Thanks for all the quick replies. After reading through the FUGUE webpage and 
epidewarp.fsl codes more carefully, I did pretty much what you suggested 
below. I wasn't sure if I needed to run prelude before finally converting the 
image in rad/s (-div .00246) so I ran Step 3 through anyway but the phase 
image values didn't change except being masked by the magnitude image. So 
the steps I ran are pretty much the same as you described.

1. Using dcm2nii to get dicom into nii.gz, which yield a phase image with range 
of -4096 to 4092.
2. fslmaths gre_pha -mul 3.14159 -div 4096 tmp1
3. prelude -a gre_mag_brain -p tmp1 -o tmp2
4. fslmaths tmp2 -mul 1000 -div 2.46 gre_pha_rads

Step 3 didn't change the values of tmp1 but just masked with the brain, so I 
guess it is not important. I am glad to confirm it with your one line command 
below since I wasn't sure what unit the original phase image is in and what is 
the right conversion to get it in rad/s unit required by FEAT.

After getting this sorted out, the FEAT results are just as good as those from 
the epidewarp.fsl script.

Cheers, XL

On Fri, 29 Aug 2008 15:43:45 -0400, Heather L. Urry 
<[log in to unmask]> wrote:

>Hi XL,
>
>After reconstructing my Siemens Trio images using MRIcron, I get a single
>real phase image like the one you describe, with a range of -4096 to 4092.
>I use the following fslmaths command to convert it to radians/second:
>
>fslmaths phase.nii.gz -mul 3.14159 -div 4096 -div .00246 fieldmap.nii.gz
>
>The .00246 refers to the difference in TEs for the two fieldmap acquisitions
>in seconds. This produces a fieldmap.nii.gz with a range of -1277.069214 to
>1275.822021.  Prior to dividing by .00246, the range is basically -pi to
>pi.  In any case, I use the output of that command (fieldmap.nii.gz) in FEAT
>for B0 unwarping. It works well for me, and so if your phase image is a
>real, unwrapped phase image, then perhaps this will work for you too.
>
>Cheers,
>
>Heather
>
>
>
>
>On Fri, Aug 29, 2008 at 1:06 PM, X Liu <[log in to unmask]> wrote:
>
>> Dear FSL developers and experts, I have been testing EPI unwarp using the
>> new version of FSL 4.1.0 within FEAT and epidewarp.fsl script (v1.32). I
>> would
>> like to confirm a couple of issues I have encountered.
>>
>> 1. Our Siemens Allegra scanner produces two GRE field maps (one 
magnitude
>> and one phase). The phase map has a range from -4096 to 4092, which 
looks
>> like the real fieldmap as at
>> http://www.fmrib.ox.ac.uk/fsl/fugue/feat_fieldmap.html (last image). When
>> using epidewarp.fsl script (with TE difference, EPI echo spacing
>> parameters),
>> without rescaling the range, prelude complains about the values out of
>> range.
>> FEAT doesn't complain and move forward since it doesn't call prelude, but
>> the
>> result looks over stretched.
>>
>> 2. By rescaling the range of the phase map (only divided by 2 so that the
>> range is -2048 to 2046, is this correct, or I have to make the range from 0
>> to
>> 4095, or I still need to go through other steps involving prelude and TE
>> difference value according to the above webpage?), both epidewarp.fsl and
>> FEAT produce reasonable unwarping results. But the results from
>> epidewarp.fsl
>> seem less stretched and more reasonably shaped (e.g., the brain stem 
seems
>> being pushed more backward by FEAT's dewarping). I looked into the 
source
>> codes of featlib.tcl and found that most of the major steps are consistent
>> across two approaches (with extra steps in FEAT) but couldn't identify 
what
>> causes the difference in results.
>>
>> 3. For the final unwarping, FEAT uses more recent functions - applywarp 
and
>> convertwarp, whereas epidewarp.fsl uses just fugue and shift map. I 
assume
>> they are equivalent but the former is more flexible in terms of combining
>> the
>> unwarp and other transformation (e.g., motion parameters, EPI to T1 affine
>> or
>> nonlinear).
>>
>> 4. For demean the unwarp map, epidewarp.fsl subtracts the mean from the
>> shift map, whereas FEAT subtracts the 50% percentile from the phase map.
>> Does this difference matter?
>>
>> Regards, XL
>>
>
>
>
>--
>Heather L. Urry
>Department of Psychology
>Tufts University
>490 Boston Avenue
>Medford, MA 02155
>
>email: [log in to unmask]
>phone:  617-627-3733
>fax: 617-627-3181
>