Questions about preprocessing: realignment and detrending

Submitted by eligva on Fri, 02/15/2013 - 00:01

 Hello!

I’m starting to use the DPARSF toolbox to preprocess my rs-fMRI data and I have a few questions regarding the preprocessing steps implemented in the toolbox. I use the advanced version because I want to do only the following steps:
1) Convert my Dicom – 2)Remove the first volumes 3)Realign and create a mean volume – 4)Segment – 5)Regress the nuisance covariates – 6)Smooth - 7)Detrend - 8)Filter

I have a few questions about realignment and detrending:
    - What are the default settings of the realignment: Does it realign all images to the first/middle/mean image? Does-it create a mean image?
    - Is there any way to modify them in your batch?
    - There is no detrending check-box in the advanced version (or did I miss it?). Is there any reason for that?
    - Should I switch then to the basic window after the smoothing in order to detrend and filter my data?

 Thank you in advance for your answers,

Best regards
Elisa

Submitted by eligva on Fri, 02/15/2013 - 00:08

 Hi!

I'm sorry I don't know what happend to my post!
here it is in a more "readable" way!

I'm starting to use the DPARSF toolbox to preprocess my rs-fMRI data and I have a few questions regarding the preprocessing steps. 
I use the advanced version because I juste want to run the following steps:
1) convert Dicom, 2) remove first volumes 3) realign and obtain a mean image 4) segment 5) regress the nuisance covariates 6) smooth, 7) detrend, 8) filter

I have a few questions regarding realignment and detrending:
- What are the default settings for the realignment: are the images realigned to the first/middle/mean image? does it create a mean image automatically? is there any way to change the settings in your batch?
- there is no detrending check-box in the advanced version of the toolbox (unless I missed it). Is there any reason for that? If I want to detrend and filter my data at the end of my preprocessing should I switch to the basic version?

Thank you again for your answers, 
Best
Elisa

Submitted by YAN Chao-Gan on Fri, 02/15/2013 - 01:47

Hi Elisa,

1. Where will you put spatial normalization? At very last step on the measure images? In that case, you can choose "calculate in original space (warp by DARTEL)".  But tick off slice timing if you want to skip this step, and tick on smooth if you want it be performed before filtering.

2. The default settings for realigment is: "Then, the time series of images for each subject were realigned using a six-parameter (rigid body) linear transformation with a two-pass procedure (registered to the first image and then registered to the mean of the images after the first realignment)." You will get the mean image, and it will be stored under RealignmentParameters directory for future use. You can modify the batch via modify the job mat files called Realign.mat under Jobmats of DPARSF directory.

3. In the advanced edition, there is "polynomial trend" option for nuisance regression. "1" means linear regressor, it should have the same effect as detrend. Nevertheless, there is still a "detrend" button before "ALFF+fALFF" and "Filter" that you can make use of.

Best,
Chao-Gan


Submitted by eligva on Sat, 02/16/2013 - 18:03

 Hello Chao-Gan, 

thank you for answering me so fast!

1. I will not normalize because I will then push my preprocessed data into another software that allow functional-structual integration using a surface-based atlas. So it is important for me to stay into the subject space. Therfore I'm using the "segment" checkbox.
What does warp by DARTEL mean? isn't it a kind of normalization? I looked for some explanations about the "New Segment + DARTEL" but I'm not sure to understand what it is. Will I still stay into the subject's space if I use it?

2-3. thank you for your answers, it is clear for me now!

I have a last problem. when I use the checkbox "segment" the masks present in the "Masks" folder miss the frontal lobe. 
and if I use the "New segment and DARTEL" I run into this error:
" Mask does not match.

Mask size is 128x128x38, not same with required size 224x256x192 "
Do you have any idea of the way I could fix this problem?

thank you in advance for your answers, 

best regards, 
Elisa

Submitted by YAN Chao-Gan on Wed, 02/20/2013 - 10:25

Hi Elisa,

1. New Segment + DARTEL means "New Segment" module in SPM8. Also perform DARTEL to generate the information needed for spatial normalization (create group template as well). You will stay at origininal space.

2. Could you have a check on the coregistration effects between functional images and T1 images? And see how well the spatial normalization effects on T1 images (w* under T1ImgSegment).

For "New Segment + DARTEL", could you provide more details, e.g., the program line for the error.

By the way, what's the resolution for your functional images? 224x256x192?

Best,

Chao-Gan 


Submitted by eligva on Sun, 05/05/2013 - 18:07

Dear Chao-Gan,

sorry for the delay in my answer but the error never happend again, except today!

In case this could help someone, I think that this error happens when the preprocessing is not started from the beginning but from a stage where the starting directory referes to a structural one (for example T1ImgCoreg). In that case, the program seems to warp the mask to the resolution of the structural image instead of the functional one.
Indeed both resolutions in the message error correspond to the resolution of my structural and functional images and it works when relaunching the program from the "regressing out covariates stages" refering to the starting directory "FunImgR".

thanks a lot for your help and for this excellent tool for preprocessing!

Best
Elisa