[FieldTrip] ft_sourceinterpolate question

Schoffelen, J.M. (Jan Mathijs) jan.schoffelen at donders.ru.nl
Thu Apr 26 09:35:39 CEST 2018


The most sensible approach to take here would be to interpolate the atlas onto the low-resolution grid that was used for the sourcereconstruction+connectivity analysis, and then parcellate.
Be aware that the functional data should be defined in the same coordinate system as the anatomical data. For the interpolation set the parameter to ‘brick0’, and the interpolation method to ‘nearest’.

Good luck,

Jan-Mathijs

On 26 Apr 2018, at 08:01, Bear, Joshua <Joshua.Bear at childrenscolorado.org<mailto:Joshua.Bear at childrenscolorado.org>> wrote:

Hi all,

It occurs to me that using ft_sourceinterpolate to decrease, rather than increase, the dimensions is not intuitive, and I found that it was actually fairly easy to get the Gorden et al. atlas loaded in as they provide an MNI-registered form of the atlas. So I repeated my steps using the selected atlas and with a smaller headmodel and leadfield resulting in a source file that had to be interpolated in the more anticipated direction. The problem persists, however.

I now have connectivity results with the following pertinent values:
          dim: [18 15 15]
          pos: [4050×3 double]
    cohspctrm: [4050×4050 double]
       dimord: 'pos_pos_freq'

My atlas has the following pertinent values:
            dim: [61 73 61]
      transform: [4×4 double]
         brick0: [61×73×61 uint16]
    brick0label: {1×333 cell}

The results of interpolating my connectivity source to the atlas lead to the same peculiar asymmetry in the cohspctrm results:
          dim: [61 73 61]
    transform: [4×4 double]
       inside: [61×73×61 logical]
    cohspctrm: [271633×4050 double]
          pos: [271633×3 double]

And as the whole point is to then use ft_sourceparcellate, I found that this carries over the asymmetrical results of the interpolation with the following source values after parcellation:
               freq: 9.9340
              label: {1×333 cell}
          cohspctrm: [333×4050 double]
      brainordinate: [1×1 struct]
                cfg: [1×1 struct]

As you can see, the first dimension of the cohspctrm results increased from 4050 to 271633 with interpolation and then decreased to 333, the number of parcels in my chosen atlas, exactly as I would have expected. The second dimension, however, did not budge from its initial 4050. The interpolation step is necessary to ensure the source matches the atlas (FT throws an error otherwise).

Could this be a bug in ft_sourceinterpolate and ft_sourceparcellate, or is there some crucial step such as converting everything into another format that I am missing?

Thanks again and sorry for the serial messages,
Josh



From: <fieldtrip-bounces at science.ru.nl<mailto:fieldtrip-bounces at science.ru.nl>> on behalf of Joshua Bear <Joshua.Bear at childrenscolorado.org<mailto:Joshua.Bear at childrenscolorado.org>>
Reply-To: FieldTrip discussion list <fieldtrip at science.ru.nl<mailto:fieldtrip at science.ru.nl>>
Date: Wednesday, April 25, 2018 at 8:14 PM
To: FieldTrip discussion list <fieldtrip at science.ru.nl<mailto:fieldtrip at science.ru.nl>>
Subject: Re: [FieldTrip] ft_sourceinterpolate question

Hi Robert,

Thank you so much for the response!

My ultimate intention is actually to use the parcellation scheme proposed by Gordon et al. (Cerebral Cortex, Volume 26, Issue 1, 1 January 2016, Pages 288–303, https://doi.org/10.1093/cercor/bhu239) though at this time I was using the atlas provided in the tutorial.

The problem I ran into is that I am trying to interpolate my brain’s mesh DOWN from 15002 to 8004 (to match the atlas), not up. As you pointed out, I have had some problems with other high resolution meshes, but in this case ft_sourceinterpolate seemed to work — it just only worked for one of the two dimensions of my connectivity data. That is, it went from a [15002×15002 double] to a [8004×15002 double]. Of course it’s possible the interpolation for that first dimension was incorrect, but what confused me was that it did not interpolate the second dimension. (As I write this, I’m wondering if I should simply try transposing the matrix and running the interpolation again!).

I will try, as you suggest, downsampling the mesh externally and trying again. I’m not able to match the 8004 vertices perfectly, but perhaps if I drop it slightly below this number it will be able to interpolate up correctly.

Any other thoughts or suggestions are greatly appreciated as well.

Best,
Josh



From: <fieldtrip-bounces at science.ru.nl<mailto:fieldtrip-bounces at science.ru.nl>> on behalf of "Seymour, Robert (Research Student)" <seymourr at aston.ac.uk<mailto:seymourr at aston.ac.uk>>
Reply-To: FieldTrip discussion list <fieldtrip at science.ru.nl<mailto:fieldtrip at science.ru.nl>>
Date: Wednesday, April 25, 2018 at 5:48 PM
To: "fieldtrip at science.ru.nl<mailto:fieldtrip at science.ru.nl>" <fieldtrip at science.ru.nl<mailto:fieldtrip at science.ru.nl>>
Subject: Re: [FieldTrip] ft_sourceinterpolate question

Hi Josh,

It seems like you are using the HCP Glasser atlas ('A multi-modal parcellation of human cerebral cortex')? The atlas presented as part of the network analysis tutorial is designed for a cortical mesh with 8004 vertices, rather than the 15002 you are using. I don’t believe ft_sourceinterpolate can currently cope with the surface-based transformation upwards from 8000—> 15000 vertices.

To use the HCP atlas you will need to downsample the original 32k atlas in fs_LR space to 15000 space and transform to whichever standard-brain template you are using. This is non-trivial and requires some functions from the HCP project (e.g. wb_command which can be found here: https://www.humanconnectome.org/software/workbench-command).

If your mesh is created using Freesurfer you can use this script here:

https://github.com/Washington-University/Pipelines/blob/master/PostFreeSurfer/scripts/FreeSurfer2CaretConvertAndRegisterNonlinear.sh

If the mesh is not constructed using Freesurfer it will be tricker & I would recommend contacting the HCP mailing list.

Good luck!

Robert Seymour (Aston Brain Centre & Department of Cognitive Science, Macquarie University).


________________________________
CONFIDENTIALITY NOTICE:  This e-mail, including any attachments, is for the sole use of the intended recipient and may contain confidential and privileged information.  If you are not an intended recipient, or the person responsible for delivering this message to an intended recipient, you are hereby notified that reading, copying, using or distributing this message is prohibited. If you are not an intended recipient, please contact the sender by reply email and destroy all copies of the original message from your computer system.
________________________________
________________________________
CONFIDENTIALITY NOTICE:  This e-mail, including any attachments, is for the sole use of the intended recipient and may contain confidential and privileged information.  If you are not an intended recipient, or the person responsible for delivering this message to an intended recipient, you are hereby notified that reading, copying, using or distributing this message is prohibited. If you are not an intended recipient, please contact the sender by reply email and destroy all copies of the original message from your computer system.
________________________________
_______________________________________________
fieldtrip mailing list
fieldtrip at donders.ru.nl<mailto:fieldtrip at donders.ru.nl>
https://mailman.science.ru.nl/mailman/listinfo/fieldtrip

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20180426/8d0ae8c4/attachment-0002.html>


More information about the fieldtrip mailing list