Essential for a quality PET/MR coregistration is that the PET and MR data be in the correct absolute units, and that both image volumes have the same orientation. MR images from the Bruker 9.4T 210/72 scanner are stored as a single raw data with separate header files.

Get MR image information:
First determine the image dimensions: look in file d3proc to find XYZ voxel dimensions, e.g.
##$IM_SIX=256
##$IM_SIY=256
##$IM_SIZ=256

In file reco you will also find the image volume dimensions (cm), e.g.
##$RECO_fov=( 3 )
8.0000 4.0000 4.0000 (cm)
##$RECO_size=( 3 )
256 256 256 (pixels)

Note that the FOV is listed as Z (axial), then X, then Y.
Convert these values to determine pixel size (mm) so that the image dimensions will be in the same units as the PET study.
E.g., from the above numbers we compute
X=Y=40/256=0.15625mm
Z=80/256=.3125

To get the axial dimension (slice thickness) for a 2D acquisition look in file ACQP and find the line: ##$ACQ_slice_thick=0.800

Which indicates that the slice thickness is 0.8mm

Read in the MR images:
1: In the main Analyze window select File->Import/Export…
2: In the ImportExport window select Tools->Raw Data…
3: In the Raw Data – ImportExport window select the “Files” tab
In the first file window:
File            2dseq (remember to hit ‘Enter’!)
In the second window
File             desired_filename (ok to leave blank)

select the “Image Parameters” tab and use the following settings:
Byte Offset: 0
Width:         256
Height:        256
Depth:         256
Volumes:    1
Data Type:  signed 16-bit
Byte Swap: None

CHECK ORIENTATION!

Voxel Dim: Width=0.15625, Height=0.15625, Depth=.3125
Select 'Load'

Writing the MR volume:
In the main Analyze window the MR volume image will be named 'RawData' by default. Select the image then go to File->Save As. Select the directory and enter the desired filename.

Before coregistration, need to Create a mean PET image

Coregister PET/MR with Analyze
Make sure both the mean PET and the MR are loaded into Analyze. Select the Base image first (usually the MR), then with the Ctrl key pressed, select the Match image (PET). Both MR and PET images will be highlighted, then select 3-D Voxel Registration. If the images look like they are almost lined up then select 'Register'. If registration looks good then 'Save' transformed image.

If the images are not lined up well, then they should first be lined up manually to get the registration pretty close. At that point the automated registration can be used.

A normalized mutual information algorithm is used for coregistration. Always inspect the coregistration before accepting it. If it's acceptable then 'Save' transformed image.