Dear Conn expert, 

I am runing minimum processed HCP data using conn doing ART and Denosing. Denosing I think the data goes well, while ART seems quite strange. Attached picture is the snapshot of the ART. 

I wonder that why the HCP data demonstrated so many outliers. Does any mistakes happen of my setting? If possible, could you help me figure it out? 

In addition, for the denosing step, I put the movementfile including 6 head motion and 6 first-derived motion paremeters. I checked the manual of conn, and found that Conn also do the same thing for movement regressors. I wonder that whether I only need input first 6 movement parameters and then conn will automically derive these first-derived, or all for denosing and conn ignore estimated first-derived?  

Thank you very much for your suggestions in advance. 
With my kind regards, 
Ray

Dear Ray,

Regarding HCP data, the file conn_batch_humanconnectomeproject.m offers an example on how to import and process HCP data. The issue that you are finding is likely due to HCP subject-movement files being saved in a somewhat unconventional format (angular components are specified in degrees, as opposed to radians). If you check the conn_batch_hummanconnectomeproject file you will see there that one way to deal with this is simply to rename or copy the movement#.txt files to movement#.deg.txt. That will let CONN know the specific format of this file. Just for reference, CONN can import the following subject-movement formats (and it discerns among them based on the file extension):

   .txt files (SPM format: three translation (x/y/z in mm), three rotations (x/y/z in radians) )
   .par files (FSL format: three rotations (x/y/z in radians), three translations (x/y/z in mm) )
   .deg.txt (HCP format: three translations (x/y/z in mm), three rotations (x/y/z in degrees) )
   .siemens.txt (Siemens format: three translations (y/x/-z in mm), three rotations (x/y/z in degrees) )

Regarding your second question, yes, typically you can just enter into CONN these 6-parameter motion files, and then during Denoising decide whether you want to automatically add first-order derivatives and/or higher-order powers to theses timeseries (e.g. for Friston24 regressors)

Hope this helps
Alfonso
Originally posted by Ruibin Zhang:

Dear Alfonso, 

Many thanks for your detailed addressment. I will follow instructions to do it. 

Here, I still one more uncertain area: 

Denosing: The default mode setting of script of conn_batch in confounds.derive is 0, that is only for raw timeseries. While the manual v17 referred that 'CONN preprocessing steps will be automatically set up to use a combination of aCompCor (White and CSF ROIs, 5 components each), scrubbing (as many regressors as identified invalid scans), motion regression (12 regressors: 6 motion parameters + 6 first-order temporal derivatives), and filtering in the Denoising step.' 

Thus, I just wonder that there are contradict areas if I only enter 6-parameter motion files, does conn realy did the first derived or still raw? 

Thanks again for your great patient. 
Ray



Originally posted by Alfonso Nieto-Castanon:

Dear Ray,

You are right about your observation, the conn_batch help description of the batch.Denoising.confounds.derive field default value ([0]) is a bit incomplete. In reality the default is [0] (raw timeseres) for all confounding effects except for realignment and task effects (for those the default is [1] -raw timeseries+first-order derivative-; to be precise any first-level covariate with a name that starts with 'effect of|realign|motion|movement' will use a default batch.Denoising.confounds.derive of [1]). Of course these default settings only apply if you do NOT specify a batch.Denoising.confounds.derive field in your batch script, otherwise the specified values will always take precedence.

In any way, if in doubt you can simply load your project in CONN's gui, go to the Denoising tab, and click there on each element in the 'confounds' list to check whether it does in fact include first-order derivative terms or not (as well as check there all of the other Denoising settings defined for your project, and/or modify those and re-run the Denoising step if you wish). 

Hope this helps
Alfonso

Originally posted by Ruibin Zhang: