[MINC-users] PET analysis on MINC files

[Christopher Bailey]

On Tue, 2004-11-30 at 12:28, Morgan Hough wrote:
> Dear Chris,
> 
> I am interested to know what it would take to do PET analysis with FSL
> Tools. With the MINC support in place, what tools would need changes?

I'll be bold and try to answer this question. In short: none. With
proper choices of parameters, the FSL tools can be used directly for PET
(blood flow) data.

> McFLIRT, FLIRT - should these work as is?

In principle, they should work. MCFLIRT is the motion correction script
calling FLIRT, the registration program, in an EPI-optimized way. You'd
probably want to consider having a look at MCFLIRT's options before
using it on PET frames. 

FLIRT is not PET2MR-optimized, so it may not produce ideal results
without some tweaking.

> FUGUE, BET and FAST - is there a inhomogenity problem to deal with? 
> 	- is the intensity modeling just the kinetics
> 	- does it help to have the tissue types?

FUGUE: No, PET has no inhomogeneity issues.
BET: Brain extraction (nothing to do with PET).
FAST: Segmentation, also not relevant for PET.

> FEAT - If GLIM is just GLM then I assume this is fine 
> 	- is everything block design in PET?
> 	- is there an equivalent to the HRF? 

There are some default settings of FEAT that should be changed when
using FEAT on a PET dataset. There has been info on this on the FSL
mailing list. Basically, highpass filtering is out. So is FILM, a
prewhitening step before FMRI model fitting. Global normalization might
make more sense in PET.

There's no PET HRF. PET acquisitions are blocked, yes. The GLM is
exactly what's needed.


> MELODIC - does anybody use ICA with PET? I don't see many papers but
> there are some.

MELODIC is a special case of ICA: it's "spatio-temporal". In PET, we
don't have the temporal dimension of FMRI, so it's not relevant. ICA can
and has been used in PET applications, but only in "spatial" mode, i.e.
looking for "spatially independent components" in 3D volumetric data. So
MELODIC's not relevant for PET either.

> I have been reading some papers about the kinetic models and it seems
> like with increasing temporal resolution there are some opportunities to
> really improve analysis. Do you think this is the case? Are these kinds
> of scanners common now?
> 

Kinetic modelling of tracer kinetics using PET is an active field, yes.
The temporal resolution involved, however, is on the order of 10
seconds. That is typically enough to study the kinetics of even
metabolised tracers in the body. I think PET is moving into higher
spatial, not temporal resolutions.

Best regards,
-Chris

-- 
Christopher Bailey <cjb@pet.auh.dk>
PET Centre and
Center for Functionally Integrative Neuroscience
Aarhus University Hospital, Denmark
http://www.cfin.au.dk/