Skip to main content Skip to navigation

John's Gems

Please note that since 2010 I no longer update John's Gems, and have instead integrated them into my blog Neuroimaging Tips & Tools. Please use the John-Gems tag in the blog to find all the Gems.

Essential code snippets by John Ashburner

John Ashburner is an author of SPM and frequent contributor to the SPM email list. Many of his answers consist of short snippets of code that are incredibly useful for SPM analyses or general data manipulation. Below are the most some of the best (the 'gems') in an accessible form. (The inspiration for this page is "Graphics Gems", a series of books containing short bits of useful code for computer graphics.)

We don't yet know about any undiscovered bugs, so we aren't sure if they exist or not.
-JA, Mar 21, 2003.

I haven't tested them all but I'll make bug fixes as I find them. I hope you find this useful.

See also the SPM2 gems page and the SPM5 gems page.

Tom Nichols



1. Image Overlay
2. Zeroing NaNs
3. NaNing Zeros
4. Histograms
5. Mesh plots
6. Reading raw data
7. Reorienting
8. Reslicing
9. Rots & Trans
10. Surf rendering
11. Editing Analyze Headers
12. fMRI Brain Threshold
13. Blob windowing
14. Down sampling
15. Brain volume (VBM)
16. Scripting figures
17. Setting the origi
18. -log10 P-values from T images
19. VBM modulation
20. Custom templates & priors
21. ImCalc Script
22. Orthviews tricks
23. Tabulating T stats
24. Un-normalizing a point

Gem 1: Adding blobs to an image


Say you have a statistic image that you created outside of SPM and you want to overlay it on a structural image. The bare bones solution is as follows.

Create a thresholded version of the image, where all subthreshold voxels are set to NaN (See how to set NaNs below). Then...

Subject: Re: Your Message Sent on Fri, 19 Jan 2001 23:28:45 -0800
From: John Ashburner <john@FIL.ION.UCL.AC.UK>
Date: Mon, 22 Jan 2001 11:40:24 +0000
To: SPM@JISCMAIL.AC.UK
 
I think it may take quite a bit of programming to get most of
the routines working with your data.  One thing to get you
started would be to try an undocumented feature of spm_orthviews.m:
 
        spm_orthviews('image',spm_get(1,'*.img','Select background image'));
        spm_orthviews('addimage',1,spm_get(1,'*.img','select blobs image'));
 
Best regards,
-John
  
Here's a different approach, which I understand will show the blobs in a monotone color (instead of hot metal). Also, there are more bells and whistles (it's less of a hack).
 
Subject: Re: overlaying resulting image
From: John Ashburner <john@fil.ion.ucl.ac.uk>
Date: Wed, 4 Oct 2000 10:37:01 +0100 (BST)
To: spm@mailbase.ac.uk, duann@salk.edu
 
| Would you please show me how to overlay a statistical map onto the brain
| template SPM uses. Let's say I have an analyze-formated statistical
| map obtained from other software. It was already normalized to the same
| coordinates as the template images SPM has. How can I overlay this image
| onto the SPM template just like the results shown in SPM convention.
 
The following is supposed to work.  It uses a hidden undocumented
feature in SPM99, so it may contain bugs....
 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 
P1 = spm_get(1,'*.img','Specify background image');
P2 = spm_get(1,'*.img','Specify blobs image');
 
% Clear graphics window..
spm_clf
 
% Display background image..
h = spm_orthviews('Image', P1,[0.05 0.05 0.9 0.9]);
 
% Display blobs in red.  Use [0 1 0] for green, [0 0 1] for blue
% [0.6 0 0.8] for purple etc..
spm_orthviews('AddColouredImage', h, P2,[1 0 0]);
 
% Update the display..
spm_orthviews('Redraw');
 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 
Best regards,
-John
  

...to the top

Gem 2: Zeroing NaN values


In addition to setting NaN values to zero, this gem is also useful as a basic skeleton for reading each slice, doing something to it, and then writing it out.
 
Subject: Re: NaN values in beta values
From: John Ashburner <john@fil.ion.ucl.ac.uk>
Date: Mon, 23 Oct 2000 17:27:24 +0100 (BST)
To: spm@mailbase.ac.uk, steffejr@umdnj.edu
 
I can't think of any nice way of doing this with ImCalc, so I'm afraid
I'll have to show you a more labour intensive method that should be
modified slightly (as the filenames need changing) before pasting into
Matlab.
 
	VI       = spm_vol('original_beta.img');
	VO       = VI;
	VO.fname = 'patched_beta.img';
	VO       = spm_create_image(VO);
	for i=1:VI.dim(3),
		img      = spm_slice_vol(VI,spm_matrix([0 0 i]),VI.dim(1:2),0);
		tmp      = find(isnan(img));
		img(tmp) = 0;
		VO       = spm_write_plane(VO,img,i);
	end;
 
  

...to the top

Gem 3: NaNing zero values


 
Subject: Re: normalization of contrast images
From: John Ashburner  
Date: Wed, 2 Aug 2000 14:37:38 +0100 (BST)
To: spm@mailbase.ac.uk, gpagnon@emory.edu
  
[...]
 
| And, while I am at it, does anybody know how to convert a mask that
| has zeros outside the brain to a mask that has NaN outside the brain?
| I tried ImCalc with something like 'i1(find(i1==0))=NaN', but it
| doesn't like it.
 
