“School of Cognitive Sciences”
Back to Papers HomeBack to Papers of School of Cognitive Sciences
| Paper IPM / Cognitive Sciences / 8483 |
|
||||||||
| Abstract: | |||||||||
|
A limitation of using linear registration methods for mapping functional magnetic resonance imaging (fMRI) data to anatomical MRI, required for both multi-subject analysis and visualization of brain activations, is its inability to correct the geometric distortion induced by field inhomogeneity. Consequently, linear methods such as MI (mutual information) do not accurately align fMRI and MRI. Nonlinear methods such as TPS (Thin Plate Spline) may do the job but they require manual identification of landmarks. Due to low speed, high cost, and user-dependency of manual landmark identification, a new algorithm for automatic landmark generation is presented in this paper. The nonlinear registration method along with new automatic algorithm is applied on real anatomical and fMRI data and the results are evaluated. Experimental results illustrate the method and its excellent performance.
Download TeX format |
|||||||||
| back to top | |||||||||
