1.  It is covered in these papers:

https://www.nature.com/articles/nature18933 — Supplemental Methods
https://www.nature.com/articles/nn.4361
https://www.sciencedirect.com/science/article/pii/S1053811917308649
https://www.sciencedirect.com/science/article/pii/S1053811914004546

2.  A standard mesh is somewhat analogous to a “standard volume space” in the 
sense that in theory there is one to one correspondence between vertices.  It 
is different in the sense that for a standard volume space in theory one has 
warped brains in 3D to all be the same—something that never actually happens 
because of the considerable variability across subjects in folding patterns and 
where areas are relative to folds.  For a standard mesh, this caveat does not 
matter because no attempt is made to move points in 3D space, one only assigns 
neuroanatomically corresponding locations (i.e. points within the same cortical 
area) the same vertex number across subjects.  The quality of this assignment 
depends on the registration algorithm with MSMAll better than MSMSulc better 
than FreeSurfer.  Once data of interest is on a standard mesh (e.g. via the 
MSMAll registration for HCP), it is trivial to compare across subjects or 
studies.

3.  This is not a recommended approach for the reasons stated earlier.

Matt.

From: Aslan Satary Dizaji <[email protected]<mailto:[email protected]>>
Date: Saturday, July 28, 2018 at 8:24 AM
To: Timothy Coalson <[email protected]<mailto:[email protected]>>
Cc: Matt Glasser <[email protected]<mailto:[email protected]>>, 
"[email protected]<mailto:[email protected]>" 
<[email protected]<mailto:[email protected]>>, Mohammad 
Reza Khodaei <[email protected]<mailto:[email protected]>>
Subject: Re: [HCP-Users] The coordinates of surface time-series of Resting 
State fMRI FIX-Denoised (Compact)

Thank you very much for your response. During this week, we had discussion on 
your comments.

>>Surface-format files, after registration and resampling to the same template 
>>(in this case, MSMAll) and standard mesh (in this case, 32k_fs_LR), can be 
>>compared/combined across subject by index, just like volume files.
1. Would you please direct us to a link with description about MSMAII template 
and how it has been created?
2. If we understand correctly, registering surfaces to a same standard template 
and standard mesh always results in corresponding indices. Is it correct?

>>However, it is possible to map such an ROI onto the surface, taking into 
>>account this blurriness, but it won't be binary anymore, and it will reveal 
>>what parts of cortex it is grabbing that you probably didn't want it to.  To 
>>do so, you can take a large number (at a guess, 30+, more is better but takes 
>>longer) of randomly-chosen HCP subjects, and use wb_command 
>>-volume-to-surface-mapping with -ribbon-constrained and each subject's 
>>MNINonLinear pial and white surfaces.  This will give you how your group ROI 
>>intersects each subject's cortical ribbon.  Then, you can average this across 
>>subjects to see on average, what parts of subjects' cortex are being captured 
>>by your ROI.  If you need it binary again, you will have to threshold it, and 
>>you might want to also use wb_command -metric-remove-islands (as your MNI 
>>space ROI will almost certainly cross to the opposite sulcal bank fairly 
>>strongly).
3. Is there any document which includes more information regarding this 
procedure?

Thank you for considering our request!


On Sat, Jul 21, 2018 at 1:10 AM Timothy Coalson 
<[email protected]<mailto:[email protected]>> wrote:
I think you are close, but there may be some terminology confusion.  Inline 
comments.

Tim

On Fri, Jul 20, 2018 at 12:34 AM, Aslan Satary Dizaji 
<[email protected]<mailto:[email protected]>> wrote:
Thank you very much for your response.

The reason that we use MATLAB based functions to read data is that we want to 
perform batch pipeline for the whole analysis of 100 unrelated subjects 
data-set.

ciftiopen.m, etc, are MATLAB based functions.  However, they don't load spatial 
information, we do all our spatial operations in scripts using wb_command 
(MATLAB is not ideal when you want to use a cluster to process a lot of 
subjects).  There are wb_command operations that can give you text and gifti 
files that can be loaded in matlab to give you this information, but from what 
you have said you wanted to do, I think you would be better off using the 
wb_command operation designed for mapping between surface and volume, rather 
than implementing similar code in matlab.  Improving matlab support for cifti 
is on the list of things to do, but with several things above it.

To sum up our conversations:

The coordinates of surface time-series of Resting State fMRI FIX-Denoised 
(Compact) data-set of 100 unrelated subjects lie in the folders 
"MNINonLinear/fsaverage_LR32k" of Structural Preprocessed data-set of 100 
unrelated subjects. It is better that we use MSMAll version of "*.dtseries.nii" 
files and midthickness_MSMAll version of "*.surf.gii" files.

Correct so far.

Also, the coordinates of surface time-series are not in the MNI space, so to 
apply a MNI mask on these coordinates, it is better that we put the MNI mask 
from the volume-space to the surface-space of a representative number of 
subjects by using HCP workbench.

