>>Hi, I have difficulty getting an undistorted 4D brain which is unbetted
using FEAT GUI. I have tried running FEAT GUI several times with the
"BET brain extraction" turned off. The files output I have are:
1. example_func.nii.gz (3D betted file)
2. filtered_func_data.nii.gz (4D betted file)
(both in the name.feat folder)
3. EF_UD_example_func.nii.gz (in the unwarp subfolder) (3D unbetted 3D file)
Is there a 4D undistorted DTI file which is unbetted, that can be obtained
from running only the B0 distortion correction function in FEAT?
Is there any benefit to use spatial smoothing function in FEAT in the B0
unwarping?
Many Thanks
Siewmin
> Hi,
>>
>> 1 - Look in the unwarp directory - there you should be able to find
>> the images that are used to generate the html page and should
>> contain the skull. You should also turn the "BET brain extraction"
>> button off in the pre-stats tab to avoid this happening.
>>
>> 2 - This is probably just a display issue. Load the image
>> unwarp/ED_UD_fmap_sigloss into FSLView and see what
>> the values are. They should hopefully all be around 1 and
>> have no (or very few) voxels containing zero. If there are
>> lots of voxels containing zero then there is a problem.
>> As to whether there is an advantage to setting it to 0% or
>> leaving it as 10%, I suspect it will not make a big difference.
>> It only affects the registration parts and not the unwarping,
>> so you will still get the unwarping in the inferior regions
>> even when these are not included in this map. Hopefully these
>> areas are unwarped successfully in the end.
>>
>> 3 - You can change the type with:
>> fslmaths imagename imagename -odt short
>> We choose to convert to 32 bit float as the interpolation which
>> is done introduces non-integer values. However, if your dynamic
>> range is large (so that the non-integer portion is negligible as a
>> percentage of the main signal) then you can convert back to
>> 16 bit integer without much loss in accuracy. There isn't any
>> advantage to doing the conversion though, so unless you really
>> need it for space reasons or for compatibility with other software
>> then I would not convert it.
>>
>> All the best,
>> Mark
>>
>>
>> On 4 May 2009, at 05:39, Siew-Min Gan wrote:
>>
>>>> Hi Mark, thanks for your replies. With the GUI output, I have a few
>>> queries, in particular with coversion of the datatype in the
>>> undistorted
>>> file.
>>>
>>> 1.The undistorted images all have skull removed (example_ func and
>>> filtered_func_data). The images of the undistorted 4D dti data on
>>> the feat
>>> report prestats_html page don't have skull removed. How can I get an
>>> undistorted 4D data with skull, or is the default output all
>>> skullless?
>>>
>>> 2. When I select % signal loss threshold as 10, the red region in the
>>> brain in row 2 of the html page( also described as the region that
>>> will be
>>> ignored during registration), are mainly in the temporal and inferior
>>> frontal lobes. This is also the region where most of the distortion
>>> and B0
>>> inhomogeniety occurs in the fieldmap. If this is the case, would it
>>> be
>>> better to select the % signal threshold as 10, so these regions with
>>> signal loss and that also tends to be distorted, would be ignored in
>>> the
>>> registration, or it doesn't matter as you have mentioned in the
>>> previous
>>> email. Next, when I select the % signal loss threshold to 0, the
>>> threshold
>>> signal loss weighting image becomes mainly all red. In the practical
>>> notes, it mentioned that the red regions will be ignored during
>>> registration. Is it just a colour display issue as I thought all the
>>> brain
>>> regions in the 2nd row would be yellow instead of red, because if I
>>> select 0, then none of the regions would not ignored during
>>> registration. I tried to attached a screenshot but the emails
>>> bounced back.
>>>
>>> 3. The result of my undistorted 4D data is of format datatype float
>>> 32. I
>>> would like to convert it to an integer 16 datatype to how the 4D data
>>> originally was? Is there an fsl command that can do that or how
>>> would you
>>> suggest I do that? If I did convert it like this, would there be
>>> some data
>>> loss during the conversion from float to integer?
