I have just plotted the timeseries and it seems that my results doesn't
really make sense - they look like square oscillations. HCP's results look
like real timeseries.

On Fri, Jul 31, 2015 at 10:48 AM, Yizhou Ma <[email protected]> wrote:

> Thank you Michael. My system is failing me at this moment and it is hard
> to plot. My dual reg results seem normal and I would not have suspected if
> I did not compare them with HCP's results.
> When I used a brainmask to mask the first 91282 "voxels" I get identical
> results as when I did not use a mask.
>
> Thanks,
> Cherry
>
>
>
> On Fri, Jul 31, 2015 at 10:25 AM, Harms, Michael <[email protected]> wrote:
>
>>
>> Yes, "_ts2" are dual regression; "_ts3" are the eigentimeseries.  See the
>> section "Description of released files" in the pdf with the PTN
>> distribution.
>>
>> Steve can comment on the matching issue, but I wonder if the 0's that get
>> added to the pseudo-NIFTI file to fill out the matrix dimensions (following
>> conversion from CIFTI) are contributing here.  You might need to use a mask
>> that defines the actual 91282 grayordinates.  How different actually are
>> your dual regression results vs. the "_ts2" node timeseries in the PTN
>> distribution?  Perhaps you could create a scatterplot to illustrate the
>> difference?
>>
>> cheers,
>> -MH
>>
>> --
>> Michael Harms, Ph.D.
>> -----------------------------------------------------------
>> Conte Center for the Neuroscience of Mental Disorders
>> Washington University School of Medicine
>> Department of Psychiatry, Box 8134
>> 660 South Euclid Ave. Tel: 314-747-6173
>> St. Louis, MO  63110 Email: [email protected]
>>
>> From: Yizhou Ma <[email protected]>
>> Date: Friday, July 31, 2015 10:16 AM
>> To: Giles Colclough <[email protected]>
>> Cc: "[email protected]" <[email protected]>
>> Subject: Re: [HCP-Users] failure to replicate dual regression 1 results
>> (PTN nodetimeseries)
>>
>> Thank you Steve and Giles.
>>
>> Giles: I believe the pdf specified that dual regression 1 results are in
>> /ts2 and eigentimeseries results are in /ts3, isn't it?
>>
>> Steve: I am confused here. I am using nifti data converted from cifti,
>> which means that the first 91282 "voxels" of my nifti matches the 91282
>> grayordinates in the cifti files. The two methods should be identical
>> because they have exactly the same info, only different shapes of matrices.
>> Isn't wb_command -cifti-convert -to-nifti the recommended way to use FSL
>> commands on cifti data and isn't it supposed to give identical results as
>> working with cifti directly?
>>
>> Best,
>> Cherry
>>
>> On Fri, Jul 31, 2015 at 10:04 AM, Giles Colclough <
>> [email protected]> wrote:
>>
>>> Hi, Cherry,
>>>
>>> Have a look at the other nodetimeseries data.
>>> The ts2 time series are estimated using the eigen-timeseries method. ts1
>>> are calculated using traditional dual regression.
>>>
>>> A fuller explanation is in the pdf available in the netmats download:
>>>
>>>
>>> Node timeseries (individual subjects)
>>>
>>> For a given “parcellation” (group-ICA decomposition), the set of ICA
>>> spatial maps was mapped onto each subject's rfMRI timeseries data to derive
>>> one representative timeseries per ICA component (for these purposes we
>>> consider each ICA component as a network "node"). For each subject, these
>>> 25 (or 50, 100, 200 or 300) timeseries can then be used in network
>>> analyses, as described below. Two distinct methods were used to estimate
>>> the node-timeseries:
>>>
>>>    1.
>>>
>>>    The more traditional "dual-regression stage-1" approach, in which
>>>    the full set of ICA maps was used as spatial regressors against the full
>>>    data, estimating one timeseries for each ICA map [Filippini 2009].
>>>    2.
>>>
>>>    A newer approach based on estimating the principal eigen-timeseries
>>>    within each ICA component; this approach aims to be more robust against
>>>    component misalignment (between the group-ICA maps and individual 
>>> subjects'
>>>    data) and artifacts [O’Reilly 2009, Smith OHBM 2014].
>>>
>>>
>>>
>>>    Best,
>>>    Giles
>>>
>>>
>>>
>>> On 31 Jul 2015, at 15:50, Yizhou Ma <[email protected]> wrote:
>>>
>>> Dear HCP experts,
>>>
>>> I am trying to run my own dual regression on HCP's rsfMRI data. What I
>>> did is use wb_command -cifti-convert -to-nifti to convert group ICA maps
>>> and individual rsfMRI scans to NifTI and use FSL's dual_regression command.
>>>
>>> *dual_regression melodic_IC.nii.gz 1 -1 0 ./test 100307_rsfMRI.nii.gz*
>>>
>>> I was able to get dual regression results. Yet the dual regression 1
>>> results I get does not match up with the nodetimeseries ts2 results
>>> provided in HCP's PTN release. I have opened my group ICA maps and rsfMRI
>>> scan in matlab to ensure that the values in them match up with the cifti
>>> files. I have tried dual regression with another subject but still the
>>> results did not match.
>>>
>>> Do anyone have an idea what might have gone wrong?
>>>
>>> Thank you very much,
>>> Cherry
>>>
>>> _______________________________________________
>>> HCP-Users mailing list
>>> [email protected]
>>> http://lists.humanconnectome.org/mailman/listinfo/hcp-users
>>>
>>>
>>>
>> _______________________________________________
>> 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

Reply via email to