Close, perhaps simply a wording issue - surface-based data doesn't inherently 
have a millimeter coordinate space it is tied to.  We provide surface files 
(which DO contain coordinates) in both MNI space and "T1w" space (really, 
distortion corrected rigidly aligned space), and you can choose between them 
depending on your task.  Because the timeseries data is already in surface 
format, moving it back to volume format would cause degradation (and worse 
alignment), it is generally better to move the ROI to the format the data is 
in.  Since the ROI is a volume-based MNI-space file, you need to use MNI-space 
surfaces to figure out where on the surface it should be, and using multiple 
subjects allows you to see the registration uncertainty in translating between 
MNI volume alignment and MSMAll surface alignment (and also avoids the issues 
with group-average surface coordinates).

One last question: if we don't use a mask and we do some analysis on the whole 
subcortical plus surface time-series, how can we summarize the results over
the whole group of subjects? I mean can we average the results over 
corresponding indices of time-series among all subjects?

Surface-format files, after registration and resampling to the same template 
(in this case, MSMAll) and standard mesh (in this case, 32k_fs_LR), can be 
compared/combined across subject by index, just like volume files.  CIFTI files 
continue this idea to both hemispheres and subcortical voxels simultaneously, 
as long as you use not only the same registrations/resamplings, but also the 
same set of brainordinates (in this case, the 91282 standard grayordinate 
space).

Thank you again for your time and consideration.

Best regards,

Aslan

On Fri, Jul 20, 2018 at 12:08 AM Timothy Coalson 
<[email protected]<mailto:[email protected]>> wrote:
ft_read_cifti.m does extra stuff that may be confusing, we don't do spatial 
operations in matlab.

I would recommend putting the volume-space mask into surface space, rather than 
trying to put the surface-based data back into volume space.  There is less 
information to be lost in a binary mask than in timeseries data.

As we have implied, defining a cortical ROI in MNI volume space is inherently 
defining it on a rather blurry map, as far as brain function is concerned, 
because MNI registration is (so far) based only on the shape of the brain, 
which is both quite variable across subjects over large expanses of cortex, and 
because function does not always line up with these brain shape features.

However, it is possible to map such an ROI onto the surface, taking into 
account this blurriness, but it won't be binary anymore, and it will reveal 
what parts of cortex it is grabbing that you probably didn't want it to.  To do 
so, you can take a large number (at a guess, 30+, more is better but takes 
longer) of randomly-chosen HCP subjects, and use wb_command 
-volume-to-surface-mapping with -ribbon-constrained and each subject's 
MNINonLinear pial and white surfaces.  This will give you how your group ROI 
intersects each subject's cortical ribbon.  Then, you can average this across 
subjects to see on average, what parts of subjects' cortex are being captured 
by your ROI.  If you need it binary again, you will have to threshold it, and 
you might want to also use wb_command -metric-remove-islands (as your MNI space 
ROI will almost certainly cross to the opposite sulcal bank fairly strongly).

Tim


On Thu, Jul 19, 2018 at 6:25 AM, Aslan Satary Dizaji 
<[email protected]<mailto:[email protected]>> wrote:
Thank you very much for both responses. Also, thank you for the insightful 
paper.

Here, I am going to describe a processing pipeline and I was wondering if you 
could tell me if there is something wrong with this pipeline:

1) We read "rfMRI_REST{1,2}_{LR,RL}_Atlas_MSMAll_hp2000_clean.dtseries.nii" 
with "ft_read_cifti.m" and 
"${subject}.{L,R}.midthickness_MSMAll.32k_fs_LR.surf.gii" with "gifti.m" of 
each subject.

2) We save four fields of the generated structure by "ft_read_cifti.m":
    2-1) "dtseries" which contains the time-series of subcortical voxels and 
cortical grayordinates.
    2-2) "pos" which, I assume, contains the MNI coordinates of subcortical 
voxels. [?]
    2-3) "brainstructure" which labels each structure of brain with different 
numbers.
    2-4) "transform" which, I assume, has the matrix for transformation of 
coordinates from MNI-mm space to MNI-matrix space (91,109,91). [?]

3) We save only the "vertices" field of the generated structure by "gifti.m" 
which, I assume, contains the coordinates of cortical grayordinates for left or 
right hemisphere in the MNI-mm space. [?]

4) Let's assume further that we want to apply a particular 3D mask in the 
MNI-matrix space (91,109,91).

5) By using "brainstructure" and "vertices", we can complete the "pos" with the 
correct coordinates of cortical grayordinates.

6) By using "transform", we can transform the coordinates of "pos" from the 
MNI-mm space to the MNI-matrix space (91,109,91).

7) Now we have a complete set of time-series from "dtseries" with their 
corresponding coordinates in the MNI-matrix space (91,109,91).

8) At this stage, we can apply our mask to this data.

Thank you in advance for your time and consideration.