>>>
>>> Thanks
>>>
>>> Siewmin
>>>
>>> Hi,
>>>>
>>>> I'm glad it is working now.
>>>>
>>>> The answer to which way you need to calculate the fieldmap
>>>> is the first way. The reason is that prelude is a tool for
>>>> unwrapping *phase* only, so it requires its input to be in radians,
>>>> in the range -pi to pi (or 0 to 2*pi). The second method you list
>>>> divides by the TE *before* running prelude, which will stop prelude
>>>> working correctly, and so it will not unwrap the phase.
>>>>
>>>> There should be no problems with running things using the
>>>> first method of preparing your fieldmap.
>>>>
>>>> All the best,
>>>> Mark
>>>>
>>>>
>>>> On 3 May 2009, at 02:07, Siew-Min Gan wrote:
>>>>
>>>>>> Hi,thanks, I tried again and it worked fine. The unwarp directory
>>>>>> and
>>>>> GUI are both looking good. I think it's because the intensity of my
>>>>> prepared fieldmap is not right the first time. I have input a
>>>>> fieldmap
>>>>> intensty of -pi to pi then and it didn't work.
>>>>>
>>>>> I have rechecked that my fieldmap would be similiar to the
>>>>> phase-difference map in the fsl practical. It worked after I have
>>>>> prepared
>>>>> it as in the instructions on the practical. However, i noticed I get
>>>>> different results if I prepare it slightly differently from the
>>>>> suggestions on the email list.
>>>>>
>>>>> I have posted a question about which way and order to best calculate
>>>>> and
>>>>> prepare the fieldmap in another email i posted today "Order of
>>>>> preparing
>>>>> Phase_Difference Map to rad/secs". Can I ask you this question on
>>>>> this
>>>>> email regarding this?
>>>>>
>>>>> I get these 2 fieldmaps which are similiar in calculations but of
>>>>> different order of calculation and unwarping, and hence different
>>>>> intensities.
>>>>>
>>>>> The first way has a phasedif map of 0 to 4094 intensity range). It
>>>>> is
>>>>> rescaled to -pi to pi, then unwarped in prelude, then changed to
>>>>> rad/
>>>>> sec
>>>>> by div on the TE dif (0.00246). The resulting fieldmap has an
>>>>> intensity
>>>>> range of -3165.861328 to 2801.69702.
>>>>> The second way has a phasedifmap of -4096 to 4092 intensity. It is
>>>>> rescaled by dv 4096, -mul by 3.14 then div 0.00246, followed by
>>>>> prelude.
>>>>> The resultant fieldmap has intensity -2801.073486 to 2790.472900.
>>>>>
>>>>> I'm not sure which one would be the more correct way, or if it
>>>>> doesn't
>>>>> matter? The undistorted dti using 1st way of preparation from the
>>>>> fsl
>>>>> practical instructions is a bit more undistorted than the results of
>>>>> the
>>>>> 2nd one in the -y direction ( unwarped direction). Why that would be
>>>>> the
>>>>> case?
>>>>>
>>>>> Thanks
>>>>>
>>>>> Siewmin
>>>>>>
>>>>>> It sounds like it hasn't worked unfortunately.
>>>>>> The errors in fslview are simply because the images are in an
>>>>>> incomplete
>>>>> feat directory. If you change the name of the directory (so it
>>>>>> doesn't end in
>>>>>> .feat) or move the files outside of the feat directory. However,
>>>>>> the
>>>>> static
>>>>>> movies and lack of difference between distorted and undistorted is
>>>>>> a much
>>>>>> worse problem.
>>>>>>
>>>>>> What files are in your unwarp directory (within the feat
>>>>>> directory)? And
>>>>> what images are you passing into the FEAT GUI?