By default, ImCalc outputs the data as 16 bit integer with scalefactors.
There is no NaN representation for this, so the data would need to be
written as floating point or double precision floating point.  I think
you can do this by typing something like:
        P     = spm_get(1,'*.img');
        Q     = 'output.img';
        f     = 'i1.*(i1./i1)';
        flags = {0,0,spm_type('float'),1};
        spm_imcalc_ui(P,Q,f,flags);
 
Best regards,
-John
  
Note you can NaN-out voxels below a threshold, say, 3:
f = 'i1 + 0./(i1>3)'

Also note that this is a general way for making spm_imcalc write out images with float or double precision; concisely

f = 'sqrt(i1)'; % whatever you like
spm_imcalc_ui('in.img','out.img',f,{[],[],spm_type('float')});

Using [],[] instead of 0,0 ensures that default values will be used for those two flags.

...to the top

Gem 4: Histogram of an image


 
Subject: Re: t-distribution display (histogram) with spm99b
From: john@fil.ion.ucl.ac.uk (John Ashburner)
Date: Fri, 30 Jul 1999 14:06:20 +0100
To: spm@mailbase.ac.uk, jovicich@humc.edu
 
 
The attached program should produce the histograms you are after.  To
call it, type:
 
	V      = spm_vol(spm_get(1,'*.img','Select image...'));
	[n, x] = histvol(V, 100);
	figure;
	bar(x,n);
  
[...]
 
Regards,
-John
The attached function is here histvol.m and below

 

 
function [n, x]=histvol(V, nbins)
% Create Histogram of an image volume
% FORMAT [n, x]=histvol(V, nbins)
% V     - mapped image volume (see spm_vol)
% nbins - number of bins to use.
% n     - number of counts in each bin
% x     - position of bin centres
%_______________________________________________________________________
% @(#)JohnsGems.html	1.42 05/02/02
 
if nargin==1, nbins = 256; end;
 
% determine range...
mx = -Inf;
mn =  Inf;
for p=1:V.dim(3),
	img = spm_slice_vol(V,spm_matrix([0 0 p]),V.dim(1:2),1);
	msk = find(isfinite(img));
	mx  = max([max(img(msk)) mx]);
	mn  = min([min(img(msk)) mn]);
end;
 
% compute histograms...
x = [mn:(mx-mn+1)/nbins:mx];
n = zeros(size(x));
for p=1:V.dim(3),
	img = spm_slice_vol(V,spm_matrix([0 0 p]),V.dim(1:2),1);
	msk = find(isfinite(img));
	n   = n+hist(img(msk),x);
end;
return;
 

...to the top

Gem 5: Mesh Plots of t-maps


 
Subject: Re: Mesh Plots of t-maps
From: John Ashburner <john@FIL.ION.UCL.AC.UK>
Date: Fri, 2 Feb 2001 11:14:20 +0000
To: SPM@JISCMAIL.AC.UK
 
| I am wishing to create Mesh plots in Matlab of the t-map produced
| from my SPM analysis. Being a somewhat Matlab virgin, I was wondering
| whether
| someone could possibly point me in the direction of exactly how to do this.
|
| Which mat file contains all t-values in the statistic image?
 
The t values are stored in spmT_????.img files, where ???? refers to
the contrast number.  You can read the values from these images into
Matlab something like:
 
        pl    = 30; % plane 30
        fname = spm_get(1,'*.img','Name of t image');
        V     = spm_vol(fname);
        M     = spm_matrix([0 0 pl]);
        img   = spm_slice_vol(V,M,V.dim(1:2),1);
 
Displaying the values can be dome something like:
        surf(img);
 
There are loads of commands for 3D visualisation in Matlab 5.x.  You can
find out what these are by typing:
        help graph3d
 
All the best,
-John
  

...to the top

Gem 6: Reading raw data


 
Subject: Re: raw data
From: John Ashburner <john@FIL.ION.UCL.AC.UK&gt
Date: Thu, 10 May 2001 14:33:19 +0100
To: SPM@JISCMAIL.AC.UK
 
> I have a very simple question.  Where does one find and how does one go
> about extracting the raw data associated with a particular voxel?
 
I would write a very short script to do this.  Try modifying and copy-pasting
the following....
* * * * * * * * * * * * * * * * * * * * * * * * * * * * --
V=spm_vol(spm_get(Inf,'*.img'));
x = 32;
y = 32;
z = 32;
 
dat = zeros(length(V),1);
for i=1:length(dat),
        dat(i) = spm_sample_vol(V(i),x,y,z,0);
end;
* * * * * * * * * * * * * * * * * * * * * * * * * * * * --
 
The vector dat will contain the raw data from the voxel at 32,32,32.
 
