Re: [HCP-Users] Extracting ROI data from HCP resting state data - 2400 data points instead of 1200 ?

[Glasser, Matthew]

That’s fine, I usually have time along rows for better or for worse…

Peace,

Matt.

From: David Hofmann >
Date: Thursday, July 14, 2016 at 12:14 PM
To: Matt Glasser >
Cc: Stephen Smith >, 
"Dierker, Donna" >, hcp-users 
>
Subject: Re: [HCP-Users] Extracting ROI data from HCP resting state data - 2400 
data points instead of 1200 ?

Hi Matthew,

I'm sorry I made a mistake. I meant that I calculated the column mean and 
subtracted it from each column. For example given a 3 by 2 matrix with 3 time 
points and 2 ROIs

Time | ROI 1 | ROI 2
  1 3  4
  2 2  2
  3 1  3

 mean  2  3

Substracting 2 from ROI 1 and 3 from ROI 2 gives me

Time | ROI 1 | ROI 2
  1 1  1
  2 0  -1
  3 -1 0

hope this clarifies it and this is what you meant. Sorry again for this mistake!

David


2016-07-13 22:53 GMT+02:00 Glasser, Matthew 
>:
2) Did you subtract the row mean from each row?

Peace,

Matt.

From: David Hofmann >
Date: Wednesday, July 13, 2016 at 9:56 AM

To: Matt Glasser >
Cc: Stephen Smith >, 
"Dierker, Donna" >, hcp-users 
>
Subject: Re: [HCP-Users] Extracting ROI data from HCP resting state data - 2400 
data points instead of 1200 ?

Hi Matthew,

thank you for the help!

I did the following:

1. I used the "clean" datasets now and extracted my ROI data from the LR and RL 
phase encodings (first two sessions)
2. I calculated the mean across time for every ROI and substracted it, i.e. I 
have a 1200 by 112 matrix and calculated the row mean (over time) and 
substracted it from each column (ROIs).
3. I did this for LR and RL seperately and then concatenated them together

Can you confirm that this is the correct?

thanks again!

David

2016-07-13 0:55 GMT+02:00 Glasser, Matthew 
>:
I would recommend using the data with _hp2000_clean in the name.  I a referring 
to taking the mean across time at each point in space and subtracting that from 
the data.

Peace,

Matt.

From: David Hofmann >
Date: Tuesday, July 12, 2016 at 6:56 AM
To: Matt Glasser >
Cc: Stephen Smith >, 
"Dierker, Donna" >, hcp-users 
>

Subject: Re: [HCP-Users] Extracting ROI data from HCP resting state data - 2400 
data points instead of 1200 ?

Hi all,

Regarding Stephen's answer:  I thought it is necessary to concatenate the LR/RL 
phase encoding directions together somehow or can I just treat every run 
seperately? What I basically want is the timecourse from a voxel or a region 
(from preprocessed data) which I can use for further analysis.

Regarding Matthew's answer: I'm afraid I'm not exactly sure what you mean by 
cleaning or removing the mean image from the data. Mean centering?

I used the preprocessed datasets:

subjectcode_3T_rfMRI_REST1_preproc.zip

\MNINonLinear\Results\rfMRI_REST1_LR
\MNINonLinear\Results\rfMRI_REST1_RL

Is this the correct data or is it necessary to use some different datasets for 
my specific purposes? (Normally I'd use the netmats datasets, but in I'm 
especially interested in the amygdala which I'm trying to extract from the 
Harvard-Oxford Atlas ROI).

Thanks for your answers!

David

2016-07-12 12:47 GMT+02:00 Glasser, Matthew 
>:
Also it appears you haven’t either cleaned or removed the mean image from the 
data.

Matt.

From: 
>
 on behalf of Stephen Smith >
Date: Tuesday, July 12, 2016 at 3:48 AM
To: David Hofmann >
Cc: "Dierker, Donna" >, hcp-users 
>

Subject: Re: [HCP-Users] Extracting ROI data from HCP resting state data - 2400 
data points instead of 1200 ?

Hi - no we do not (in general for resting-state) ever recommend temporal 
contatenation like this before further analyses - for the reason you're seeing 
here.
For example, for the HCP released netmats, we take the 4 runs, one at a time, 

[HCP-Users] HCP Course 2016 Update: Full Course Schedule

[Elam, Jennifer]

The full Course 
Schedule
 is now available for the 2016 HCP Course: "Exploring the Human 
Connectome".
 Attendee spots are filling up fast, 
Register today!



The 2016 HCP Course curriculum includes over 20 Lectures and 15 Practicals and 
will be presented by 19 HCP investigators and 
staff
 and and will be held over five full days: August 28-September 1 
(Sunday-Thursday) at the Joseph B. Martin Conference Center at Harvard Medical 
School, in Boston, Massachusetts, 
USA.

