List readers - here is a helpful clarification from Avi on my question to Nick. Thx Avi! carl
------ Forwarded Message From: Avi Snyder avi@npg.wustl.edu Reply-To: Avi Snyder avi@npg.wustl.edu Date: Fri, 24 Apr 2009 01:37:50 -0400 To: nicks@nmr.mgh.harvard.edu, "Schwartz, Carl R.E." carl_schwartz@hms.harvard.edu Cc: avi@npg.wustl.edu, zonglei.fsl@gmail.com Conversation: Question about Tal QC logs Subject: Re: Question about Tal QC logs
Carl,
tal_QC_AZS uses a different algorithm from the QA script now in FreeSurfer. Both scripts provide QA numbers but the numbers won't match.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
Nick,
Could you please help me reconcile your reply to Zhen below with Avis note below about the tal_QC_AZS script?
Below we have pasted a sample of our recon all log. Is the talAviQA in the default log the same as the eta term that Avi describes as output of the tal_QC_AZS script below in his note to Zhen? What does the z refer to in the recon all log?
It appears that the default tal check does not generate the atlas_transform_error term, correct? ( I also dont see any term less than 15).
Does one need to run Avis tal_QC_AZS script to generate the atlas_transform_error term ?
While it looks like tal check not not automatically output the value of the atlas_transform_error term, Is the z score that DOES appear (automatically) in the log file the same z score that Avi mentions below, that compares the atlas_transform_error term to the training set?
Thanks!
Carl
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
------------------------------------------------------------------------------------------------------------------------------ Mon Apr 13 15:14:55 CST 2009 talairach_avi done
cp transforms/talairach.auto.xfm transforms/talairach.xfm
#-------------------------------------------- #@# Talairach Failure Detection Mon Apr 13 15:14:56 CST 2009 /nfs/h2/r6/subjects/A39/mri
talairach_afd -T 0.005 -xfm transforms/talairach.xfm
talairach_afd: Talairach Transform: transforms/talairach.xfm OK (p=0.8116, pval=0.8469 >= threshold=0.0050)
awk -f /nfs/h1/r1/jliu/new_soft_test/freesurfer/bin/extract_talairach_avi_QA.awk /nfs/h2/r6/subjects/A39/mri/transforms/talairach_avi.log
TalAviQA: 0.97187 z-score: -1 #-------------------------------------------
On 4/22/09 4:04 PM, "Nick Schmansky" nicks@nmr.mgh.harvard.edu wrote:
Zhen,
The -tal-check flag included with recon-all will perform the QA check that Avi mentions, with the addition of creating a z-score comparing the subject's QA score to a set of QA scores of known-good talairach alignments. The -tal-check flag runs by default in recon-all (in -all and -autorecon1), and will error exit if the alignment is below threshold.
Nick
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
On Tue, 2009-04-21 at 19:02 +0800, Zhen Zonglei wrote:
Dear Avi:
Thanks for your kind and the script.I will use it to check tal QC of all my data .
Zognlei
2009/4/21 Avi Snyder avi@npg.wustl.edu Zonglei,
1. Re: how to detemine Talairach transform success from the logfile: Attached is a simple csh script (tal_QC_AZS) and a required awk script. tal_QC_AZS reads the log file and writes out two numbers. The first number (eta) is the spatial correlation between the (blurred) input image and the atlas representative target image. eta values less than 0.96 are highly suspect. The second number ("atlas_transform_error") is computed on the basis of registration objective function curvature in parameter space. The 12 present numbers are compared to values derived from a training set. atlas_transform_error values less than ~15 are good. Avi ----------------------------------------------------------------------- >Date: Tue, 21 Apr 2009 12:52:45 +0800 >Subject: Re: [Freesurfer] t4_read error (fwd) >From: Zhen Zonglei <zonglei.fsl@gmail.com> >To: Avi Snyder <avi@npg.wustl.edu> >Cc: freesurfer@nmr.mgh.harvard.edu, fischl@nmr.mgh.harvard.edu, nicks@nmr.mgh.harvard.edu > > Hi avi: > >Thanks for your help. Still have two questions. > >1) How to determine whether the Talairach transform is reasonable >based on recon-all.log?