Parameters

<< Click to Display Table of Contents >>

Navigation:  Registration Methods Reference > Diffeomorphic Elastic Matching (ANTS) >

Parameters

Most of the parameters correspond to the parameters of the original ANTS software, for which a manual page is available. Please refer to the ANTS user community for hints regarding tailoring of the parameters for specific tasks.

Basic Parameters

DiffeoBasicParameters

Initial Matching offers three methods to roughly align the data, before the elastic deformations start. The choices are

DiffeoInit

with Affine SPM from Template-based Normalization (SPM5) as default. Rigid will apply the standard Rigid Registration procedure, whereas Manual (Initial) will use the current location of the input image and not do any further alignment. The parameters of these methods (if applicable) appear after selecting them.

Nonlinear Warping should be enabled, otherwise the elastic part will be skipped. This can be helpful to assess the effect of the initial step.

The Cross correlation radius determines the number of pixels in the neighborhood, which are used for calculating the cross correlation cost function.

The Gradient Step characterizes the gradient descent optimization.

Advanced Parameters

DiffeoAdvancedParameters

With Outlier replacement enabled, extreme pixel values are replaces by the winsorization method. Histogram matching normalizes the pixel values, and is recommended when matching images from the same modality.

Number of levels determines the levels of hierarchical matching, working from coarse towards fine resolution. On each level, three parameters are configured: Shrink factor defines the sub-sampling in each direction. A factor of 8 reduces the number of pixels by a factor 83. Smoothing sigma is the Gaussian smoothing kernel size, and Number of iterations the number of optimizations at each level.

The Convergence parameters determine whether the iterations can be stopped before the configured number of iterations are exhausted. The Field variance parameters may be useful when adjusting to published methods.