First, let us clear some of the unneeded variables from the workspace:
clear PETsummed1 ratio clipneg clippos
Now, we will use PETsummed2 for a very simple form of image
masking: namely, we will use only pixels whose values in the
integrated image are greater than the mean of the integrated image.
For CBF studies, this usually creates a usable mask that contains only
in-brain voxels. It can be done by calculating the mean of the
integrated image, and creating another image consisting entirely of
1's and 0's (a ``binary'' image), where a 1 indicates that the pixel
is inside the head, and a zero indicates outside:
avg = mean (PETsummed2);
mask = PETsummed2 > avg;
We can view the mask as an image itself to see a silhouette of the brain at this slice:
figure;
viewimage (mask);
Or, we can apply the mask to the summed image and view it. To apply
the mask to an image, simply perform an element-by-element multiply
using the
.* operator on the mask and the image; since mask
consists entirely of ones and zeros, every point in the image will
either be set to zero or unchanged. Thus,
PETsummed2 = PETsummed2 .* mask;
figure;
viewimage (PETsummed2);
will display the trapezoidally integrated image with out-of-head data
set to zero.
Finally, to see just how many pixels are ``on'' in the mask, you can type
size (find (mask))
The fact that only a fraction of the entire image consists of image
data means that masking can be used to great advantage when
performing analysis that requires computations for every pixel.
(Generally, this should be avoided if it is all possible to take
advantage of MATLAB's vectorized structure.)