Best regards,
-John

...to the top

Gem 7: Reorienting images


 
Subject: Re: AC-PC positions
From: John Ashburner <john@fil.ion.ucl.ac.uk>
Date: Fri, 27 Oct 2000 15:00:05 +0100 (BST)
To: spm@mailbase.ac.uk, spm@fil.ion.ucl.ac.uk
[...]
 
The best that I can suggest is you try manually reorienting your
images via the <Display&gt button.  Try different rotations and
translations until the image is displayed how you want it.  The
attached Matlab function can then be used for reslicing the image(s)
in the transverse orientation, with 1mm isotropic resolution.
 
Best regards,
-John
Attached file: reorient.m
Error in c matrix fixed, June 6, 2001 -TEN

Subsequently, John offered modifications to use the native voxel size, which have been incorporated...

See also SPM5 Gem2 & Gem3

 
Subject: Re: reorient.m
From: John Ashburner <john@FIL.ION.UCL.AC.UK>
Date: Thu, 14 Dec 2000 17:04:38 +0000
To: SPM@JISCMAIL.AC.UK
 
| I would like to reorient some images in two ways,
| using the usefull function reorient() of John Ashburner.
|
|   - first is to keep the voxel size intact,
| but still reorienting in the transverse plane
 
Without testing the code, and without too much thought, I think
the modification to do this is involves something like changing from:
        mat = spm_matrix([mn-1]);
        dim = (mat\[mx 1]')';
to something like:
        vox = spm_imatrix(V.mat);
        vox = vox(7:9);
        mat = spm_matrix([0 0 0  0 0 0  vox])*spm_matrix([mn-1]);
        dim = (mat\[mx 1]')';
 
|
|   - second is to reorient in the coronal plane,
|     with 1x1x1 mm resol.
 
I think this is involves something like changing from:
        mat = spm_matrix([mn-1]);
to something like:
        mat = spm_matrix([0 0 0 pi/2])*spm_matrix([mn-1]);
or maybe:
        mat = spm_matrix([0 0 0 -pi/2])*spm_matrix([mn-1]);
 
|
| In the two cases, the final image should have no .mat file
| and be resliced using sinc ...
 
To reslice using sinc interpolation, you change from:
                img = spm_slice_vol(V,M,dim(1:2),1);
to something like:
                img = spm_slice_vol(V,M,dim(1:2),-6);
 
I hopehese suggestions work.
Best regards,
-John
  
If you want to set arbitrary voxel size, just set vox as desired in the fix above, but then be sure to force dim to be an integer.

For example, I wanted to increase the resolution of my images by a factor of three, so I did

 
        vox = spm_imatrix(V.mat);
        vox = vox(7:9)/3;
        mat = spm_matrix([0 0 0  0 0 0  vox])*spm_matrix([mn-1]);
        dim = ceil(mat\[mx 1]')');
                

...to the top

Gem 8: Reslicing images


This is for soley reslicing images, but it has a nice description of how to create a basic transformation matrix.
 
Subject: Re: how to write out a resampled volume with new user specified
              voxel size
From: John Ashburner <john@FIL.ION.UCL.AC.UK&gt
Date: Thu, 4 Jan 2001 10:57:52 +0000
To: SPM@JISCMAIL.AC.UK
 
 
|       I would like to batch resample some volume images to isotropic
| voxels.  Is there an SPM99 command line instruction (or menu option) that
| would permit me to write out a resampled volume with a user specified voxel
| size dimensions?  I know this core function is implicit in several
| modules.  I considered using Coregister or Spatial Normalization modules
| but these require a target or template volume of the desired resolution
| which doesn't always exist in my application.
|      Many thanks for any suggestions.
 
You could try the attached function that I've just scribbled together.  It
hasn't been tested, so I hope it works.
 
Best regards,
-John
The attached function is here reslice.m and below

 

 
function reslice(PI,PO,dim,mat,hld)
% FORMAT reslice(PI,PO,dim,mat,hld)
%   PI - input filename
%   PO - output filename
%   dim - 1x3 matrix of image dimensions
%   mat - 4x4 affine transformation matrix mapping
%         from vox to mm (for output image).
%         To define M from vox and origin, then
%             off = -vox.*origin;
%              M   = [vox(1) 0      0      off(1)
%                     0      vox(2) 0      off(2)
%                     0      0      vox(3) off(3)
%                     0      0      0      1];
%
%   hld - interpolation method.
%___________________________________________________________________________
% @(#)JohnsGems.html	1.42 John Ashburner 05/02/02
 
VI          = spm_vol(PI);
VO          = VI;
VO.fname    = deblank(PO);
VO.mat      = mat;
VO.dim(1:3) = dim;
 
VO = spm_create_image(VO); end;
for x3 = 1:VO.dim(3),
        M  = inv(spm_matrix([0 0 -x3 0 0 0 1 1 1])*inv(VO.mat)*VI.mat);
        v  = spm_slice_vol(VI,M,VO.dim(1:2),hld);
        VO = spm_write_plane(VO,v,x3);
end;
 

...to the top

Gem 9: Rotations and translations from .mat files


 
Subject: Re: rot+trans
From: John Ashburner <john@fil.ion.ucl.ac.uk>
Date: Fri, 6 Oct 2000 16:15:23 +0100 (BST)
To: spm@mailbase.ac.uk, pauna@creatis.insa-lyon.fr
 
| 1 how can I find the transformations
| (rot+translations) from the transformation matrix?
 
