Template Construction using a Patch-based Robust Estimator

Pierrick Coupé (1) , Vladimir Fonov (1) , José V. Manjón (2) , Louis D. Collins (1)

1 MNI - McConnell Brain Imaging Centre
2 ITACA

In MR image analysis, template construction is often required to perform voxel-based morphometry analysis in order to create a reference used to register a population of different subjects. Usually, after a non-linear registration step used to align the MR images together, simple voxel-wise averaging is used to construct the template. The averaging approach has the advantage of being fast and simple. However this technique is not robust to registration errors, and tends to degrade small structures and contrast between tissues producing an over-smoothed result. To overcome these limitations, we propose to use a robust patch-based estimation which is less sensitive to outliers during template construction and preserves image structure continuity.

Domaines: Imagerie médicale

Pierrick Coupé, Vladimir Fonov, José V. Manjón, Louis D. Collins. Template Construction using a Patch-based Robust Estimator. Organization for Human Brain Mapping 2010 Annual Meeting, Jun 2010, Spain. ⟨hal-00645471⟩