JHUlogo
Small Animal Imaging Resource Program @ Johns Hopkins

JHU/JHMI SAIRP

Image realignment

The Siemens HRRT PET camera has excellent resolution ~2.4mm, which enables it to image the brain in exquisite detail. However, because of the superior resolution, subtle head movements, which are unavoidable during a typical 60-90 minute study, will be quite noticeable and detract from the quality of the image data. E.g., a subject may move their eyes, talk or otherwise open and close their mouth, or adjust their arms and legs. For brain scanning all of these motions can cause slight tilting or rotation of the head. In principle, the best way to correct for such motion is to measure the motion through a tracking device (e.g. the ‘Polaris’ system) and rebin the sinogram data. However, motion tracking has not been implemented thus we will do a ‘post-hoc’ motion correction instead where the reconstructed images will be realigned frame-by-frame. For that purpose we will make use of the spm_realign utility in SPM2.

Using Analyze
Step 1 : Create the reference image
The reference image is what all images will be normalized to. An appropriate reference image is a mean image averaged from multiple scans. Let’s try averaging slices 5-10 then apply motion correction to the remaining slices.


Step 2 : Realign all images to the reference image
Initially read in only a single late image frame. E.g., if there are 22 frames you can read in frame 19 or 20. This gives a good view of the overall uptake and image boundaries. Use the SubRegion tab in the Load As window to select only a single frame 'V'. At this point you can also do a conservative truncation of X, Y, and Z, making sure that nothing in the subject is truncated.


Step 3 : Create mean from realigned images
At this stage the realigned images are treated like an ordinary data set. Use ‘Load As’ to read in last 8 volumes of the motion corrected data. A mean image is created, which is then coregistered to MR, etc.