The relative translations and rotations between a pair of images
is encoded in the .mat files of the images.  So for images F.img and
G.img, you would find the transformation by:
 
	% The mat file contents are loaded by:
	MF = spm_get_space('F.img');
	MG = spm_get_space('G.img');
	
	% A rigid body mapping is derived by:
	MR = MF/MG; % or MR = MG/MF;
	
	% From this, a set of rotations and translations
	% can be obtained via:
	params = spm_imatrix(MR)
	
	% See spm_matrix for an explanation of what these
	% parameters mean
  

...to the top

Gem 10: Surface Renderings: Rolling your own (w/ blobs!)


 
Subject: Re: rendering SPM96
From: John Ashburner <john@FIL.ION.UCL.AC.UK&gt
Date: Tue, 20 Feb 2001 16:22:05 +0000
To: SPM@JISCMAIL.AC.UK
 
SPM96 shows voxels that are considered to be between 5mm in front of and 20mm
behind the surface that is displayed.  This is the same as for the "old style"
rendering of SPM99.
 
To change this depth, try changing line 131 of spm_render.m from:
        msk = find(xyz(3,:) &lt (z1+20) &amp xyz(3,:) &gt (z1-5));
to something like:
        msk = find(xyz(3,:) &lt (z1+5) &amp xyz(3,:) &gt (z1-5));
 
 
I would rather see the rendering done on the brain of either the same
subject that the data were acquired, or on a nice smooth average brain
surface.  Rendering can be done on a smooth average in SPM99, and there
is also the capability of doing the rendering on to the individual subjects
MR.  The rendering in SPM96 is done onto a single subject brain, which can
be quite misleading as it is not the same subject that the data comes
from.  Depths behind the surface are derived from the single subject brain,
which can cause problems as spatial normalisation is not exact.  An activation
that could be on the brain surface may appear a few mm in front of or behind
the single subject brain surface.
 
Another option within SPM99 is to use the brain extraction feature to
identify the approximate brain surface of a segmented structural image.
This would be saved as surf_*.mat.  A routine similar the one attached
can then be used to render the blobs on to this surface.
 
Best regards
-John
  
Attached file: fancy_rendering.m

...to the top

Gem 11: Editing Analyze Headers


OK, this mail is actually from Matthew Brett, but it references code written by John.
 
Subject: Re: Scale factor
From: Matthew Brett <matthewbrett@YAHOO.COM>
Date: Tue, 22 May 2001 18:47:15 -0700 (21:47 EDT)
To: SPM@JISCMAIL.AC.UK
 
Dear Masahiro,
 
> I would like to incorporate scaling factors to many header files.
> It would be nice if I can incorporate multiple scaling factors saved
> in a text file into multiple header files.  Is there a way to do
> this? 
 
For the same problem, I have used an utility written
by John Ashburner many eons ago, called dbedit:
 
http://www.mrc-cbu.cam.ac.uk/Imaging/dbedit.html 
 
You need to script it somehow, obviously, but the line in the file
setting your scale factor might look like:
 
    dbedit myfile.hdr dime.funused1=$myval
 
where your scale factor is in the variable $myval
 
I've also been playing with some perl functions to do this kind of
thing, written by Andrew Janke:
 
http://www.cmr.uq.edu.au/~rotor/software/ 
 
Best,
 
Matthew
  
dbedit tarball: dbedit.tar.gz

...to the top

Gem 12: fMRI Analysis Threshold


OK, this mail isn't from John either, and it doesn't even reference John's code, but it's useful info that I've needed on more than one occation.

The difficulty is that the fMRI interface doesn't querry about masking; this email spells out how to overcome this.

 
Subject: Re: explicit masking
From: Stefan Kiebel <skiebel@fil.ion.ucl.ac.uk&gt
Date: Tue, 27 Jun 2000 11:43:52 +0100
To: "Kevin J. Black" <kevin@npg.wustl.edu>, SPM <spm@mailbase.ac.uk&gt
 
Dear Kevin,
 
> Is it possible to instruct spm99 to search all voxels within a given
> mask image rather than all above a fixed or a %mean threshold? 
 
Yes, with SPM99 it's possible to use several masking options. 
 
To recap, there are 3 sorts of masks used in SPM99:
   1. an analysis threshold
   2. implicit masking
   3. explicit masking
 
1: One can set this threshold for each image to -Inf to switch off this
   threshold.
2: If the image allows this, NaN at a voxel position masks this voxel
   from the statistics, otherwise the mask value is zero (and the user
   can choose, whether implicit masking should be used at all).
3: Use mask image file(s), where NaN (when image format allows this) or
   a non-positive value masks a voxel.
 
On top of this, SPM automatically removes any voxels with constant
values over time.
 
So what you want is an analysis, where one only applies an explicit
mask. 
 
In SPM99 for PET, you can do this by going for the Full Monty and
choosing -Inf for the implicit mask and no 0-thresholding. Specify one
or more mask images. (You could also define a new model structure,
controlling the way SPM for PET asks questions).
 
