RapidArt

[Gideon]

Hi,


I want to use ArtifactDetect from "algorithms.rapidart" for localizers, rest and task scans artifact detection .

I was wondering whether there are any guidelines or best practices available for using it?
What are the scenarios which you should and should not use it?
Moreover, are there any limitations on using it for functional connectivity analysis?


Thank you,

Gideon
 

[Satrajit Ghosh]

hi gideon,

I want to use ArtifactDetect from "algorithms.rapidart" for localizers, rest and task scans artifact detection .

I was wondering whether there are any guidelines or best practices available for using it?

in general we use a z threshold of 2 or 3 and a composite norm threshold of 0.5 mm with motion using differences and intensity not using differences.

 

What are the scenarios which you should and should not use it?

you can use it for any functional task. using this algorithm by itself does not change anything in your analysis, simply gives you a quantitative observation of outliers. however, where it becomes useful is adding those outliers (one per column) to your design matrix or nuisance filtering. in nipype we generally do this using SpecifyModelgen and its variants.

 

Moreover, are there any limitations on using it for functional connectivity analysis?

you want to be careful that the outliers are not biased towards any specific condition. essentially by checking the correlation between a single vector containing your outliers and your conditions of interest.

this will in effect create different designs for each subject as they will have different numbers of outliers, but our empirical evidence shows that it improves characteristics of residuals in task analysis.

cheers,

satra

[Gideon]

Thank you very much for the response. I will give it a try.

Gideon

בתאריך יום שישי, 24 במאי 2013 17:04:01 UTC+3, מאת satra: