Hi Andreas,
>
>> you don't need to acquire all your EPI images with both blip-directions.
>
> yes, but I meant something different:
> Given you'd have the choice of acquiring - let's say - either 80 dirs blip
> up/down or 160 dirs in one phase encoding only (along with a set of blip
> up/down B0 images) for bedpostx, what would you choose?
I think the reality is that we haven't done enough testing to say for certain yet. I think if you can afford more directions, say 500+ then I think that collection 250+250 is a good idea. You may still have an increased uncertainty as you pass through a heavily distorted area, but hopefully you should at least reduce any bias tracing through it if you have both blip-directions. This is the decision we have taken in the HCP project for example.
If you can only afford a total of 80 directions then I think I would suggest sampling 80 different directions instead of 40+40.
Your example is in that middle ground where I'm just not certain.
It is possible in eddy to have the cake and eat it. You can collect 80 different directions and still have them be 40+40 in terms of phase-encode blips. You would then chose them such that if you have one dwi with pe+ you should then chose the dwi with the "closest" (largest inner product) diffusion direction to be pe-. What you loose in that case is the possibility to do pairwise least-squares restoration.
So, lots of options, lots of testing and no real ground truth ;-)
> Thanks and best regards from Germany,
Any less cold there?
Jesper
> Andreas
>
> Am 23.03.13 09:11 schrieb "Jesper Andersson" unter <[log in to unmask]>:
>
>> Hi again Andreas,
>>
>>> yes - that's why I was wondering, i.e. Mark must have downsampled /
>>> preregistered or recorded a T2 scan at low res.
>>> Given that such approach may work, would you invest scan time in full
>>> blip up / down recordings (of all diffusion directions) or prefer to
>>> double the dirs in one phase encoding dir?
>>
>> what you can get from the b0-EPI<->T2 combo is an estimate of the field
>> which you can then use to correct all your EPIs. You can achieve exactly
>> the same thing with b0-EPI+<->b0-EPI- where the sign indicates the
>> blip-direction. I.e. you don't need to acquire all your EPI images with
>> both blip-directions.
>>
>> Given that you would actually loose time by acquiring a T2 instead of a
>> single EPI with reversed blips, which would only take you ~5sec. The
>> potential problem with the T2 option is that the contrast might not be
>> perfectly matched which might mean a somewhat poorer estimate of the
>> field. So my vote would always be for reversed blips for at least a
>> couple of b=0 images.
>>
>> But it is nice to know that the T2 option works for when people forget.
>>
>> Jesper
>>
>>> Cheers,
>>> Andreas
>>>
>>>
>>>
>>> Am 23.03.2013 um 08:34 schrieb Jesper Andersson <[log in to unmask]>:
>>>
>>>> Hi Andreas,
>>>>
>>>>> thanks. It does look reassuring. I will try it myself, I quite like
>>>>> the
>>>>> idea.
>>>>> One question: Your T2 was of the same dimensions like your B0 image
>>>>> ord
>>>>> did you downsample it to the B0 resolution? Often, the T2 will be of a
>>>>> higher res than the B0, at least when it come to in plane resolution.
>>>>
>>>> topup expects all images to have the same matrix size, so you will
>>>> typically need to downsample the T2. It is probably also a good idea to
>>>> combine that with an RB registration to make sure they are not too far
>>>> apart initially. There is a movement model inside topup but that is for
>>>> small movements between the different scans and is unlikely to work
>>>> with big position differences that you might get from different
>>>> sequences/FOV.
>>>>
>>>> Jesper
>>>>
>>>>> Cheers,
>>>>> Andreas
>>>>>
>>>>> Am 22.03.13 17:02 schrieb "Mark Drakesmith" unter
>>>>> <[log in to unmask]>:
>>>>>
>>>>>> Hi Andreas
>>>>>>
>>>>>> I've attached images of the T2w, distorted and corrected B0 images.
>>>>>>
>>>>>> The topup call was:
>>>>>>
>>>>>> topup --imain=T2w_images_topup.nii.gz --datain=topup_dti_params.txt
>>>>>> --config=b02b0.cnf --out=my_topup_results --fout=my_field
>>>>>> --iout=my_unwarped_images -v
>>>>>>
>>>>>> The T2w_images_topup.nii.gz image contained the original B0 and the
>>>>>> T2w
>>>>>> images. Note that the B0 image was rigidly transformed with flirt so
>>>>>> it
>>>>>> is in the same voxel space a the T2w. the topup_dti_params.txt file
>>>>>> contained the following lines, corresponding to the B0 and T2w images
>>>>>> respectively. As per Jespers, suggestion the readout time for the T2w
>>>>>> was set close to zero:
>>>>>>
>>>>>> 0 1 0 0.0864
>>>>>> 0 1 0 0.00000001
>>>>>>
>>>>>> I've not yet applied this the full set of DWIs yet, but so far it
>>>>>> looks
>>>>>> good.
>>>>>>
>>>>>> I hope this is of help.
>>>>>>
>>>>>> Mark
>>>>>>
>>>>>>
>>>>>> Dr. Mark Drakesmith
>>>>>> Research Associate
>>>>>>
>>>>>> Caridff University Brain Research Imaging Centre (CUBRIC)
>>>>>> School of Psychology
>>>>>> Cardiff University
>>>>>> Park Place
>>>>>> Cardiff
>>>>>> CF10 3AT
>>>>>>
>>>>>> Tel: +44 (0) 29 2087 0354
>>>>>> Fax: +44 (0) 29 2087 0339
>>>>>>
>>>>>> On 22/03/13 15:34, Andreas Bartsch wrote:
>>>>>>> Hi Mark,
>>>>>>>
>>>>>>> I have followed this with interest.
>>>>>>> Could you post an example (e.g. a slicer output) along with your
>>>>>>> topup
>>>>>>> call?
>>>>>>> You can also email me directly.
>>>>>>> Cheers,
>>>>>>> Andreas
>>>>>>>
>>>>>>> Am 22.03.13 15:53 schrieb "Mark Drakesmith" unter
>>>>>>> <[log in to unmask]>:
>>>>>>>
>>>>>>>> Hi Jesper
>>>>>>>>
>>>>>>>> I just tried it. It works beautifully! I didn't realise you could
>>>>>>>> use
>>>>>>>> topup in that way (I had just followed the example in the FSL
>>>>>>>> wiki).
>>>>>>>>
>>>>>>>> Thanks very much! :)
>>>>>>>>
>>>>>>>> Mark
>>>>>>>
>>>>>>
>>>>>
>>>
>
|