With fMRI data/models, SPM99 is fully capable of doing explicit masking,
but the user interface for fMRI doesn't ask for it. One way to do this
type of masking anyway is to specify your model, choose 'estimate later'
and modify (in matlab) the resulting SPMcfg.mat file. (see spm_spm.m
lines 27 - 39 and 688 - 713).
 
   1. Load the SPMcfg.mat file, set the xM.TH values all to -Inf,
      set xM.I to 0 (in case that you have an image format not 
      allowing NaN). 
   2. Set xM.VM to a vector of structures, where each structure
      element is the output of spm_vol. For instance: 
               xM.VM = spm_vol('Maskimage');
   3. Finally, save by
               save SPMcfg xM -append
 
> If so, does the program define a voxel to be used as one which has
> nonzero value in the named mask image?
 
Not nonzero, but any positive value and unequal NaN. Note that you can
specify more than one mask image, where the resulting mask is then the
intersection of all mask images.
 
Stefan
  

...to the top

Gem 13: Setting Blob Color Bar Window


 
Subject: Re: colour bar
From: John Ashburner &ltjohn@FIL.ION.UCL.AC.UK&gt
Date: Thu, 28 Mar 2002 11:23:16 +0000 (06:23 EST)
To: SPM@JISCMAIL.AC.UK
 
> I would like to use the same colour bar for two different VBM experiments..
> Please, does anyone know how can I do this?
 
If this is for the display of orthogonal views, then you can tweek the
values that the blobs are scaled to by using a little bit of extra
Matlab code.  First of all, display an image, with superimposed blobs.
Then in Matlab, type:
 
        global st
        st.vols{1}.blobs{1}.mx
 
This will give the maximum intensity of the set of blobs.  Then do the
same with the other set of results.  Find the largest of both results
(suppose it is 5.6) and scale the blobs so they are displayed with
this maximum by:
 
        global st
        st.vols{1}.blobs{1}.mx = 5.6;
 
Best regards,
-John
 

...to the top

Gem 14: Down Sampling


 
Subject: Re: resample to lower resolution
From: John Ashburner <john@FIL.ION.UCL.AC.UK&gt
Date: Wed, 3 Jul 2002 12:35:22 +0100 (07:35 EDT)
To: SPM@JISCMAIL.AC.UK
 
> Is there a way to resample images of 512x512 spatial resolution down to
> 256x256 resolution?
 
If you copy and paste the following into Matlab, then it should do the job
for you.  Note that I haven't fully tested the code, but it worked on the
one image I tried it on.  Note that it only reduces the data using Fourier
transforms in two directions.  Combining slices in the 3rd direction is
just by averaging.
 
Best regards,
-John
%* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * 
 
% Select image to reduce
V           = spm_vol(spm_get(1,'*.img','Select image'));
 
% Create new image header information
VO          = V;
VO.fname    = 'reduced.img';
VO.dim(1:3) = floor(VO.dim(1:3)/2);
VO.mat      = V.mat*[2 0 0 -0.5 ; 0 2 0 -0.5 ; 0 0 2 -0.5 ; 0 0 0 1];
 
% Write the header
VO          = spm_create_image(VO);
 
% Work out which bits of the Fourier transform to retain
d1  = VO.dim(1);
d2  = VO.dim(2);
if rem(d1,2), r1a = 1:(d1+1)/2; r1b = [];     
         r1c = [];          r1d = (d1*2-(d1-3)/2):d1*2;
else,         r1a = 1:d1/2;     r1b = d1/2+1; 
         r1c = d1*2-d1/2+1; r1d = (d1*2-d1/2+2):d1*2;
end;
if rem(d2,2), r2a = 1:(d2+1)/2; r2b = [];     
         r2c = [];          r2d = (d2*2-(d2-3)/2):d2*2;
else,         r2a = 1:d2/2;     r2b = d2/2+1; 
         r2c = d2*2-d2/2+1; r2d = (d2*2-d2/2+2):d2*2;
end;
 
for i=1:VO.dim(3),
 
        % Fourier transform of one slice
        f = fft2(spm_slice_vol(V,spm_matrix([0 0 (i*2-1)]),VO.dim(1:2)*2,0));
 
        % Throw away the unwanted region
        f1  = [f(r1a,r2a) (f(r1a,r2b)+f(r1a,r2c))/2 f(r1a,r2d)
             ([f(r1b,r2a) (f(r1b,r2b)+f(r1b,r2c))/2 f(r1b,r2d)]+...
              [f(r1c,r2a) (f(r1c,r2b)+f(r1c,r2c))/2 f(r1c,r2d)])/2
               f(r1d,r2a) (f(r1d,r2b)+f(r1d,r2c))/2 f(r1d,r2d)]/4;
 
        % Fourier transform of second slice
        f = fft2(spm_slice_vol(V,spm_matrix([0 0 (i*2  )]),VO.dim(1:2)*2,0));
 
