Hi Jiasong and Glenn,
I have a suggestion and a question:

Suggestion:
One other way of creating a mask is to consider voxels where FA> 0.15 for
example. Usually, this creates quite a good mask. The idea I took out of
last sentence on the first page of
Derek K. Jones et al. Magnetic Resonance in Medicine 42: 37-41 (1999).

When you use SPM99 you would do this at the step when it says
"Threshold masking   none     absolute    prop'nal"
Then click on absolute and type in your threshold (example 0.15)

Question:
I tried to segment the b0 (non-diffusion-weighted) images. My problem
started when I had to binarize the image. Suppose one takes the white matter
probability maps. What is considered white matter? Those voxels with
probability > 0.8 for example? or everything within that segment even if the
probability is 0.1? I realize that masking with FA > 0.2 has a similar
problem so it would be good to hear some comments if others have been
thinking about this. Thanks.

Good luck Jiasong,
Zoltan

**********************************
Zoltan Nagy
Karolinska Institutet
Neonatology
Astrid Lindgren's Children's Hospital Q2:07
171 76 Stockholm, Sweden

Phone:  +46-8-517-77354
Fax:    +46-8-517-77353
E-mail: [log in to unmask]
**********************************


----- Original Message -----
From: "Glenn Stebbins" <[log in to unmask]>
To: <[log in to unmask]>
Sent: Friday, May 14, 2004 4:21 PM
Subject: Re: [SPM] t-test on FA maps


> Hi Jiansong,
>         One possibility is to use the 0 diffusion-weighted image (the T2
> image).  You could segment this image into separate compartments.  You
> could then take the white matter segment and digitalize it to create
> the mask.  You could then apply the mask to the FA volumes.  Note
> carefully Volkmar Glauche's comments about normalization.  This is
> particularly tricky with EPI images.
>
> -Skip
>
> On May 14, 2004, at 9:12 AM, Xu, Jiansong wrote:
>
> > Dear Sir:
> >
> > Would you tell me how to create and use the mask?
> >
> > Thanks a lot.
> >
> > Jiansong
> >
> > -----Original Message-----
> > From: Glenn Stebbins [mailto:[log in to unmask]]
> > Sent: Thursday, May 13, 2004 7:05 PM
> > To: [log in to unmask]
> > Subject: Re: [SPM] t-test on FA maps
> >
> >
> > Hi Sunil,
> >         It would seem best to use the two-sample t-test module (or the
> > ANOVA),
> > but you should be very careful in your interpretation of the results.
> > If the images are not masked in some way, you may not be comparing
> > similar tissue types between groups. For example, in elderly subjects
> > you would have increased atrophy with increased ventricles.  When
> > compared to younger brains, you would get a significant difference in
> > FA around the ventricles because you would be comparing white matter
> > (in the younger brains) to CSF (in the elderly).  If you mask the
> > images first, you may be able to avoid these issues.
> >
> > -Skip
> >
> >
> > On May 13, 2004, at 6:34 PM, Sunil Patel wrote:
> >
> >> Hello,
> >>
> >> I'm a relatively new user of SPM(99).  I have two sets of FA maps and
> >> wanted to compare each individual patient with the control group (10
> >> normals) as a whole.  The SPM99 manual says that a one-sample t-test
> >> will determine if the mean of a group is significantly different from
> >> 0.  Thus, I could subtract the patient FA map from each of the control
> >> FA maps and then do a one sample t-test on the difference images.
> >> Alternatively, could I simply do a two sample t-test and have one
> >> subject (patient) in one of the two groups?
> >>
> >> Thanks in advance for any advice.
> >>
> >> Sunil
> >>
> >>
> > Glenn T. Stebbins, Ph.D.
> > Associate Professor of Neurological Sciences
> > Tel:  312-563-3854
> > Fax: 312-563-4009
> > email: [log in to unmask]
> >
> >
> Glenn T. Stebbins, Ph.D.
> Associate Professor of Neurological Sciences
> Tel:  312-563-3854
> Fax: 312-563-4009
> email: [log in to unmask]
>