[FieldTrip] Source Point Count & Atlas Point Count Discrepancy

[Konstantinos Tsilimparis]

Hi Sarah,
I am happy that my suggestions helped you!
Regarding your new question, you need to consider two different parameters:

  1.  The inward shift
The inward shift parameter adjusts the volume conductor boundary and determines whether dipoles are inside or outside the head model. A negative inward shift will increase the "inside" points as you correctly pointed out. Using a negative inward shift let’s you align better the surface of your singleshell headmodel with the surface of the brain and can help your source reconstruction group analysis later on. The inward shift is relevant for the source reconstruction, where accurately identifying sources within and on the surface of the brain is crucial. However, it won't help improve the interpolation of the atlas with the source model (see: ft_sourceinterpolate).

  1.  The resolution of your sourcemodel
Increasing the resolution of your sourcemodel, improves the interpolation of the atlas with the source model. Now your interpolated atlas has higher resolution and has less “gaps” close to the brain’s surface.
I am not an expert, but my suggestion would be to use an inward shift and a high source model resolution (as high as possible without making it too computationally heavy). This approach will allow you to better identify sources on the brain's surface and determine the AAL ROI in which these sources belong.

Best,
Konstantinos Tsilimparis


From: fieldtrip <fieldtrip-bounces at science.ru.nl> On Behalf Of Sarah via fieldtrip
Sent: Monday, July 8, 2024 8:53 PM
To: fieldtrip at science.ru.nl
Cc: Sarah <s83728498 at gmail.com>
Subject: Re: [FieldTrip] Source Point Count & Atlas Point Count Discrepancy

Hi Konstantinos,

Thank you for your reply, the article link, and the nice visualization code suggestions!

I agree with you that one reason for the discrepancy is the lack of some brain regions being part of the atlas itself. I also agree that the 7.5mm grid is quite coarse so naturally not many source points can fit within an anatomical region. However, what I am still uncertain about is when you mention "the boundary between inside and outside is not well defined."

The discrepancy in grid points, that I notice, becomes apparent when you compare these two observations:
1) when I plot my sourcemodel's grid positions that are marked as "inside" and my headmodel, I notice a cloud of grid points that appear outside of the surface of headmodel (even while being considered as "inside"). In fact, I also use the cfg.inwardshift option when preparing my sourcemodel to specify an outward shift (because I use a singleshell model), and so I believe this also increases the number of sources marked as inside.
2) when I retrieve all the atlas points that correspond to my template grid and plot those grid points with my headmodel, I notice that the grid points only appear within the volume of the headmodel (ie the cloud of grid points at or above the surface of the headmodel is not included within the atlas points even while being considered as "inside")

My question is that given the atlas only seems to be including grid points that are within the volume of my headmodel, does this mean that specifying the outer-shift when preparing my sourcemodel to increase source points that are marked as "inside" is rendered as not useful? In other words, would there be no point to specify an inward/outward shift in this case even though I've learned that it may be helpful/recommended for single shell models in some cases?

Thanks again,
Sarah

On Wed, Jul 3, 2024 at 5:12 PM Sarah <s83728498 at gmail.com<mailto:s83728498 at gmail.com>> wrote:

Hi all,



I have a question about mapping source points to atlas ROIs.



I have computed my source analysis using FT’s 3D grid of 7.5mm. I have also interpolated the FT shipped AAL atlas (ROI_MNI_V4.nii) onto this 7.5mm template grid.



I have now been mapping only some relevant source points to the atlas to determine which ROI they exist in. However, I noticed that many of my source points didn’t “exist” within the atlas. In fact, the # of my subject’s source points that exist inside/outside are reported as: 6871 dipoles inside, 6791 dipoles outside brain, whereas I notice the atlas interpolated source model only has 3475 points total.



I was wondering how to handle this apparent discrepancy (as I assume the atlas point count should at least be somewhat closer to the total of source points marked as inside the brain).



Thanks,

Sarah



P.S. I included some of my relevant code below to produce this, but it is simply the code of the FT tutorials.



Subject Sourcemodel Code:

cfg           = [];

cfg.method    ='basedonmni';

cfg.template  = template_grid;

cfg.nonlinear = 'yes';

cfg.mri       = mri;

cfg.unit      ='mm';

grid          = ft_prepare_sourcemodel(cfg);



Interpolate Atlas Code:

cfg = [];

cfg.interpmethod = 'nearest';

cfg.parameter = 'tissue';

sourcemodel2 = ft_sourceinterpolate(cfg, atlas, template_grid);

sourcemodel2 = ft_convert_units(sourcemodel2, 'mm');
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20240709/ab6fb02b/attachment.htm>