        % Throw away the unwanted region
        f2  = [f(r1a,r2a) (f(r1a,r2b)+f(r1a,r2c))/2 f(r1a,r2d)
             ([f(r1b,r2a) (f(r1b,r2b)+f(r1b,r2c))/2 f(r1b,r2d)]+...
              [f(r1c,r2a) (f(r1c,r2b)+f(r1c,r2c))/2 f(r1c,r2d)])/2
               f(r1d,r2a) (f(r1d,r2b)+f(r1d,r2c))/2 f(r1d,r2d)]/4;
 
        % Create a simple average of two FTs and do an inverse FT
        f   = real(ifft2((f1+f2)))/2;
 
        % Write the plane to disk
        VO  = spm_write_plane(VO,f,i);
end;
 

...to the top

Gem 15: Computing Cerebral Volume (VBM)


 
Subject:      Re: smoothed modulated image
From: John Ashburner <john@FIL.ION.UCL.AC.UK>
Date:         Fri, 12 Apr 2002 13:45:01 +0000
To: SPM@JISCMAIL.AC.UK
 
> considering a smoothed modulated image, which is the right
> interpretation of the matrix:  "each value of the matrix denotes the
> volume, measured in mm3, of gray matter within each voxel" or "each
> value of the matrix is proportional to the volume, measured in mm3,
> of gray matter within each voxel" or something else? 
 
The contents of a modulated image are a voxel compression map
multiplied by tissue belonging probabilities (which range between zero
and one).
 
The units in the images are a bit tricky to explain easily (so I would
suggest you say that intensities are proportional).  To find the
volume of tissue in a structure in one of the modulated images, you
sum the voxels for that structure and multiply by the product of the
voxel sizes of the modulated image.
 
The total volume of grey matter in the original image can be
determined by summing the voxels in the modulated, spatially
normalised image and multiplying by the voxel volume (product of voxel
size).
 
For example, try the following code for an original image and the same
image after spatial normalisation and modulation. Providing the
bounding box of the normalised image is big enough, then both should
give approximately the same answer.
 
V = spm_vol(spm_get(1,'*.img'))
tot = 0;
for i=1:V(1).dim(3),
        img = spm_slice_vol(V(1),spm_matrix([0 0 i]),V(1).dim(1:2),0);
        tot = tot + sum(img(:));
end;
voxvol = det(V(1).mat)/100^3; % volume of a voxel, in litres
tot    = tot; % integral of voxel intensities
 
tot*voxvol
 
 
I hope the above makes sense.
 
All the best,
-John

...to the top

Gem 16: Scripting Figures


OK, this isn't a John email, but rather a tip of my own that uses one of John's functions. When preparing a manuscript you often want to display a "blobs on brain" image, where a reference image underlies a colored significance image. You can do this within the SPM Results facility, but since you never get a figure right the first time, I prefer to do it on the command line.

The code snippet blow scriptizes the blobs-on-brain figure. You'll get a large orthgonal viewer in the graphics window, so it's then easy to print (or grab a screen snapshot) to then create your figure.

 

 
% Make sure to first clear the graphics window
 
% Select images
Pbck = spm_get(1,'*.img','Select background image')
Psta = spm_get(1,'*.img','Select statistic image')
 
% Set the threshold value
Th   = 4;  
 
% Create a new image where all voxels below Th have value NaN
PstaTh = [spm_str_manip(Psta,'s') 'Th'];
spm_imcalc_ui(Psta,PstaTh,'i1+(0./(i1>=Th))',{[],[],spm_type('float')},Th);
 
% Display!
spm_orthviews('image',Pbck,[0.05 0.05 0.9 0.9]);
spm_orthviews('addimage',1,PstaTh)
 
% Possibly, set the crosshairs to your favorite location
spm_orthviews('reposition',[0 -10 10])
 
                
This assumes that you just want to threshold your image based on a single intensity threshold. To make it totally scripted, replace the spm_get calls with hard assignments.

 

...to the top

Gem 17: Origin maddness


A source of confusion is where the origin (the [0,0,0] location of an image) is stored. When there is no associated .mat file, the origin is read from the Analyze originator field. If this is zero it is assumed to match the center of the image field of view. If there is a .mat file, then the origin is the first three values of
 
        M\[0 0 0 1]'
                
where M is the transformation matrix in the .mat file.

One limitation is that the origin stored in the Analyze header is a (short) integer, and so cannot represent an origin with fractional values. To set the origin to specific, fractional value, use this code snippet:

 
  Orig = [ x y z ]; % Desired origin in units of voxels
  P = spm_get(Inf,'*.img'); % matrix of file names
 
  for i=1:size(P,1)
 
    M = spm_get_space(deblank(P(i,:)));
    R = M(1:3,1:3);
    % Set origin
    M(1:3,4) = -R*Orig(:);
    spm_get_space(deblank(P(i,:)),M);
 	
  end

...to the top

Gem 18: -log10 P-values from T images


P-value images are difficult to visualize since "important" values are small and clumped near zero. A -log10 transformation makes for much better visualization while still having interpretability (e.g. a value of 3 cooresponds to P=0.001).

This function, T2nltP, will create -log10 P-value image based on either a contrast number (which must be a T contrast) or a T statistic image and the degrees of freedom.

