Program name: 3dwarper

Function: Nonlinear registration between 3D high-resolution T1-weighted MR images of the brain.

Author: Babak A. Ardekani, Ph.D.

Reference: Ardekani BA, Guckemus S, Bachman A, Hoptman MJ, Wojtaszek M, Nierenberg J.

 

Usage:

            3dwarper –obj  object  –trg  target  [–iter  N]  [–R]  [–u  filename]  [–o  filename]

            [–cubicspline]  [–c]  [–T  filename]  [–sd  N]  [–w  N]  [–s  N] [-A]

 

Required arguments:

 

obj   object

Specifies the image to be registered to the target image. object is expected to be an image of type ‘short’ in ANALYZE format.

 

obj   target 

Specifies the image to which the object image is to be registered. target is expected to be an image of type ‘short’ in ANALYZE format.

 

Optional arguments:

 

iter   N        

Specifies the number of multi-resolution levels.  N must be an integer between 1 and 4. Default N=4.

 

R      

Performs an initial 6-parameter linear rigid-body registration before starting the non-linear registration process.

 

u   filename         

Stores the displacement vector field in the specified filename.  Default filename=object.wrp

 

o   filename         

Stores the transformed (registered) object image in the specified filename.  Default filename=Cobject.img

 

cubicspline     

The output transformed (registered) object image is generated using the cubic spline interpolation (default is trilinear interpolation).

 

c       

Approximates the displacement vector field as a truncated Fourier-Legendre series.

 

T   filename         

Applies the linear transformation (4×4 matrix) specified in filename to the object image before starting the registration process.

 

sd   N

Specifies the degree of smoothing applied to the displacement vector field.   Typically ranges between 4.0 to 12.0 mm (default = search window size minus 1 mm).

 

w   N

Correlation window size in voxels (default=5; minimum=3).

 

s   N  

Search window size in voxels (default=5; minimum=3).

 

A

When this option is specified, the program first goes through three resolution levels. It then computes an initial affine transformation based on the results and restarts the optimization process.  Slightly more accurate registrations may be expected with this option.