[FieldTrip] How to apply ROI mask to grandavg (Sonja Suntrup-Kr?ger)

[RICHARDS, JOHN]

Sonja:

The x,y,z of the dim are the dimensions of a 3D matrix; the pos and pow are linear matrices, and are stacked as if they were an XYZ atlas.

How to do the other stuff is MATLAB programming.
If you have a 3D matrix, you can treat is like a linear matrix for multiplication.
  e.g., ROIMATRIX(x, y, z) can be accessed linearly with ROIMATRIX(:).
If you have a 4D matrix (nROI,x,y,z), it could be reshaped 
   reshapematrix=reshape(ROIMATRIX,nROI,size(ROIMATRIX,2)*size(ROIMatrix,3)*size(ROIMATRIX,4));
   This will give a matrix with (nROI,nVoxels), similar in dimension to the pos or pow matrix.

Or.  Keep your ROI atlas in the FT format in the first place (e.g. nROI X length(pos)) and then you won't have to transform it for the operations with the pow vector.

There are several other ways to do this in MATLAB.  One way is to use indices rather than 3D matrices.  You can translate each 3D ROI into an index matrix with sub2ind; then use the resulting indices to access the corresponding elements in the pow  (e.g., extractroi=pow(indices)), and simply sum or get an average of the elements in extractroi.

This is MATLAB program rather than FT programming.

John

***********************************************
John E. Richards
Carolina Distinguished Professor
Department of Psychology
University of South Carolina
Columbia, SC  29208
Dept Phone: 803 777 2079
Fax: 803 777 9558
Email: richards-john at sc.edu
HTTP: jerlab.psych.sc.edu
*************************************************


-----Original Message-----
From: Sonja Suntrup-Krüger [mailto:s.suntrup-krueger at uni-muenster.de] 
Sent: Tuesday, August 22, 2017 3:16 AM
To: RICHARDS, JOHN <RICHARDS at mailbox.sc.edu>; fieldtrip at science.ru.nl
Subject: Re: How to apply ROI mask to grandavg (Sonja Suntrup-Kr?ger)

Dear John,
hello everyone,
thank you very much for your valuable response!! Your suggestion is exactly what I had in mind. However, as I see it, if I want to do the multiplication you suggested (multiply the ROI matrix * the pow), the question that remains is how to reshape the ROI-matrix correcly. In other words:

How do the Voxel indices of the linear pos (or pow) -matrix relate to the voxel indices of the dim-cube matrix? I think this is what I need to know to correctly reshape my ROI-matrix from cube (N-ROI X 181 X 217 X 181)  to linear (N-ROI x 7109137) prior to the multiplication step.

Further advice is greatly appreciated!
Thank you very much!

Sincerely,
Sonja



RICHARDS, JOHN schrieb am 2017-08-19:
> Sonja. You are definitely on the right track.

>      pow: [7109137x6 double]  (because there are 6 datasets in the grandavg)
>      pos: [7109137x3 double]
>      dim: [181 217 181]

> The dim are the size of the cube for your source model matrix.  The pow/pos and other matrices are linear.   The linear arrays are a little easier to manipulate, may be the reason that FT decided to use these rather than have 3D matrices.  181*217*181 = 7109137.

> If you create a ROI matrix, it would either be of dimensions, N-ROI X 181 X 217 X 181, or more conveniently you could reshape it into a N-ROI x 7109137.

> Then multiply the ROI matrix * the pow, e.g., [N-ROI X 7109137] * 
> [7109137 X 6] = N-ROI * 6.

> This would sum across each voxel in the source model matrix for each roi.  You can imagine a single ROI of the entire source model (1 x N-dipoles), this would sum across all voxels in the source model.

> Or a single ROI, say of the IFG, in the ROI matrix, would give you the sum of the CDR pow data in the IFG.  Or 10 selected ROIs;  etc etc.

