Hello again,
Thanks a lot for your advices. So I have been attempting to run HCP_fix on the
tfMRI data but am not sure about the proper procedure. I would really
appreciate it if you could at least point me to the right documentation.
1. Running HCP_fix directly on the CIFTI file
tfMRI_WM_LR_Atlas_MSMAll.dtseries.nii does not seem to work. Would the
volumetric file tfMRI_WM_LR.nii.gz be the correct starting point? If yes, what
happens then to the MSM registration?
2. Besides the file Movement_Regressors.txt, what other input data does the
script need? This is the error message I got:
nmrmac31:fix1.06 adrian$ ./HCP_fix ~/Downloads/tfMRI_WM_LR.nii.gz 2000
processing FMRI file tfMRI_WM_LR with highpass 2000
running highpass
running MELODIC
running FIX
FIX Feature extraction for Melodic output directory: tfMRI_WM_LR_hp2000.ica
create edge masks
run FAST
Image Exception : #22 :: ERROR: Could not open image
tfMRI_WM_LR_hp2000.ica/reg/highres
libc++abi.dylib: terminating with uncaught exception of type
RBD_COMMON::BaseException
/Users/adrian/Downloads/fix1.06/fix_0b_apply_fast: line 11: 6072 Abort trap: 6
fast -t 1 -o ${folders}/fix/fastsg ${folders}/reg/highres
Image Exception : #22 :: ERROR: Could not open image
tfMRI_WM_LR_hp2000.ica/fix/fastsg_pveseg
libc++abi.dylib: terminating with uncaught exception of type
RBD_COMMON::BaseException
I would be very grateful for your suggestions. Thanks a lot!
Best,
Adrian
________________________________________
Von: Glasser, Matthew [[email protected]]
Gesendet: Mittwoch, 3. August 2016 12:50
An: Chan, Adrian Ka Ho
Cc: [email protected]; Burgess, Gregory
Betreff: Re: AW: [HCP-Users] FIX-ICA denoised task data?
Not at this time, though I would like to see this done. Note that I used
the term “structured noise” rather than physiological noise, though
physiological noise will be part of structured noise. It is worth noting
that ICA+FIX is based on spatial ICA and is not able to remove global
physiological noise (it removes spatially specific physiological and other
structured noise). We are working on this issue however. ICA+FIX removes
many kinds of structured noise, including that from movement. Another way
of thinking about z-stats going down in the face of known structured noise
in the data is that the z-stats in the current data are likely positively
inflated to some degree because of correlation between the structured
noise and the task design.
The phase encoding direction results in many differences in the mapping of
neural activity to fMRI. Using both RL and LR results in a more unbiased
estimate than using one phase encoding direction alone. In an ideal
world, one would like to sample from all possible phase encoding
directions using the same scanning parameters to get the most spatially
unbiased estimate, but this is non-trivial to do
Peace,
Matt.
On 8/3/16, 12:03 AM, "Chan, Adrian Ka Ho" <[email protected]> wrote:
>Hi Matt and Greg,
>
>Thank you so much for your super-fast responses. To follow up what you
>have written, is there any plan to release FIXed tfMRI data in the near
>future? If the physiological noise does lead to false positives in the
>z-stats, presumably there would be enough interest in the community to
>compare the two methods? I tried to run HCP_fix on a local machine
>overnight and the computing time does not seem to be scalable at all...
>
>Regarding what you have said about combining runs of the same encoding
>direction, is there a particular reason that explains why the LR data is
>sometimes so different from its RL counterpart, to the extent that
>sometimes REST1 LR resembles REST2 LR more than REST1 RL itself?
>
>Thank you very much again for your help!
>Adrian
>
>________________________________________
>Von: Glasser, Matthew [[email protected]]
>Gesendet: Dienstag, 2. August 2016 22:29
>An: Chan, Adrian Ka Ho
>Cc: [email protected]; Burgess, Gregory
>Betreff: Re: [HCP-Users] FIX-ICA denoised task data?
>
>I would recommend ICA+FIX for HCP task fMRI data and it is unfortunate
>that it hasn¹t already been run (however doing this is a big enough
>undertaking that I haven¹t attempted it). When carefully going through
>the task data to create the multi-modal parcellation, there were a number
>of false positive findings in the task fMRI data that were very
>reminiscent of known structured noise in the resting state fMRI. Properly
>denoising the task fMRI data could cause z-statistics to go up or to go
>down depending on the degree to which the structured noise is correlated
>with the task design. What is more important is whether the change in
>beta maps resulting from ICA+FIX denoising shows a likely neural or likely
>artifactual pattern. ICA+FIX denoising of the task fMRI data might also
>benefit from combining across runs of the same phase encoding direction
>that were acquired within the same scanning session, as the task fMRI runs
>are shorter than the resting state fMRI runs and the more time points one
>has to work with, the better ICA is able to split the temporal variance
>into components.
>
>Peace,
>
>Matt.