>>>>>>
>>>>>> As for the rescaling - that is simply something that FEAT does
>>>>>> automatically (and unfortunately can't be turned off) but it won't
>>>>> affect any processing, as the diffusion fitting process will remove
>>>>> any
>>>>> global scaling.
>>>>>>
>>>>>> Let me know about the unwarp directory and GUI setup and
>>>>>> hopefully we can sort out the problem.
>>>>>>
>>>>>> All the best,
>>>>>> Mark
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> On 1 May 2009, at 19:55, Siewmin Gan wrote:
>>>>>>
>>>>>>> Hi,
>>>>>>> I have a couple of questions regarding the output data when I
>>>>>>> used the FEAT Gui to do
>>>>>>> distortion correction of DTI data. The movie section comparing
>>>>>>> distorted and undistorted
>>>>>>> example_func data (third last rows) showed a stationary row of
>>>>>>> images ( i.e there is no
>>>>>>> flipping between one image to another). When I tried to fslview
>>>>>>> the
>>>>> example_func.nii.gz,
>>>>>>> filtered_func_data.nii.gz or the
>>>>>>> example_func_orig_distorted.nii.gz, i
>>>>> could only view the
>>>>>>> first two and not able to view the last image. Also in all cases,
>>>>>>> there
>>>>> would be error
>>>>>>> messages as below for all three:
>>>>>>> ERROR (nifti_image_read): failed to find header file for
>>>>>>> '/Users/siewmingan/Desktop/DATASET/patient_1/tryfieldmap_y.feat/
>>>>> stats/pe1'
>>>>>>> ** ERROR:
>>>>>>> nifti_image_open(/Users/siewmingan/Desktop/DATASET/patient_1/
>>>>>>> tryfieldmap_y.feat/stat
>>>>>>> s/pe1): bad header info
>>>>>>> Error: failed to open file
>>>>>>> /Users/siewmingan/Desktop/DATASET/patient_1/tryfieldmap_y.feat/
>>>>>>> stats/ pe1.
>>>>>>> Why does the error message occurs and how I can resolve it?
>>>>>>> Because the movie section did not display the contrast between the
>>>>> distorted and
>>>>>>> undistorted images, and the 4th row displaying the original
>>>>>>> distorted images looks
>>>>>>> similiar to the 5th row displaying the undistorted images , and I
>>>>> couldn't view the
>>>>>>> example_func.nii with the example_func_orig_distorted.nii.gz, I
>>>>>>> then
>>>>> tried to compare the
>>>>>>> example_func.nii.gz with my original 4D data file. I noticed the
>>>>> intensity in the
>>>>>>> example_func.nii.gz file is around 2-3 times lower than the
>>>>>>> original 4D
>>>>> data. May I ask
>>>>>>> why that would be so and would that be a problem? Also, if both
>>>>>>> the
>>>>> example_func.nii.gz
>>>>>>> file and the filtered_func_data.nii.gz are both the undistorted 4D
>>>>>>> DTI
>>>>> data from running
>>>>>>> FEAT, why do they have different intensity with the former in 100s
>>>>>>> and
>>>>> the latter in
>>>>>>> 10000s?
>>>>>>> Apart from the difference in the intensity in the original 4D data
>>>>>>> and the
>>>>>>> example_func.nii.gz, it was quite hard to detect differences due
>>>>>>> to the
>>>>> different intensity
>>>>>>> of the brain on display ( I've tried both distortion correction
>>>>>>> with
>>>>> the "y" and "-y" option.)
>>>>>>> Is there any way to resolve the previous error messages so I can
>>>>> directly compare the
>>>>>>> distorted and undistorted images with the same intensity, and
>>>>>>> check if
>>>>> I have done the
>>>>>>> distortion correction step correctly?
>>>>>>> Many Thanks
>>>>>>> Siewmin
>>>>>>
>>>>>
>>>>
>>>
>>
>
|