> I also often put the POW data into a MRI volume.  In the MATLAB format a NIFTI MRI volume struct representing this source model would have a img member that is 181 x 217 x 181.  You can simply assign the POW to the img model to get the CDR data into the MRI volume, e.g., MRIVOLUME.img = pow.  Then save or manipulate the MRI volume struct.

> John



> ***********************************************
> John E. Richards
> Carolina Distinguished Professor
> Department of Psychology
> University of South Carolina
> Columbia, SC  29208
> Dept Phone: 803 777 2079
> Fax: 803 777 9558
> Email: richards-john at sc.edu
> HTTP: jerlab.psych.sc.edu
> *************************************************


> -----Original Message-----
> From: fieldtrip-bounces at science.ru.nl 
> [mailto:fieldtrip-bounces at science.ru.nl] On Behalf Of 
> fieldtrip-request at science.ru.nl
> Sent: Saturday, August 19, 2017 6:00 AM
> To: fieldtrip at science.ru.nl
> Subject: fieldtrip Digest, Vol 81, Issue 16

> Send fieldtrip mailing list submissions to
> 	fieldtrip at science.ru.nl

> To subscribe or unsubscribe via the World Wide Web, visit
> 	https://mailman.science.ru.nl/mailman/listinfo/fieldtrip
> or, via email, send a message with subject or body 'help' to
> 	fieldtrip-request at science.ru.nl

> You can reach the person managing the list at
> 	fieldtrip-owner at science.ru.nl

> When replying, please edit your Subject line so it is more specific than "Re: Contents of fieldtrip digest..."


> Today's Topics:

>    1. How to apply ROI mask to grandavg (Sonja Suntrup-Kr?ger)
>    2. Problem with coordinate systems when creating head model
>       (Maria Hakonen)
>    3. bias in cluster stats from density of frequency	sampling?
>       (Nick Ketz)


> ----------------------------------------------------------------------

> Message: 1
> Date: Fri, 18 Aug 2017 12:45:31 +0200 (CEST)
> From: Sonja Suntrup-Kr?ger  <s.suntrup-krueger at uni-muenster.de>
> To: <fieldtrip at science.ru.nl>
> Subject: [FieldTrip] How to apply ROI mask to grandavg
> Message-ID:
> 	
> <permail-201708181045318218e1ae00002392-s_sunt01 at message-id.uni-muenst
> er.de>

> Content-Type: text/plain; charset=iso-8859-1

> Dear all,

> I want to compare task-related brain activation in a ROI  in two groups of subjects. I performed LCMV beamformer analysis on single subject data and, after normalization, created a grandaverage of each group, which I need to compare statistically within a defined ROI.
> Since the option to create a ROI in ft_sourcestatistis is unfortunately no longer available since 2015, I had the following idea to create my ROI:
> By using ft_volumelookup I created a binary mask from an anatomical atlas indicating my ROI in standard space. Now I would like to apply this mask on my grandaverage data. In specific, I want to multiply my mask values (zeros and ones) with the "pow" values in my grandavg.
> This would set all the power values outside my ROI to zero. With the result from this procedure I can then calculate my statistc.

> My binary mask has the dimensions: x,y,z
>      mask [181x217x181 logical]
>      coordsys: 'spm'

> In my grandavg, the dimension "dim" is the same, but the power values ("pow") and their positions ("pos") are listed as a single long  vector for each subject:
>      pow: [7109137x6 double]  (because there are 6 datasets in the grandavg)
>      pos: [7109137x3 double]
>      dim: [181 217 181]
>      coordsys: 'spm'

> My question is: How does the "x,y,z"-data structure of "dim" relate to the "single long vector"-data structure of "pow"/"pos"? If I had this information, I could transform my mask into the same structure and apply it as explained above.

> Your help is greatly appreciated.
> Thank you very much!
> Sonja Suntrup-Kr?ger

> --
> Sonja Suntrup-Krueger, MD
> Assistant Professor of Neurology
> Institute for Biomagnetism and Biosignalanalysis University of 
> Muenster, Germany


> *******************************