Nonlinear Group-wise Registration by Congealing
The goal of this work, being completed in conjunction with UMass Amherst, MGH, and MIT, is to use the congealing alignment model to simultaneously register group data for building probabilistic atlases. We have developed software to handle large-scale experiments and are currently studying how the unbiased atlases obtained from group registration outperform pairwise registration atlases on high-level tasks, such as segmentation. Also, we are developing techniques for congealing on data types other than intensities, such as tensors, to enable group registration focused on white matter.
 |
 |
 |
 |
 |
 |