Thanks, Aura.
What exact version of FreeSurfer are you using? I’m assuming it’s a development version, but from when? The results are indeed weird…
--
Juan Eugenio Iglesias
ERC Senior Research Fellow
Translational Imaging Group
University College London
Thank you for replying so quickly to my lengthy message. I correct some of my statements,
1. Matlab 8.4 was used (not 8.2)
2. When I ran the cross-sectional subfield command, I did not get exit errors as I did with the longitudinal command (kvlGetOptimizer). I got output errors and warnings of non convergence in the log file.
-I see your point. I mean the built-in hippocampal module (v10) worked when reconall was run with fs60 (hippoSf). When run independently, the hippocampal module (v21) exited with errors when using segmentHA_T1_long.sh, and
the output was off when using segmentHA_T1.sh (this is better seen in pic 3). I hope this clarifies a bit. Please tell me if it doesn’t.
These are screenshots of the images output by jobs that exited without processing errors:
rh.hippoAmygLabels-T1 and lh.hippoAmygLabels-T1 output by segmentHA_T1_long.sh
SUBJ1_timepoint1.long.SUBJ1_basetemplate
aseg and T1 output
by fs60 reconall -long
(I am now noticing these *.long.* directories contain hippoSfVolumes text files, but not hippoSfLabels mgz files...)
rh.hippoAmygLabels-T1 and lh.hippoAmygLabels-T1 output by segmentHA_T1.sh
SUBJ1_timepoint1.long.SUBJ1_basetemplate
aseg and T1 output by
fs60 reconall
rh.hippoAmygLabels-T1 output by segmentHA_T1.sh SUBJ1_timepoint1
aseg, T1 and lh.hippoSfLabels-T1.v10 output
by fs60 reconall
1. What do you mean by “loading Matlab 8.2”? It should be Matlab 8.4 (2014b).
2. “Some still did”: what error did you get?
- Also, I’m confused by what you wrote, that subfields worked but hippocampal module didn’t (aren’t they the same thing?)
- Finally: can you please send us a screenshot of one of those abruptly cut hippocampi, so I better understand what might be going on?
ERC Senior Research Fellow
Translational Imaging Group
University College London
I am attempting to use the longitudinal pipeline of the "hippocampal subfields and nuclei of amygdala”, but keep getting a flawed output.
My dataset consists of images from a longitudinal study with some dropouts. First, I ran the reconall longitudinal pipeline in fs60 for all subjects. I got no errors. Then, I tried
1. Running the longitudinal pipeline of the hippocampal module, loading Matlab 8.2 to run
segmentHA_T1_long.sh SUBJ1_basetemplate
segmentHA_T1_long.sh SUBJ2_basetemplate
I got an error in most of these jobs (mostly for templates of subjects with multiple visits) that read
"ERROR: MatlabObjectArray(0x7fd7ce84cea0): Trying to access non-existing object (25)
Error in kvlGetOptimizer (line 11)
Error in SegmentSubfieldsT1Longitudinal (line 1155)”
2. I then tried running the cross sectional pipeline for each one of the timepoints in subjects already processed by the reconall longitudinal pipeline (i. e. segmentations in the .long. directories)
segmentHA_T1.sh SUBJ1_timepoint1.long.SUBJ1_basetemplate
segmentHA_T1.sh SUBJ1_timepoint2.long.SUBJ1_basetemplate
segmentHA_T1.sh SUBJ2_timepoint1.long.SUBJ2_basetemplate
segmentHA_T1.sh SUBJ3_timepoint1.long.SUBJ3_basetemplate
Most of them ran without errors, but some still did.
3. I then tried to run the cross sectional hippocampal module for all subjects/timepoints without using the reconall longitudinal pipeline
segmentHA_T1.sh SUBJ1_timepoint1
segmentHA_T1.sh SUBJ1_timepoint2
segmentHA_T1.sh SUBJ2_timepoint1
segmentHA_T1.sh SUBJ3_timepoint1
They were processed without errors. But when inspecting the processing log, there was a message that appeared 4 times in the file
"WARNING: Registration did not converge in 5 steps!
* Problem might be ill posed.
* Please inspect output manually!
It took 766.0135 seconds “
With manual inspection, subfield segmentations output by fs60 seemed fine in all the directories of the reconall cross sectional and longitudinal outputs (hippoSfLabels-T1). For the Hippocampal Module outputs (hippoAmygLabels-T1), the hippocampus was abbruptly
cut, sometimes laterally, sometimes dorsally. This was the case for all three analyses.
I wonder what is the problem and what I can do to fix it so I can successfully run the longitudinal pipeline of the "hippocampal subfields and nuclei of amygdala”