On Sat, Mar 06, 2010 at 03:09:54PM -0500, Judd Storrs wrote:
> On Fri, Mar 5, 2010 at 1:36 AM, Søren Hauberg <so...@hauberg.org> wrote:
> >     2. It would be nice if the help text more clear explains what a
> >        'phantom' is. Personally, I don't know what it is.
> 
> FYI,
> 
> A "phantom" is basically a non-human object used to assess image
> reconstruction quality in medical imaging. The Shepp-Logan is a
> digital 3D phantom that is designed to have grossly brain-like
> features but also at the same time have a precisely defined
> mathematical structure so that reconstruction algorithms can be tested
> by simulation and comparision to the known ground truth. (The concept
> of truth is problematic in medical imaging because there are multiple
> (if not infinite) plausible images that correspond to the acquired
> data) I think Shepp-Logan is in designed to even be analytically
> useful.

Not quite. The original Shepp-Logan phantom is a 2D phantom, showing a
slice through the head:
http://bigwww.epfl.ch/thevenaz/shepplogan/
This is what is (hopefully) implemented in phantom.m (the fastest way is
looking whether the description talks about 'ellipses' versus
'ellipsoids').

There is an extension to the 3D case, where a certain slice through the
head has the properties of the 2D original.

Shepp-Logan is characterized by very small jumps in density, making it
difficult for algorithms at that time (in other words, a real
challenge). Unfortunately, instead of improving their algorithms, people
started creating 'modified' phantoms, with higher jumps to get more
contrast in their results.

        Thomas

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
_______________________________________________
Octave-dev mailing list
Octave-dev@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/octave-dev

Reply via email to