co-registration, it integrates brain stripping, bias correction and segmentation into one optimization framework. A brain atlas is used as spatial priors for tissue categories. We want to use ABC to do efficient co-registration of multiple MR modalities, joint visualization of multiple co-registered modalities.
After the project week, we'll do the following work.
- Improve the GUI of ABC for co-rigistration.
- Incorporate the registration of the references of different time points into ABC
- add an option which asks the users to specify the paths of references at different time points to do the registration
Progress
We did the following works during the project week.
- Understand the requirements of neurosurgeons and clinicians
- Co-registration of multi-modal longitudinal data is important, this would allow the neurosurgeons and clinicians to compare the pathology (lesion, bleeding) across different modalities and time points.
- The ABC uses the first input image as reference and registers all other images to the first one, the co-registration of different modalities in the same time point is done automatically. For the co-registration of the data with different time points, we need to do the registration of the references of different time points firstly, and then run the ABC. In this way, co-registration is done across time and different modalities. We tried the following registration modules in Slicer.
- Robust Multiresolution Affine Registration
- Fast Rigid Registration
We found the results of rigid registration of different modules are good for doing the registration of the references of different time points (the results are slightly different).
We're aware that the Engineering core are working on improving the display module of Slicer. The new module will support the flexible visualization of multi-modal data.
