Respect on the output image that I can’t see, as you said, I load them in the mrview and there is a black screen. But when I run mrstats they aren’t empty, so it could be a visualisation problem.
One possible misunderstanding is that you’re using the parcellation image to initialise the connectome tool, and expecting to see the parcellation image. That’s not quite how it works: that image is used to calculate certain properties of the parcellation, and the parcels are then displayed using whatever is selected for the “Node geometry” control in the connectome tool (which by default is the spheres that you see). It is possible to see the actual parcellation within the connectome tool by selecting “Overlay” for the node geometry, turning on “Crop to slab” in the connectome tool, ensuring that the slab thickness is set to zero, and preferably using something other than “Fixed” for node colouring.
But for you, the more pertinent test (if you haven’t done so already) is to simply open the parcellation image as the sole main image in mrview. That will ensure that the position of the camera is right in the middle of the image and the intensity windowing is appropriate; so if there is something in there, you should then see it.
The other thing to look for is the values reported on the colour bar in the bottom-right corner of the main window (top-right for images loaded using the overlay tool). If it shows a range of (0, -1), then mrview has been unable to set appropriate intensity windowing as the first loaded slice is empty - it will reset this windowing as soon as you encounter a non-empty slice (assuming your software is up-to-date). If it shows a sensible numerical range, but the main window still shows nothing but black, you might have an inadequate video card / driver, in which case you can report the contents shown when pressing “Info” (button at top-right corner of mrview - “OpenGL Information”.
I understand that this spatial offset is there because the parcellation node image isn’t generated from my data set, I mean, I use a generic parcellation file (aparc+aseg.nii) from the data set using on this tutorial. Is that correct? Or do I have to generate the parcellation image with my own data set?
If you’re displaying a b=0 image from one dataset and then initialising the connectome tool using a label image from a completely different subject / dataset, then yes, there will inevitably be a spatial offset between the displayed brain and the calculated positions of the nodes. The nodes will only appear in the correct positions relative to the b=0 image if you’ve performed parcellation of that subject’s anatomical image, and explicitly performed image registration between the DWI and T1 volumes (and preferably performed EPI distortion correction before that as well).
Coming back to the parcellation image I used, I’ve tried to aligne it with my DWI data using the command FLIRT of the FSL software but I haven’t been successful. How could I do this registration? Also, I tried to overlay in mrview the DWI and the parcellation node image, but, again, I only can see the DWI.
Trying to register a parcellation image to anything is almost certainly going to fail. Image registration relies on optimising some measure of image similarity in order to drive the alignment. With a label image, the ‘image contrast’ is a set of brain parcels, each labeled with a unique integer; you’re then asking the algorithm to align this information with an image where each type of brain tissue has an approximately constant intensity. So it’s pretty much impossible here to define a metric that obtains a maximal / minimal value when the two images are well-aligned. It is more common to perform image registration using the T1 image, with the derived transformation being applied to the parcellation image.
It’s also unusual to want to align a single-subject parcellation to an entirely different subject. Usually one would want to use the connectome tool to visualise generated connectome data; but this intrinsically requires a definition of nodes in order to calculate a connectome in the first place, which means a parcellation image must have been defined for that subject at some point, which is then logically the image you should be using to initialise the connectome tool. Freesurfer-based parcellations should not be registered between subjects and used for connectome construction.
Rob