(See also the equivalent SPM2 function.)

T2nltP.m

 

 
function T2nltP(a1,a2)
% Write image of -log10 P-values for a T image
%
% FORMAT T2nltP(c)
% c     Contrast number of a T constrast (assumes cwd is a SPM results dir)
%
% FORMAT T2nltP(Timg,df)
% Timg  Filename of T image
% df    Degrees of freedom
%
%
% As per SPM convention, T images are zero masked, and so zeros will have
% P-value NaN.
%
% @(#)T2nltP.m	1.2 T. Nichols 03/07/15
 
if nargin==1
  c = a1;
  load xCon
  load SPM xX
  if xCon(c).STAT ~= 'T', error('Not a T contrast'); end
  Tnm = sprintf('spmT_%04d',c);
  df = xX.erdf;
else
  Tnm = a1;
  df  = a2;  
end
 
 
Tvol = spm_vol(Tnm);
 
Pvol        = Tvol;
Pvol.dim(4) = spm_type('float');
Pvol.fname  = strrep(Tvol.fname,'spmT','spm_nltP');
if strcmp(Pvol.fname,Tvol.fname)
  Pvol.fname = fullfile(spm_str_manip(Tvol.fname,'H'), ...
			['nltP' spm_str_manip(Tvol.fname,'t')]);
end
 
 
Pvol = spm_create_image(Pvol);
 
for i=1:Pvol.dim(3),
  img         = spm_slice_vol(Tvol,spm_matrix([0 0 i]),Tvol.dim(1:2),0);
  img(img==0) = NaN;
  tmp         = find(isfinite(img));
  if ~isempty(tmp)
    img(tmp)  = -log10(max(eps,1-spm_Tcdf(img(tmp),df)));
  end
  Pvol        = spm_write_plane(Pvol,img,i);
end;
 

...to the top

Gem 19: VBM modulation script


This is the famed script to modulate spatially normalized probability images. For gray matter probability images, modulated images have units of gray matter volume per voxel, instead of gray matter concentration adjusted for differences in local brain size.

In John's own words, here's a better explaination.

Note that script below is SPM99 specific. In SPM2, it is done via spm_write_sn(V,prm,'modulate') (See spm_write_sn help fore more.)

 
Date: Thu, 27 Jul 2000 14:28:39 +0100 (BST)
Subject: Re: matlab question & VBM
From: John Ashburner <john@fil.ion.ucl.ac.uk>
To: spm@mailbase.ac.uk, x.chitnis@iop.kcl.ac.uk
 
 
| Following on from John Ashburner's recent reply, is there a matlab function
| that enables you to adjust spatially normalised images in order to preserve
| original tissue volume for VBM?
 
The function attached to this email will do this.  Type the following bit of
code into Matlab to run it:
 
	Mats   = spm_get(Inf,'*_sn3d.mat','Select sn3d.mat files');
	Images = spm_get(size(Mats,1),'*.img','Select images to modulate');
 
	for i=1:size(Mats,1),
	        spm_preserve_quantity(deblank(Mats(i,:)),deblank(Images(i,:)));
	end;
 
[...]
 
Best regards,
-John
The attached script is here spm_preserve_quantity.m

 

...to the top

Gem 20: Creating customized templates and priors


 
From: John Ashburner <john@FIL.ION.UCL.AC.UK>
Subject: Re: Help for constructing Template images
Date: Wed, 18 Dec 2002 16:22:59 +0000
To: SPM@JISCMAIL.AC.UK
 
> What are the advantages of customized template images
> in VBM analysis?
 
Customised templates are useful when:
 
1) The contrast in your MR images is not the same as the
   contrast used to generate the existing templates.  If
   the contrast is different, then the mean squared cost
   function is not optimal.  However, for "optimised" VBM
   this only really applies to the initial affine
   registration that is incorporated into the initial
   segmentation.  Contrast differences are likely to have
   a relatively small effect on the final results.
 
2) The demographics of your subject population differ
   from those used to generate the existing templates and
   prior probability images.  For example, serious problems
   can occur if your subjects have very large ventricles.
   In these data, there would be CSF in regions where the
   existing priors say CSF should not exist.  This would
   force some of the CSF to be classified as white matter,
   seriously affecting the intensity distribution that
   is used to model white matter.  This then has negative
   consequences for the whole of the segmentation.
 
> Can any one please explain the detailed steps to
> construct a customized template image (gray and white
> matter images) for VBM analysis?
 
The following script is one possible way of generating your
own template image.  Note that it takes a while to run, and
does not save any intermediate images that could be useful
for quality control.  Also, if it crashes at any point then
it is difficult to recover the work it has done so far.
 
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * 
make_template.m

 

 
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * 
 
You may also wish to do some manual editing of the images
afterwards - especially to remove extra-skull CSF.  When
everything has finished, simply smooth the images by 8mm
and call them templates and prior probability images.
You can modify the default priors for the segmentation step
in order that the customised ones are used.  This can be done
either by changing spm_defaults.m, or by typing the following
in Matlab:
 
spm_defaults
global defaults
defaults.segment.estimate.priors = ...
     spm_get(3,'*.IMAGE','Select GM,WM & CSF priors');
 
Note that this will be cleared if you reload the defaults.  This
could be done when you start spm, reset the defaults or if the
optimised VBM script is run, as it calls spm_defaults.m.
Alternatively the optimised VBM script could be modified to
include the above.
 
