Dear readers,

Here is the answers I received after my request of December about
the use of fiducial points in NMR. So, fiducials points don't
seem to be necessary in image processing of NMR pictures ( as I
hoped ! ).

Serge Van Sint Jan


>From Ton van den Bogert,

Until now, I have always assumed that the voxels in a scan are
a nice rectanbular 3-D grid, so no reference markers are
necessary. Considering how the resonance data are collected, I
can't imagine how a shift between slices could occur.

I have considered that there may be non-linearities in the
relationship between pixel coordinates and real (x,y,z)
coordinates, and should (real soon now) do the experiment with
a precise grid of phantom objects to test the linearity of the
scanner's coordinate system. The result will, of course, depend
on the quality of the scanner (linearity of the magnetic field
gradient etc.).


>From William Coles,

I am currently working on imaging bird and fish heads
looking for magnetic anomallies. I am using preserved heads now
so they don't move. I am collecting multislice images, alligning
the images has not been a problem in this case, because the
samples are in a fixed position.

When using live animals I will be using a 3D imaging
sequence. The 3D data is collected at once and the images are
proccessed together. In my case I will have 36 sequenced images,
so pixel 2,2,1 will be on the same line as pixel 2,2,N through

I am using a SUN IPX with SunVision and XV-2.21 to do my
image processing. SunVision supposedly can manipulate multislice
images before concatenation into a 3D format, I have not tried
this feature as my images have always been alligned.

As for using a reference element I found that with
sequenced slices there was enough of a pattern in the image to
tell me whether they were alligned.

I would recommend looking into using 3D imaging rather
than using multislice images and superimposing them. It will
probably make life a little easier.

Good Luck.

William Coles