[Insight-users] RE: mri ct reg.

[Lydia Ng]

Hi Nasser,

> mri-LUT_159x191x95_1channels_8bit.raw  (@150ucrons)

Is your data 8bit? I think the application was hard-coded for signed
short?
You would have to modify the image/pixel type if you data is 8bit.

> To start with the data sets are not that misaligned but of course they
are
> of different scale, and this is my problem.  I read the different
> heuristics mentioned on starting the registration process but I don't
know
> what's a good way to set the shrink levels.

My personal heuristic is to start with the two images being of similar
resolution and not downsample too much (i.e. so that the image size does
not fall below say 64 pixels per dimension).

So for your problem -
Fixed:=20
> mri-LUT_159x191x95_1channels_8bit.raw  (@150ucrons)
> 0
> 159 191 95
> 0.943396 0.78534 1.57895

Moving:
> ct_319x319x159_1channels_8bit.raw  (@86.5ucrons)
> 0
> 319 319 159
> 0.27116 0.27116 0.544025

I would try something like
Fixed start shrink factors: 2 2 1
Moving start shrink factors: 8 8 4
Then means starting off with images that are approx isotropic of 2mm^3

> I got the code running but
> when I try to view the results the fixed and registered images are not
of
> same size. Conceptually, what am I missing?
That's strange the registered image should be the same as the fixed.
I wonder if it has to do with the 8bit problem?

- Lydia

> -----Original Message-----
> From: Nasser Kashou [mailto:kashoun at ee . eng . ohio-state . edu]
> Sent: Tuesday, October 21, 2003 3:45 PM
> To: Lydia Ng
> Subject: mri ct reg.
>=20
> Hi Lydia,
> I would like to thank you in advance and the rest of the guys for your
> help.  I have read almost all the emails from the archives related to
> multiresMIregistration example.  I'm now trying to align  the
following
> data sets:
> mri-LUT_159x191x95_1channels_8bit.raw  (@150ucrons)
> 0
> 159 191 95
> 0.943396 0.78534 1.57895
> ct_319x319x159_1channels_8bit.raw  (@86.5ucrons)
> 0
> 319 319 159
> 0.27116 0.27116 0.544025
>=20
> To start with the data sets are not that misaligned but of course they
are
> of different scale, and this is my problem.  I read the different
> heuristics mentioned on starting the registration process but I don't
know
> what's a good way to set the shrink levels. I got the code running but
> when I try to view the results the fixed and registered images are not
of
> same size. Conceptually, what am I missing?
> Thanks