Note that I have only tried the script with three images, so
I don't have a good feel for how robust it is likely to be.
 
>
> Please let me know the number of subjects required to
> construct one?
 
Its hard to say, but more is best.  The 8mm smoothing means
that you can get away with slightly fewer than otherwise.
 
Best regards,
-John
Note: The make_template is an updated version of what was originally posted. It is current as of Sep 9, 2003.

...to the top

Gem 21: ImCalc Script


As posted, this snippet is for SPM2; I've edited it to work with SPM99.
 
Subject: Re: a script to use ImaCal
From: John Ashburner <john@FIL.ION.UCL.AC.UK>
Date: Thu, 16 Oct 2003 11:24:17 +0000 (07:24 EDT)
To: SPM@JISCMAIL.AC.UK
 
> Brain image for each subject to mask out CSF signal was generated by
> using MPR_seg1.img (i1) and MPR_seg2.img (i2) with (i1+i2)>0.5 in
> ImaCal.(called brainmpr.img for each subject)
>
> Then I have more than two-hundred maps, which need to mask out
> CSF. I think I can use ImaCal again with selecting brainmpr.img (i1)
> and FAmap.img (i1), and then calculating (i1.*i2) to generate a new
> image named as bFAmap.img.  Unfortunately, if I use the ImaCal, it
> take so long time to finish all subjects. Could anyone have a script
> to generate  a multiplication imaging with choosing a brain
> image(brainmpr.img, i1) and a map image (FAmap.img, i2) and writing
> an output image (bFAmap.img) from i1.*i2? 
 
You can do this with a script in Matlab.  Something along the lines of the
following should do it:
 
P1=spm_get(Inf,'*.img','Select i1');
P2=spm_get(size(P1,1),'*.img','Select i2');
 
for i=1:size(P1,1),
    P = strvcat(P1(i,:),P2(i,:)));
    Q = ['brainmpr_' num2str(i) '.img'];
    f = '(i1+i2)>0.5';
    flags = {[],[],[],[]};
    Q = spm_imcalc_ui(P,Q,f,flags);
end;
 
Note that I have not tested the above script.  I'm sure you can fix it if
it doesn't work.
 
Best regards,
-John

...to the top

Gem 22: spm_orthviews tricks


IMHO, after spatial normalization, John's key contribution to SPM is spm_orthviews, the function behind the 'Check Reg' button which let's you view many volumes simeltaneously. Some of the Gems use spm_orthviews (e.g. Gems 1 and 16) but listed here are some generally useful tricks for spm_orthviews.

These tricks are useful anytime a three-view orthogonal slice view is shown, whether from useing 'Check Reg', 'Display' or when overlaying blobs from the 'Results' window.

Turn off/on cross-hairs
spm_orthviews('Xhairs','off')
spm_orthviews('Xhairs','on') .

 

Return current x,y,z world space, mm location
spm_orthviews('pos')'
(Note that I transpose the returned value into a row vector, for easier copying and pasting.)

 

Move to x,y,z world space, mm location
spm_orthviews('reposition',[x y z])

 

...to the top

Gem 23: Tabulating T Statistics


 
Subject: Fwd: Re: tabulating all statistics
From: John Ashburner <john@FIL.ION.UCL.AC.UK>
Date: Tue, 1 Jul 2003 11:47:07 +0000 (07:47 EDT)
To: SPM@JISCMAIL.AC.UK
 
> I was wondering if it would be possible to write t values for an
> entire volume into a file? 
 
Try this:
 
   fid=fopen('t-values.txt','w');
   P=spm_get(1,'*.img','Select statistic image');
   V=spm_vol(P);
 
   [x,y,z] = ndgrid(1:V.dim(1),1:V.dim(2),0);
   for i=1:V.dim(3),
     z   = z + 1;
     tmp = spm_sample_vol(V,x,y,z,0);
     msk = find(tmp~=0 & finite(tmp));
     if ~isempty(msk),
       tmp = tmp(msk);
       xyz1=[x(msk)'; y(msk)'; z(msk)'; ones(1,length(msk))];
       xyzt=V.mat(1:3,:)*xyz1;
       for j=1:length(tmp),
         fprintf(fid,'%.4g %.4g %.4g\t%g\n',...
                 xyzt(1,j),xyzt(2,j),xyzt(3,j),tmp(j));
       end;
     end;
   end;
   fclose(fid);
 
best regards,
-John
                
As noted in a 2 Feb 2004 email, to eliminate all voxels below a certain threshold, change
 
     msk = find(tmp~=0 & finite(tmp));
to
 
     msk = find(tmp>0 & finite(tmp));

 

...to the top

Gem 24: Un-normalizing a point


This script of John's will find the corresponding co-ordinate in un-normalised image: get_orig_coord.m

 

...to the top