What you really want to simulate is an underlying geometric deformation
(like atrophy) that gets propagated through an image formation model to
create an atrophic image. I don't think multiplying the image intensities
is a good model for what you want. Atrophy doesn't look like a uniform
darkening of the gray matter.
On Tue, 4 Aug 2015, Brent Womble wrote:
Yes, I'm multiplying the image intensities by a scaling factor. Sorry if
that wasn't clear.
-Brent
On Aug 4, 2015 12:09, "Bruce Fischl" <fis...@nmr.mgh.harvard.edu> wrote:
Hi Brent
a quick look and it seems that your answer is "yes"? You are
multiplying the image intensities by some scale factor?
Bruce
On Tue, 4 Aug 2015, Brent Womble wrote:
Yes. Here is the MATLAB script:
for i = [1:20]
%Load the original image and gray matter mask
raw=load_nii([pwd,'/', num2str(i),
'/pre.nii'])
a=single(squeeze(raw.img));
c1_raw=load_nii([pwd,'/', num2str(i),
'/c1pre.nii']);
c1=squeeze(c1_raw.img);
%Mask is gray around edges. Threshold to make
it binary
c1(c1>0.1)=1;
%Convert mask to logical
c1=logical(c1);
%%Generate a 3D gaussian kernel
%Specify the origin, size, and intensity of
the kernel
k_origin=origins(i,:); %origins is an i by 3
vector to store
all coordinates
k_size=20;
k_intensity=100;
k=1-fspecial3('gaussian',k_size).*k_intensity;
k(k<0)=0;
%Expand the kernel to the size of the original
image
padsize_pre=k_origin - (k_size/2);
padsize_post=size(a) - k_origin - (k_size/2);
k=padarray(k,padsize_post,1,'post');
k=padarray(k,padsize_pre,1,'pre');
%Restrict the kernel to the gray mask
b=k.*c1;
b(~c1)=1;
%%Apply the kernel to the original image
c=b.*a;
%Write to .nii
raw.img=c;
save_nii(raw,[pwd,'/', num2str(i),
'/dense-synth.nii']);
end
On Tue, Aug 4, 2015 at 11:52 AM, Bruce Fischl
<fis...@nmr.mgh.harvard.edu>
wrote:
Hi Brent
are you saying you just multiplied the gray
matter intensities
by some scale factor (>1)?
Bruce
On Tue, 4 Aug 2015, Brent Womble wrote:
Hi everyone,
I've been making synthetic brains to
test how
Freesurfer handles various
structural changes.
One of the changes I'm testing is
increased density
(like in VBM). To
simulate increased density, I used a
spherical
Gaussian multiplication
kernel, with a radius of 20 voxels. I
centered this
kernel at a point in
the right superior frontal gyrus and
masked the
changes to grey matter using
the segmentation output from SPM12.
Here is an example, a difference image
(synth.nii -
pre.nii):
[IMAGE]
Around that area, Freesurfer didn't
detect a change
(as I expected):
Inline image 2
Original in red and blue, synthetic in
pink and
light blue.
The problem is that the longitudinal
streams in
Freesurfer found a bunch of
changes in random parts of the brain.
We looked at the recons for each subject
individually, and saw some noise
around the clusters that were
significant in the
two-stage model. This
doesn't make sense, because the actual
voxels were
exactly the same in that
region. For example:
[IMAGE]
I redid the recon-all with the
-nonormalization
flag. It didn't fix the
problem. For example (same subject as
above):
[IMAGE]
I thought it might have been an issue
caused by my
increased density method,
so I ran it again, this time with the
kernel
centered in the cerebellum
(still masked to grey matter). I still
got weird
changes in unrelated parts
of the cortex. For example (same subject
as above):
[IMAGE]
Note: light blue/red and blue/red are
reversed in
this one, just because of
the order I loaded it in Freeview.
In summary, Freesurfer found a bunch of
weird
changes that were nowhere near
the changes I actually made.
Where should I go from here?
-Brent
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
The information in this e-mail is intended
only for the person
to whom it is
addressed. If you believe this e-mail was sent
to you in error
and the e-mail
contains patient information, please contact
the Partners
Compliance HelpLine at
http://www.partners.org/complianceline . If
the e-mail was sent
to you in error
but does not contain patient information,
please contact the
sender and properly
dispose of the e-mail.
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
The information in this e-mail is intended only for the person
to whom it is
addressed. If you believe this e-mail was sent to you in error
and the e-mail
contains patient information, please contact the Partners
Compliance HelpLine at
http://www.partners.org/complianceline . If the e-mail was sent
to you in error
but does not contain patient information, please contact the
sender and properly
dispose of the e-mail.
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
The information in this e-mail is intended only for the person to whom it is
addressed. If you believe this e-mail was sent to you in error and the e-mail
contains patient information, please contact the Partners Compliance HelpLine at
http://www.partners.org/complianceline . If the e-mail was sent to you in error
but does not contain patient information, please contact the sender and properly
dispose of the e-mail.
