Hello Experts,
When viewing the Melodic statistical images in the usual output report, we see that thresh_zstat images overlaid onto the bg_image file. The bg_image file is the background image file chosen in the registration step within melodic. The bg_image is not exactly the same image as the origional, as it is only a fraction of the kilobyte size of the original. Also, to overlay the bg_image onto the origional anatomical image, requires a registration step to get it back to the same "size" as as the original.
As such, if we want to overlay the statistical images onto the original anatomical image, then the statistical images need also need to go through this second step of being registered to the original anatomical image. To do this, I have been running FLIRT after MELODIC so that I can view the stat images overlaid on the original full anatomic image. As I chose the linear registration step in MELODIC to begin with I assume that using FLIRT is okay, and don't need to use a non-linear registration step here. And this seem to work well, especially with the exterior cortical surfaces.
But regularly, I get a statistical image which shows bilateral activity in the posterior mesial temporal surfaces, which strangely also includes the superior surface of the cerebellum within the same two "blobs". I can't think of what network this would be, yet it regularly shows up, and the frequency characteristics are similar to other resting state networks. This makes me wonder that the registration steps I describes above may be not the best approach, especially for this area. Maybe there is a non-linear contraction of the bg_image fromt he original anatomical that effects the most medial interior areas, such that even small miscalculation can push a network into an incorrect area. Or, maybe this is just a network I am not familiar with.
Thanks in advance for your thoughtful help,
Varina
|