Registration

Node Icon

Uses elastix to register images using specified parameters. It requires at least one fixed and one moving image as input, and will output the moving image registered and resampled to the fixed image coordinate system, as well as the registration transform and the progress.

The default registration method is rigid multimodal registration using mutual information. The parameters are saved with the Registration (Elastix) node in the workflow (.ice-file). Parameters can also be imported from a .txt file from within the Parameter editor.

For more information about elastix, visit their homepage or look at Image registration for a brief overview.

Inputs

Fixed

Fixed Image.

Type: Image, List, Required, Single

Moving

Moving Image.

Type: Image, List, Optional, Single

Outputs

Output

Moving image registered and resampled to the fixed image coordinate system.

Type: Image

T(x)

The final transformation found during the registration.

Type: Struct

Progress

Progress metrics, such as Time, Metric value etc.

Type: List of Lists

Settings

Inputs

Fixed Mask Boolean

Use a fixed mask to restrict the registration optimizer to a specific region in the fixed image.

Moving Mask Boolean

Use a moving mask to restrict the registration optimizer to a specific region in the moving image.

Landmarks Boolean

Use fixed and moving landmarks.

Initial Transform Boolean

Use an initial transform.

Auxiliary Image Boolean

Use an auxiliary image to guide the registration. It can describe e.g. the rigidity of the moving image. The use of the auxiliary image is defined in the Elastix Parameters by using the input name, e.g. ‘AUX0’.

Num Aux Images Integer

Set the number of auxiliary images to be used in the registration.

Outputs

Image Boolean

Define if the node will return an image output.

Progress Boolean

Define if the node will return registration progress information.

Configure Registration

2D/2D Reg Boolean

Use this option to register 2D images.

Multi-Metric Reg Boolean

Use this option to register a set of fixed and moving images simultaneously using different metrics simultaneously. A metric needs to be specified for each pair of fixed and moving images.

Number Of Metrics Integer

Set the number of metrics to be used in the registration. Each metric requires one pair of images.

Groupwise Reg Boolean

Use this option to register a time series simultaneously to all 3D volumes in the moving input. This incorporates all image information in registration process and eliminates bias towards a chosen reference frame.

Parameter Source

Source Selection

How to set the registration parameters. Settings lets you define the parameters directly in the node settings. Load Parameter File lets you import a parameter file from disk. Parameters Editor lets you open an editor where you can define the parameters using elastix syntax. From Input lets you connect a Structure containing elastix parameters.

Values: Settings, Load Parameter File, Parameters Editor, From Input

Paramter File Filepath

Specify the filepath to the paramater file.

Templates Selection

Choose a template for the registration parameters. The default parameters for each template can be changed in the settings.

Values: Affine (Affine Matrix), Affine (Rot+Shear+Scale+Trans), Non-Rigid, Rigid Time Series, Rigid

Editor Elastix Parameters

Open the editor. You can access it by double clicking the node as well.

Components

Transform Selection

Choose which transform to use.

Values: Translation, Rigid, Similarity, Affine, Affine (R+H+S+T), BSplines

Metric Selection

Choose which metric to use.

Values: Mean Squares Difference, Normalized Correlation Coefficient, Mutual Information, Normalized Mutual Information, Kappa Statistics

Transform Settings

BSplines Grid Spacing [mm] Numbers

Set the resolution of the Bsplines grid in [mm].

Metric Settings

Number of Bins Integers

Set the number of bins used by the metric. Be mindful of intensity outliers in the input image, e.g. a metallic implant in a CT image which has significantly higher intensity than the tissue. Such images can be clipped before registering to improve the registration quality.

Bending Penalty Boolean

Use bending energy penalty. See elastix documentation for further information.

Penalty Weight Numbers

Set the relative weights of the metrics.

Sampler Settings

Number of Samples Integers

Set the number of samples to calculate the metric for.

Number of Sampling Attempts Integers

A certain fraction of valid samples (i.e. samples that are within the moving image buffer) are required to give a valid result. If a valid results is not obtained, the sampler will try again for the number of times specified.

Fraction of Valid Samples Numbers

Specify the number of valid samples (i.e. samples that are within the moving image buffer) required for a valid result.

Pyramid Settings

Pyramid Steps Integer

Set the number of downsampling steps to use in the registration.

Fixed Pyramid Scaling Integers

Set the downsampling divisor for each pyramid step, specified as a Numeric Array of integers.

Moving Pyramid Scaling Integers

Set the downsampling divisor for each pyramid step, specified as a Numeric Array of integers.

Fixed Sampling Selection

Set the sampling strategy for the fixed image pyramid.

Values: Smoothing and Downsampling, Smoothing, Downsampling

Moving Sampling Selection

Set the sampling strategy for the moving image pyramid.

Values: Smoothing and Downsampling, Smoothing, Downsampling

General Settings

Initial Alignment Boolean

Use initial alignment as a starting point for the registration process.

Number of Iterations Integers

Set the number of iterations for each step in the image pyramid.

Extrapolation Voxel Value Float

Set the voxel value that should be applied to voxels outside of the moving image if those voxels are moved into the fixed frame of referece.

Resampling Order Selection

Set the order of the final resampling.

Values: Nearest Neighbour, Linear, BSpline Order 2, BSpline Order 3

See also

References

  1. Klein, M. Staring, K. Murphy, M.A. Viergever, J.P.W. Pluim, “elastix: a toolbox for intensity based medical image registration”, IEEE Transactions on Medical Imaging, vol. 29, no. 1, pp. 196 - 205, January 2010.
  2. D.P. Shamonin, E.E. Bron, B.P.F. Lelieveldt, M. Smits, S. Klein and M. Staring, “Fast Parallel Image Registration on CPU and GPU for Diagnostic Classification of Alzheimer’s Disease”, Frontiers in Neuroinformatics, vol. 7, no. 50, pp. 1-15, January 2014.

Keywords: rigid, affine, transform

Copyright © 2023, Hero Imaging AB