>
>On 8/2/16, 1:52 PM, "[email protected] on behalf of
>Burgess, Gregory" <[email protected] on behalf of
>[email protected]> wrote:
>
>>The FIX code includes a hcp_fix wrapper script, which you could run on
>>the task fMRI time series data. It should be pretty straightforward.
>>
>>However, the fact is that the 3T task data was not denoised by FIX
>>because the zstat maps generally _decrease_ in statistical significance
>>after FIX in regions where task activation is expected. This reduction
>>could be the result of removing task-correlated signal or removing
>>task-correlated noise. We have not determined that the reduced zstats are
>>due to noise, so we have elected to forego FIX on task fMRI for now.
>>
>>--Greg
>>
>>____________________________________________________________________
>>Greg Burgess, Ph.D.
>>Staff Scientist, Human Connectome Project
>>Washington University School of Medicine
>>Department of Psychiatry
>>Phone: 314-362-7864
>>Email: [email protected]
>>
>>> On Aug 2, 2016, at 1:42 PM, Chan, Adrian Ka Ho <[email protected]>
>>>wrote:
>>>
>>> Dear HCP experts,
>>>
>>> In the "900 subject" release, the rfMRI preprocessed data comes in two
>>>flavors, for example:
>>>
>>>
>>><SubjectID>/MNINonLinear/Results/rfMRI_REST1_LR_Atlas_MSMALL.dtseries.ni
>>>i
>>> and
>>>
>>><SubjectID>/MNINonLinear/Results/rfMRI_REST1_LR_Atlas_MSMALL_hp2000_clea
>>>n
>>>.dtseries.nii
>>>
>>> On the other hand, the tfMRI data seems to be released only in the
>>>first form (ie, ending with _Atlas_MSMALL.dtseries.nii). Looking at the
>>>power spectrum, one can see that most of the spikes that are cleared
>>>away by FIX-ICA in rfMRI also exist in the task data.
>>>
>>> The older posts in the mailing list suggest that the task data has not
>>>been denoised at all since the z-score results do not change much.
>>>However, instead of the common activation mapping, I would like to
>>>analyze the time series per se and hope to get rid of as much structured
>>>noise as possible. Is there a way to remove the equivalent "bad" ICA
>>>components from the task time series?
>>>
>>> Thank you very much for your help (and of course, many thanks also for
>>>sharing these really high-quality data with all of us)!
>>>
>>> Best wishes,
>>> Adrian
>>>
>>>
>>>
>>>------------------------------------------------------------------------
>>>-
>>>-----------------------
>>>
>>>------------------------------------------------------------------------
>>>-
>>>-----------------------
>>> Forschungszentrum Juelich GmbH
>>> 52425 Juelich
>>> Sitz der Gesellschaft: Juelich
>>> Eingetragen im Handelsregister des Amtsgerichts Dueren Nr. HR B 3498
>>> Vorsitzender des Aufsichtsrats: MinDir Dr. Karl Eugen Huthmacher
>>> Geschaeftsfuehrung: Prof. Dr.-Ing. Wolfgang Marquardt (Vorsitzender),
>>> Karsten Beneke (stellv. Vorsitzender), Prof. Dr.-Ing. Harald Bolt,
>>> Prof. Dr. Sebastian M. Schmidt
>>>
>>>------------------------------------------------------------------------
>>>-
>>>-----------------------
>>>
>>>------------------------------------------------------------------------
>>>-
>>>-----------------------
>>>
>>> _______________________________________________
>>> HCP-Users mailing list
>>> [email protected]
>>> http://lists.humanconnectome.org/mailman/listinfo/hcp-users
>>
>>
>>________________________________
>>The materials in this message are private and may contain Protected
>>Healthcare Information or other information of a sensitive nature. If you
>>are not the intended recipient, be advised that any unauthorized use,
>>disclosure, copying or the taking of any action in reliance on the
>>contents of this information is strictly prohibited. If you have received
>>this email in error, please immediately notify the sender via telephone
>>or return mail.
>>
>>_______________________________________________
>>HCP-Users mailing list
>>[email protected]
>>http://lists.humanconnectome.org/mailman/listinfo/hcp-users
>
>
>________________________________
>The materials in this message are private and may contain Protected
>Healthcare Information or other information of a sensitive nature. If you
>are not the intended recipient, be advised that any unauthorized use,
>disclosure, copying or the taking of any action in reliance on the
>contents of this information is strictly prohibited. If you have received
>this email in error, please immediately notify the sender via telephone
>or return mail.
________________________________
The materials in this message are private and may contain Protected Healthcare
Information or other information of a sensitive nature. If you are not the
intended recipient, be advised that any unauthorized use, disclosure, copying
or the taking of any action in reliance on the contents of this information is
strictly prohibited. If you have received this email in error, please
immediately notify the sender via telephone or return mail.
_______________________________________________
HCP-Users mailing list
[email protected]
http://lists.humanconnectome.org/mailman/listinfo/hcp-users
