Print

Print
Hi,

1) It wouldn't be easy to create a rectangular prism or cylinder with orientation that varied, depending on the orientation of the cortex.  Also, given the nature of how things would spread inside the head I'm not sure whether it is worthwhile going to this effort versus just sticking with a sphere.  So I would advise that you stick with the sphere for now.

2) If you are concerned about the accuracy of the GM mask, especially in the sulci, then I would advise you to take the GM PVE segmentation from the individual subject, in their native space and at that resolution, and threshold it yourself at a point where you are happy with the resulting accuracy.  It is unlikely to be perfect, but it is easier for you to adjust the relative errors between over-segmentation and under-segmentation if you choose your own threshold.  Using the native resolution should also help a lot.

3) The 3D viewer in FSLView is very sensitive to using the right thresholds and doing operations in the correct order.  If you go to the first practical in the FSL Course then you will find a section on running the 3D viewer.  Follow the instructions there very carefully, especially with respect to the order of the operations, and you should find that things will work out.

All the best,
Mark


On 16 Sep 2014, at 19:53, Tudor Popescu <[log in to unmask]> wrote:

Hi Mark,

Many thanks. I got the coordinates and used these commands to create a spherical mask of radius 20mm:
   fslmaths /usr/local/fsl/data/standard/MNI152_T1_2mm -mul 0 -add 1 -roi 127 1 168 1 110 1 0 1 mask_F3_point -odt float
   fslmaths mask_F3_point -kernel sphere 20 -fmean mask_F3_sphere -odt float

A couple of issues:

1) My electrodes were 40x40mm squares, and as a first attempt, I'd like to have a mask that has this surface area (40x40mm) at the level of the cortex, and that in terms of depth goes all the way till the WM-GM border. I guess that, better than to start with a spherical mask, would be to start with a cubical (or rather: parallelepipedal) one, but how can I still control its centre point and the size of its 3 edges (40x40xcortex_depth)?

2) As you indicated previously, I understand that after creating the mask, it then needs to be multiplied with a BET mask to exclude non-brain voxels. Since I want the mask to extend all the way down into the cortex till the WM-GM border, I figured that multiplying with the "GM_mask" produced by FSLVBm would be best, as it also takes care of the out-of-brain problem.
However, GM_mask seems to include much more than just GM voxels: for instance, the space between sulcal walls corresponds to white voxels in GM_mask. That might be due to the insufficient 2mm resolution that I know is standard for masks used with FSLVBM, but still - is this the proper mask to multiply by?

3) I tried to see how the mask looks like with regard to the MNI template brain in the 3D Viewer, however the 3D viewer (opened from the Tools menu) is always blank. Both the mask and the underlying MNI template were visible from the layers menu. Is the 3D viewer not able to render such a scenario?

Many thanks!
Tudor

On 5 September 2014 13:00, Mark Jenkinson <[log in to unmask]> wrote:
Hi,

I can't help you with finding the MNI coordinates, but if you have access to existing packages then I think you should persevere with them.

As for removing voxels that are not inside the brain mask, the easiest way to do this is use a binary brain mask (e.g. from running BET) and then multiply the two masks together with fslmaths.  This will zero all voxels outside of the brain mask.

All the best,
Mark


On 3 Sep 2014, at 23:47, Tudor Popescu <[log in to unmask]> wrote:

Dear FSL team,

I have structural and diffusion data from a training study involving different tDCS groups. For by VBM and TBSS analyses, I would like to define ROIs (masks) in MNI space corresponding to the approximate cortical area underlying the tDCS electrodes, which were placed on F3&F4 for one group and on P3&P4 in another group (10-20 system positions). I would therefore need to find the approximate centre coordinates, in MNI space, for the F3, F4, P3, P4 electrode positions; however, this is proving challenging.

Other papers that have done this (1, 2) have obtained coordinates for their electrodes (other positions than mine, unfortunately) using various types of converters, such as Brainstorm, Comet and the Muenster T2T Converter. However, it's proving very time-costly to get any of these to actually produce MNI coordinates at the cortical (not scalp) level for the electrode positions I mentioned. I suppose a scalp-to-cortex projection would have to be done, but I'm still working out how to do it. Clearly many approximations are involved so these won't be very accurate masks, but I want to at least try to define them first.

Also, assuming I find the centre coordinates and, based on them, I use fslmaths to define a cubical mask of cube edge equal to the electrode edge (40 mm), what is the easiest way of deleting from it voxels that are outside of the brain (assuming the centre of the cube will be right on the surface of the brain)?

Has anyone done this before and/or can share some advice? Many thanks!

Best wishes,
Tudor