Hi Jesper,
Thanks for your quick response. The documentation online is definitely enough for me to run it with --repol, but I do have two questions. First, how does it determine what the threshold for replacement is? That is, does it perform replacement for every slice that is identified as an outlier in the first place, no matter how severe it is? I ask because the contents of the outlier_report from running Eddy without any replacement produce the following:
Slice 32 in scan 67 is an outlier -43.4565 standard deviations off
Slice 33 in scan 47 is an outlier 7.62952 standard deviations off
Slice 57 in scan 59 is an outlier -5.1364 standard deviations off
Slice 59 in scan 59 is an outlier -4.25245 standard deviations off
Slice 62 in scan 63 is an outlier -14.7206 standard deviations off
As you can see the slice I identified is clearly the glaring one. But it also identifies 4 other slices in addition. Actually, after looking at that last one, slice 62 in scan 63, that one looks like there is severe signal dropout, too. However, axial slice 62 out of 70 contains barely any brain, it is the very top of cortex. Anyway, the other 3 slices it identifies seem fine to me. Is it just that my eyes mislead me, they really are a problem, and in turn I should want --repol to take care of all of them?
Second question: if the answer to the first question is yes, do I need to turn on ol_pos because the 2nd outlier slice is a positive 7.629 STD off?
Thanks again!
-Ben
> On Jan 18, 2017, at 9:58 AM, Jesper Andersson <[log in to unmask]> wrote:
>
> Dear Ben,
>
>>
>> I am concerned about a potential artifact but am hoping that it is something I don't need to worry about after pre-processing. The data come from the following sequence: First, 1 B0 image with P->A phase encoding. Immediately after, 10 B0 images with A->P phase encoding. Immediately after, 60 DWI images with A->P phase encoding at b=1000. There are 70 diffusion directions total. For preprocessing, I use TOPUP (b02b0.cnf) followed by Eddy (w/ TOPUP output, otherwise default settings).
>>
>> I noticed that in one slice of the DWI, solely in the fourth to last volume, there is a slice that looks very off. Ive uploaded several screenshots showing what it looks like before as well as after preprocessing. They can be found here: https://urldefense.proofpoint.com/v2/url?u=http-3A__www.dropbox.com_sh_6pmk7sl94bk8koz_AACx5io2-5FGZc1mIxf4GBjc2ia-3Fdl-3D0&d=DwIFAg&c=kbmfwr1Yojg42sGEpaQh5ofMHBeTl9EI2eaqQZhHbOU&r=W4J-adUY7z1HJlARJ25wulCS4wTkKmprrGB3MGB5JN4&m=dPw4AcNTruUvGylKCI8LbSrQnOyAjaw1oRipoIs7gVE&s=uFp04kug5FnPNRMgthSLsXjxaUunr8HXuMIOp80uaSY&e=
>>
>> As you can see, before preprocessing it appears to be on one single slice. Afterwards, it appears to be on half of each of two slices (still just one volume). The right hemisphere side of the originally affected slice, and the left hemisphere side of one slice superior. The "_Left" files refer to the slice where the left hemisphere appears to be affected after preprocessing whereas "_Right" is the slice where the right hemisphere seems affected.
>>
>> I apologize for asking such a subjective question, but does this look like an artifact that is a problem even after preprocessing? If so, any advice for how to fix it before I run bedpostX? Would this be something that outlier replacement in Eddy is good for, for example?
>
> it looks like exactly the kind of problem that the outlier replacement is there for. I think/hope there should be enough documentation online to help you with that. If not, get back to us.
>
> Jesper
>
>
>>
>> Thank you all very much!
>>
>> -Ben Chernoff
|