Best regards,

Aslan

On Thu, Jul 19, 2018 at 1:15 AM Timothy Coalson 
<[email protected]<mailto:[email protected]>> wrote:
As an additional note, if you do use the non-MSMAll data, you should not use an 
MSMAll surface with it.  The data and surface file should go through the same 
registration and resampling.  The files without MSMAll in their names are 
registered/resampled using MSMSulc.

Tim


On Wed, Jul 18, 2018 at 8:16 AM, Glasser, Matthew 
<[email protected]<mailto:[email protected]>> wrote:
1.  rfMRI_REST1_LR_Atlas_MSMAll_hp2000_clean.dtseries.nii
2-3. 100307.L.midthickness_MSMAll.32k_fs_LR.surf.gii

Note that the surface coordinates or only valid for single individuals.  There 
is no “standard space” (like MNI space) that properly lines up the cortical 
areas, instead surface registration brings the data onto a standard 32k mesh 
where vertex 1000 has the same neuroanatomical location across subjects.  You 
probably don’t actually need the surface coordinates for your analysis unless 
you are doing something unusual.

Peace,

Matt.

From: Aslan Satary Dizaji <[email protected]<mailto:[email protected]>>
Date: Wednesday, July 18, 2018 at 12:57 AM
To: Matt Glasser <[email protected]<mailto:[email protected]>>
Cc: "[email protected]<mailto:[email protected]>" 
<[email protected]<mailto:[email protected]>>, Mohammad 
Reza Khodaei <[email protected]<mailto:[email protected]>>
Subject: Re: [HCP-Users] The coordinates of surface time-series of Resting 
State fMRI FIX-Denoised (Compact)

Dear Matt,

Thank you for your email.

In "Resting State fMRI FIX-Denoised (Compact)" of 100 unrelated subjects, 
inside folder "${StudyFolder}/${Subject}/MNINonLinear", there is only one 
folder, "Results", which inside it there are four folders "rfMRI_REST1_LR", 
"rfMRI_REST1_RL", "rfMRI_REST2_LR", "rfMRI_REST2_RL". And finally inside each 
one of these folders, there are two "*.dtseries.nii" files and two 
"*.dtscalar.nii" files. So basically, in "Resting State fMRI FIX-Denoised 
(Compact)" of 100 unrelated subjects, there is not any folder with the name of 
"fsaverage_LR32k". However, I checked the other data-sets of 100 unrelated 
subjects, and I found that "Structural Preprocessed" of 100 unrelated subjects 
has the  "fsaverage_LR32k" folder with the files that you mentioned. So my 
question is that, do we need to download this data-set too so to be able to get 
the coordinates of surface time-series?

Also, I have three other questions:

1) Which one of these "*.dtseries.nii" files do you recommend that we use:

rfMRI_REST1_LR_Atlas_hp2000_clean.dtseries.nii
rfMRI_REST1_LR_Atlas_MSMAll_hp2000_clean.dtseries.nii

2) For example, for subject 100307 and for left hemisphere, which one of these 
three "*.surf.gii" files do you recommend that we use:

100307.L.inflated.32k_fs_LR.surf.gii
100307.L.midthickness.32k_fs_LR.surf.gii
100307.L.very_inflated.32k_fs_LR.surf.gii

3) I assume that with "rfMRI_REST1_LR_Atlas_hp2000_clean.dtseries.nii", we 
should use one of the three above "*.surf.gii" files and with 
"rfMRI_REST1_LR_Atlas_MSMAll_hp2000_clean.dtseries.nii", we should use one of 
the three below "*.surf.gii" files:

100307.L.inflated_MSMAll.32k_fs_LR.surf.gii
100307.L.midthickness_MSMAll.32k_fs_LR.surf.gii
100307.L.very_inflated_MSMAll.32k_fs_LR.surf.gii

Am I right?

Many many thanks in advance for your time and consideration.

Best regards,

Aslan

________________________________
The materials in this message are private and may contain Protected Healthcare 
Information or other information of a sensitive nature. If you are not the 
intended recipient, be advised that any unauthorized use, disclosure, copying 
or the taking of any action in reliance on the contents of this information is 
strictly prohibited. If you have received this email in error, please 
immediately notify the sender via telephone or return mail.

_______________________________________________
HCP-Users mailing list
[email protected]<mailto:[email protected]>
http://lists.humanconnectome.org/mailman/listinfo/hcp-users




________________________________
The materials in this message are private and may contain Protected Healthcare 
Information or other information of a sensitive nature. If you are not the 
intended recipient, be advised that any unauthorized use, disclosure, copying 
or the taking of any action in reliance on the contents of this information is 
strictly prohibited. If you have received this email in error, please 
immediately notify the sender via telephone or return mail.

_______________________________________________
HCP-Users mailing list
[email protected]
http://lists.humanconnectome.org/mailman/listinfo/hcp-users

Reply via email to