images/v_36.pngVINCI 4.24.0
Docu Home

Landmark Registration

Table of Contents

see also: Index, Plugins

Overview

The Landmark plugin was developed for registering two images by a least square fit of two sets of user defined points. Our approach for solving this Generalized Procrustes Problem uses Singular Value Decomposition to yield translations and rotations without reflections and scaling.

Usage

Open the plugin

you can find the landmark plugin in the plugins list below each OrthoDisplay (hammer icon). You can also open it by pressing <Q> after positioning the mouse cursor somewhere on the OrthoDisplay that displays the reference image. We recommend that the image you want to reslice (reslicing image) is shown in the OrthoDisplay below the one showing the reference image.

Define “landmarks” (points)

The points are defined by cutting the images at the desired position and using keyboard shortcuts or the landmark tool’s user interface. If you press CTRL-<1>, …, CTRL-<5> on the keyboard while the mouse cursor is positioned on an OrthoDisplay showing either the reslice or reference image, a landmark with an automatic name (e.g. <Point “3">) is generated. You can also define new landmarks by typing in a new name in the appropriate dropdown box and pressing the Set button. Points correspond through their names using a case sensitive comparison.

Managing landmarks

landmark positions can be changed by pressing the Set button with an already existing point name in the combobox or using the keyboard shortcut mentioned above, e.g. pressing key CTRL-<3> will overwrite <Point “3"> if that landmark was already defined. Their position can be shown as current cut position of the image by pressing the Find button. One or all already defined points can be deleted from the list with the Delete This and Delete All button, resp.

Register

initiates the registration process. Details depend on the number of independant points in the two point sets i.e. the rank of the point correlation matrix (see below). After registration the reslicing image is shown fused on the display of the reference image. The fused OrthoDisplay is put in the red link channel. Registration, fusion, link channel and mouse mode settings can be undone/redone in one step: either use the keyboard shortcuts <Ctrl-Z>, <Ctrl-Shift-Z> or use the protocol tool (Edit menu).

One point in each set

translate the reslicing image so that the common landmark are co-registered. Cut position and rotation center are set at the common point to start immediate manual registration. Mouse mode is changed to mouse induced image rotation. The link channel is in cut synchronization only mode.

Two points in each set

translate the reslicing image so that the centers of the points (“centers of gravity”) are co-registered, then rotate the reslicing image so that the connecting lines are parallel. Optionally (hidden by the “More…"-button), you can let Vinci prepare the images for manual registration. The program then rotates both images so as to make the connecting line vertical on the sagittal display. Cut position and rotation centers are set to the common “center of gravity” to start immediate manual registration. Mouse mode is changed to mouse induced image rotation. The link channel is in cut synchronization only mode.

Three or more points in each set

translate and rotate the reslicing image so that the image are fully co-registered - a warning message is issued if the mean square distance of the point sets exceeds twice the greater voxel-diagonal of the two images. Mouse mode is left unchanged. Cut Synchronization only is de-activated in the link channel.

Uncorrelated points

if, after the registration attempt, the mean square distance of the (rotated) point sets is larger than before only the translation is applied and an error message is issued. Cut position and rotation center are set at the common point to start immediate manual registration. Mouse mode is changed to mouse induced image rotation. The link channel is in cut synchronization only mode.

(C) 2005-2014 Max Planck Institute for Neurological Research Cologne, Germany