The course will provide hands-on experience in working with a new multi-modal 
human cortical parcellation (Glasser et al., Nature, in press) and with the 
"HCP-Style" paradigm for data acquisition, analysis, and sharing (Glasser et 
al., Nature Neuroscience, in press).

If you have any questions, please contact us at 
hcpcou...@humanconnectome.org

Hope to see you in Boston!



Best,

2016 HCP Course Staff


Jennifer Elam, Ph.D.
Scientific Outreach, Human Connectome Project
Washington University School of Medicine
Department of Neuroscience, Box 8108
660 South Euclid Avenue
St. Louis, MO 63110
314-362-9387
e...@wustl.edu
www.humanconnectome.org


___
HCP-Users mailing list
HCP-Users@humanconnectome.org
http://lists.humanconnectome.org/mailman/listinfo/hcp-users

Re: [HCP-Users] Exact Meaning of ICA maps in 800 subjects release

[Stephen Smith]

Hi  [comment also for Jenn below]

Sorry, we didn't calculate z-stat versions of the volumetric group-average 
maps.  The volumetric maps are really just intended as a useful visual 
reference - the CIFTI versions are the "real thing".

-

BTW - I think that you are not asking about subject-specific RSN maps - but for 
completeness: just the CIFTI z-stat versions of those are available online at 
HCP website.
Jenn - I've just noticed that the descriptions of these 3 recent packages are 
totally wrong - that's probably my fault sorry - please could you change all of 
the following text from:

The following links contain volumetric NIFTI versions of the CIFTI MSMall 
Group-ICA parcellations, with various ICA dimensionalities applied.

 Volumetric Parcellations for 10-, 25-, 50-, 100-dimensionalities (57GB) 

 Volumetric Parcellations for 200-dimensionality (55GB) 

 Volumetric Parcellations for 300-dimensionality (85GB) 

to being:

The following links contain subject-specific CIFTI maps: subject-specific 
versions of the group-ICA parcellations, with various ICA dimensionalities 
applied. These are z-statistic maps generated using dual-regression.

 CIFTI subject-specific Parcellations for 10-, 25-, 50-, 100-dimensionalities 
(57GB) 

 CIFTI subject-specific Parcellations for 200-dimensionality (55GB) 

 CIFTI subject-specific Parcellations for 300-dimensionality (85GB) 


Thanks!





> On 14 Jul 2016, at 06:49, Nicola Toschi  wrote:
> 
> Dear Prof. Smith,
> Dear List,
> 
> one more question: would the Z-maps related to the second-last step step 
> (dual regression to create volumetric representations if the IC) be available 
> (they are not in the public release)?
> 
> This would greatly aid in interpreting/threshold the volumetric betas.
> 
> Thanks a lot in advance,
> 
> Nicola
> 
> On 07/06/2016 10:09 AM, Stephen Smith wrote:
>> Hi
>> 
>> Group-ICA is carried out in grayordinate space. Then:
>> 
>> For each subject, subject-wise grayordinate AND volumetric maps (versions of 
>> the group-ICA) are estimated using dual-regression, with node-timeseries 
>> normalisation, and outputs in units of GLM betas (parameter estimates) not 
>> Z.   
>> 
>> These volumetric maps are then averaged across subjects to generate the 
>> group-average volumetric maps distributed from HCP.
>> 
>> So how you threshold these is up to you and depends what the intended usage 
>> for that is...
>> 
>> Cheers.
>> 
>> 
>> 
>>> On 6 Jul 2016, at 07:26, Nicola Toschi < 
>>> tos...@med.uniroma2.it 
>>> > wrote:
>>> 
>>> Hi, 
>>> 
>>> The volumetric version!
>>> 
>>> Thanks in advance, 
>>> 
>>> Nicola
>>> 
>>> 
>>>  Original message 
>>> From: Stephen Smith > 
>>> Date: 04/07/2016 12:37 PM (GMT+01:00) 
>>> To: Nicola Toschi > 
>>> Cc: hcp-users >> > 
>>> Subject: Re: [HCP-Users] Exact Meaning of ICA maps in 800 subjects release 
>>> 
>>> HI - do you mean the grayordinate or volumetric versions?  
>>> Cheers
>>> 
>>> 
>>> 
 On 4 Jul 2016, at 11:16, Nicola Toschi > wrote:
 
 Hi list, 
 
 This may be more of a MELODIC question, but I would be grateful for any 
 input:
 
 In the ICA release on 800+ subjects, what is the exact mathematical 
 meaning / definition of the (non-binary) intensities of the different 
 component volumes?
 
 Are they easily relatable to a statistical map (I am looking to threshold 
 them to study main clusters and possibly correct for multiple comparisons)?
 
 Thanks in advance, 
 
 Nicola
 
 ___
 HCP-Users mailing list
 HCP-Users@humanconnectome.org 
 http://lists.humanconnectome.org/mailman/listinfo/hcp-users 
 
>>> 
>>>