Dear All,

I am just a beginner of AFNI.
Recently, I read the below board about rmsold & rmsnew.
I think that it is good to decide whether the 3dvolreg works well or not depending on the rmsold & rmsnew values. And if it do not work well, I can discard the whole or some part of mri signals for my final analysis.

My questions are;

Q1. What is the usual value of rmsnew if motion correction worked well?
- I think it may be arbitary, however, there may be some consensual values such as 25, 30, 40 ???

Q2. Is it rational to censor the data of high rmsnew during the deconvolution ?


Sukkyoon AN
--------------------------------------
- the below -
Posted by Michael S. Beauchamp on May 17, 2002 at 07:08:15:

In Reply to: 3dvolreg output rmsnew question posted by Vinod Rao on May 16, 2002 at 14:55:33:

Dear Vinod,
3dvolreg attempets to correct for head motion by translating and rotating each volume to match the base (canonical) volume.
The way it settles on the correct amount of transformation and rotation is by minimizing the root mean square (rms) difference between each volume and the base volume--i.e. the rms difference at every voxel.
Therefore, the "rmsold" is the rms difference (averaged across all voxels) between the unregistered volume and the base volume. If this is high, it means that there is a big amount of translation/rotation required (or something else bad happened to that volume).
The "rmsnew" is the rms difference AFTER alignment (transformation and rotation to minimize rmsnew). This will always be smaller than rmsold, and if motion correction worked, rmsold will be high and rmsnew will be small.
If something else bad happened, motion correction will not help much and both rmsold and rmsnew will be large.
----------------------------------------------------------