Skip to content

TOPUP configuration files

This page contains listings of the configuration files that are provided with TOPUP. A TOPUP configuration file is simply a plain-text file containing command-line arguments, one argument per line. Lines beginning with a # character are ignored.

Three configuration files are provided with TOPUP; the resolution of your data set will dictate which file you should use.

b02b0_1.cnf

If your data has an odd number of voxels along any direction (e.g. 96x96x51), then you will need to use this configuration file.

# Resolution (knot-spacing) of warps in mm
--warpres=20,16,14,12,10,6,4,4,4
# Subsampling level (a value of 2 indicates that a 2x2x2 neighbourhood is collapsed to 1 voxel)
--subsamp=1,1,1,1,1,1,1,1,1
# FWHM of gaussian smoothing
--fwhm=8,6,4,3,3,2,1,0,0
# Maximum number of iterations
--miter=5,5,5,5,5,10,10,20,20
# Relative weight of regularisation
--lambda=0.0005,0.0001,0.00001,0.0000015,0.0000005,0.0000005,0.00000005,0.0000000005,0.00000000001
# If set to 1 lambda is multiplied by the current average squared difference
--ssqlambda=1
# Regularisation model
--regmod=bending_energy
# If set to 1 movements are estimated along with the field
--estmov=1,1,1,1,1,0,0,0,0
# 0=Levenberg-Marquardt, 1=Scaled Conjugate Gradient
--minmet=0,0,0,0,0,1,1,1,1
# Quadratic or cubic splines
--splineorder=3
# Precision for calculation and storage of Hessian
--numprec=double
# Linear or spline interpolation
--interp=spline
# If set to 1 the images are individually scaled to a common mean intensity 
--scale=1

b02b0_2.cnf

This configuration uses a factor 2 sub-sampling scheme to speed up the field map estimation. If your data has an even number of voxels in all directions (e.g. 96x96x50) you can use this file.

# Resolution (knot-spacing) of warps in mm
--warpres=20,16,14,12,10,6,4,4,4
# Subsampling level (a value of 2 indicates that a 2x2x2 neighbourhood is collapsed to 1 voxel)
--subsamp=2,2,2,2,2,1,1,1,1
# FWHM of gaussian smoothing
--fwhm=8,6,4,3,3,2,1,0,0
# Maximum number of iterations
--miter=5,5,5,5,5,10,10,20,20
# Relative weight of regularisation
--lambda=0.005,0.001,0.0001,0.000015,0.000005,0.0000005,0.00000005,0.0000000005,0.00000000001
# If set to 1 lambda is multiplied by the current average squared difference
--ssqlambda=1
# Regularisation model
--regmod=bending_energy
# If set to 1 movements are estimated along with the field
--estmov=1,1,1,1,1,0,0,0,0
# 0=Levenberg-Marquardt, 1=Scaled Conjugate Gradient
--minmet=0,0,0,0,0,1,1,1,1
# Quadratic or cubic splines
--splineorder=3
# Precision for calculation and storage of Hessian
--numprec=double
# Linear or spline interpolation
--interp=spline
# If set to 1 the images are individually scaled to a common mean intensity 
--scale=1

b02b0_4.cnf

This configuration uses a factor 4 sub-sampling scheme to speed up the field map estimation even more. Your image data needs to have dimensions that are a multiple of four in all directions (e.g. 96x96x48) in order to use this configuration.

# Resolution (knot-spacing) of warps in mm
--warpres=20,16,14,12,10,6,4,4,4
# Subsampling level (a value of 2 indicates that a 2x2x2 neighbourhood is collapsed to 1 voxel)
--subsamp=4,4,2,2,2,1,1,1,1
# FWHM of gaussian smoothing
--fwhm=8,6,4,3,3,2,1,0,0
# Maximum number of iterations
--miter=5,5,5,5,5,10,10,20,20
# Relative weight of regularisation
--lambda=0.035,0.006,0.0001,0.000015,0.000005,0.0000005,0.00000005,0.0000000005,0.00000000001
# If set to 1 lambda is multiplied by the current average squared difference
--ssqlambda=1
# Regularisation model
--regmod=bending_energy
# If set to 1 movements are estimated along with the field
--estmov=1,1,1,1,1,0,0,0,0
# 0=Levenberg-Marquardt, 1=Scaled Conjugate Gradient
--minmet=0,0,0,0,0,1,1,1,1
# Quadratic or cubic splines
--splineorder=3
# Precision for calculation and storage of Hessian
--numprec=double
# Linear or spline interpolation
--interp=spline
# If set to 1 the images are individually scaled to a common mean intensity 
--scale=1