Hi Mark,
THanks for the advice. I still am not having much success with the
distortion correction. THe images are slightly corrected but the
improvement is not much. Are there any obvious things we could be doing
wrong? The fieldmap looks sensible so I doubt the spin echo acquisition is
at fault. I've given it a pretty tought test with a phantom with a lot of
distortion and maybe I should use something simpler...
Thanks,
Gaby
>Hi Gaby,
>
>I'm not sure if there is a "best" echo offset given the variations in
>scanner and field inhomogeneities. Obviously you don't want too
>much phase wrapping which means a small difference, but you
>want a large signal, hence a big difference. Somewhere in there
>is a compromise which will work the best. Ideally, however, you
>could take a range of times and fit the best curve. I'd like to
>generalise fugue to allow for this, but it isn't a high priority right
>now.
>
>As for underestimating the distortion, there might be some problems
>associated with point spread functions when the voxels are large,
>leading to blurring of the complex signal and bad phase estimation.
>To check the acquisition and analysis, I would use a phantom of
>precisely
>known geometric shape in with something to give some B0 inhomogeneity,
>then take images with different offsets and different phase encode
>directions
>(especially the sign - that is, top to bottom vs bottom to top). If
>there is some
>systematic underestimate it should show up different when the region is
>compressed or stretch, as the resultant "unwarped" image would still
>have
>residual compression or stretch. However, signal loss would always just
>leave a dark patch. Comparing the fieldmap values at different echo
>offsets
>should also give you a feeling for the precision and repeatability of
>the
>estimate and might show a problem at short/long differences.
>
>I know that is all pretty common sense, but hopefully it will be of
>some help.
>Best of luck with it Gaby, and let me know if you still can't get things
>working well.
>
>Cheers,
> Mark
>
>
>
>On Wednesday, September 3, 2003, at 06:19 am, Gaby Pell wrote:
>
>> Dear FSL-gurus (but especially Mark!),
>>
>>
>>
>> Do you have a feeling for what is the best aysmmetric echo offset
>> gives the best field map for distortion creation? I am using 10ms and
>> am just about to try collecting a wider range. The corrections I am
>> getting with FUGUE at the moment look reasonable but there is still a
>> fair bit of distortion that is not corrected for (and this isn’t
>> signal loss). Is there anything else I could be doing wrong with the
>> analysis. Can you suggest any tests that would help me to localise
>> this problem either to the acquisition or the analysis?
>>
>>
>>
>> Thanks,
>>
>>
>>
>> Gaby
>>
>>
>>
>> P.S. I have tried different levels of regularisation
>>
>>
>>
>> ***************************************
>> Gaby Pell, PhD
>> Brain Research Institute
>> Ground Floor,NeurosciencesBuilding
>> Austin & Repatriation Medical Centre
>> Banksia Street
>> West Heidelberg,
>> Melbourne,
>> Victoria, 3081
>> Australia
>> Tel: (+61 3) 9496-2868 (W)
>> Tel: {+61 3) 9527 6441 (H)
>> Fax: (+61 3) 9496-4071
>> Email: [log in to unmask]
>>
>>
>>
>>
>>
>
>Hi Gaby,
>
>
>I'm not sure if there is a "best" echo offset given the variations in
>
>scanner and field inhomogeneities. Obviously you don't want too
>
>much phase wrapping which means a small difference, but you
>
>want a large signal, hence a big difference. Somewhere in there
>
>is a compromise which will work the best. Ideally, however, you
>
>could take a range of times and fit the best curve. I'd like to
>
>generalise fugue to allow for this, but it isn't a high priority right
>
>now.
>
>
>As for underestimating the distortion, there might be some problems
>
>associated with point spread functions when the voxels are large,
>
>leading to blurring of the complex signal and bad phase estimation.
>
>To check the acquisition and analysis, I would use a phantom of
>precisely
>
>known geometric shape in with something to give some B0 inhomogeneity,
>
>then take images with different offsets and different phase encode
>directions
>
>(especially the sign - that is, top to bottom vs bottom to top). If
>there is some
>
>systematic underestimate it should show up different when the region is
>
>compressed or stretch, as the resultant "unwarped" image would still
>have
>
>residual compression or stretch. However, signal loss would always
>just
>
>leave a dark patch. Comparing the fieldmap values at different echo
>offsets
>
>should also give you a feeling for the precision and repeatability of
>the
>
>estimate and might show a problem at short/long differences.
>
>
>I know that is all pretty common sense, but hopefully it will be of
>some help.
>
>Best of luck with it Gaby, and let me know if you still can't get
>things
>
>working well.
>
>
>Cheers,
>
> Mark
>
>
>
>
>On Wednesday, September 3, 2003, at 06:19 am, Gaby Pell wrote:
>
>
|