Dear Freesurfer Experts,

Please excuse my long email yesterday.

Now, I finished code update of the outlier detection part for brainstem substructures (bs) and Hippocampal Subfields (hc) modules. Code updated in the similar way as my last email except that I combined the output of both modules to simplify the process. The following log files might show some clue of my exploration:

YongbshcLUT (like asegLUT_112511):

Medulla 0.287988 +/- 0.0304379
Pons 0.864817 +/- 0.128019
SCP 0.0161611 +/- 0.0030778
Midbrain 0.388938 +/- 0.038842
Whole_brainstem 1.8246 +/- 0.184472
Left-Hippocampal_tail 0.0291617 +/- 0.00437039
Left-subiculum 0.0237801 +/- 0.0044348
Left-CA1 0.0387709 +/- 0.00612168
Left-hippocampal-fissure 0.00944849 +/- 0.00188347
Left-presubiculum 0.0177329 +/- 0.00272706
Left-parasubiculum 0.00372024 +/- 0.000876862
Left-molecular_layer_HP 0.0320802 +/- 0.00806223
Left-GC-ML-DG 0.0186971 +/- 0.00308382
Left-CA3 0.0120106 +/- 0.00198279
Left-CA4 0.0132618 +/- 0.00261687
Left-fimbria 0.00371701 +/- 0.00166146
Left-HATA 0.0031189 +/- 0.000692806
Left-Whole_hippocampus 0.190022 +/- 0.0246413
Right-Hippocampal_tail 0.030817 +/- 0.00496078
Right-subiculum 0.0248689 +/- 0.00427691
Right-CA1 0.0370719 +/- 0.00873442
Right-hippocampal-fissure 0.0103329 +/- 0.00187117
Right-presubiculum 0.0178487 +/- 0.00260947
Right-parasubiculum 0.00380418 +/- 0.000861778
Right-molecular_layer_HP 0.032938 +/- 0.00448073
Right-GC-ML-DG 0.0168729 +/- 0.00304788
Right-CA3 0.0130893 +/- 0.00239818
Right-CA4 0.0143492 +/- 0.00281808
Right-fimbria 0.00438099 +/- 0.00176098
Right-HATA 0.00338842 +/- 0.0006037
Right-Whole_hippocampus 0.198797 +/- 0.0243668

an example subject's bshc.stats (like aseg.stats):

Medulla 4288.386728
Pons 12836.783824
SCP 194.446727
Midbrain 8306.988464
Whole_brainstem 22293.878740
Left-Hippocampal_tail 462.086606
Left-subiculum 307.970217
Left-CA1 816.878818
Left-hippocampal-fissure 118.986208
Left-presubiculum 227.392060
Left-parasubiculum 84.381831
Left-molecular_layer_HP 841.496111
Left-GC-ML-DG 238.308049
Left-CA3 183.383724
Left-CA4 196.098618
Left-fimbria 73.990998
Left-HATA 36.237082
Left-Whole_hippocampus 2838.187477
Right-Hippocampal_tail 408.832729
Right-subiculum 388.667888
Right-CA1 801.601831
Right-hippocampal-fissure 132.071091
Right-presubiculum 242.179137
Right-parasubiculum 43.414171
Right-molecular_layer_HP 487.298072
Right-GC-ML-DG 246.283617
Right-CA3 164.849646
Right-CA4 209.108368
Right-fimbria 90.291400
Right-HATA 42.073898
Right-Whole_hippocampus 2791.300419

icv.txt (actually eITV instead of ICV):

1420463.178113

bshc_outliers.log:

Subject 008 Right-HATA value .002080900 is an outlier ...

Note: all data here are kind of simulation results because of data safety reason and these logs only serve as signs of correct code update.

I guess that both the only output of bs and hc (bshc.stats) is segment volume, not the 'normMean normStdDev ...' in aseg.stats.
So is it senseful to include those 'norm*' measures in the bs & hc modules in the future?
BTW, is there an advantage to use normMax / eITV rather than normMean / eITV in outlier detection in theory?

Thank you very much again!

I wish all of you a nice weekend!

Best regards,

Yong
_________________________
Yong Li, PhD
Department of Neurology,
Klinikum rechts der Isar,
Technische Universität München,
Munich, Germany

On Thu, Apr 6, 2017 at 11:36 AM, Yong Li <yongli86@gmail.com> wrote:

Dear Freesurfer Experts,

As one of the new freesurfers, I really appreciate the excellent work of authors as well as great support of users' forum.

A good guess of mine is that authors, e.g., Vasanth or Louis etc., are busy updating QA tools after the FSv6 release since the QA Tools v1.2 was written for FSv5.3. If so, is it possible to provide an estimated release date please so as to adjust my plan? Thanks for understanding!

In case, it might take a while to release the next QA Tool, may I dare to provide my attempts of code updating of some functions in QA Tools? The following are the brief list of code adapting steps:
1. modified codes for generating file order list in recon_checker;
2. modified codes for matching segment names for instance from DefaultAsegMeans.txt in gparcmean_from_table;
3. adapted default_FOF and default_status_log for FSv6.

After these adaptation, QA Tools ran through and I only encountered some 'syntax error' in terminal for 2 segments, 5th-Ventricle and WM-hypointensities, that I have to ask the experts to guide me through this point if it's of your interests. To ensure the QA Tools functioning well, I tested 116 elderly patients and detected reasonable outliers per patient. Therefore, I would like to double that my code updating based on my general understanding of QA Tool still fit to the basic purpose of authors by asking the following theoretical questions:

1. SNR and segments outliers detection are measures of quality assessment in QA Tools. Have other factors such as subject's motion, image bias, resolution etc. been considered in QA Tools or future release? The default SNR threshold is 16 based on any reference? The max SNR of my patients so far is 32. Any idea of scores of healthy elderly controls?

2. The QA Tools v1.2 analyzes aseg.stats output. Is it possible to include output of brainstem substructures and HippocampalSubfields of FSv6? If yes, I would like to adapt the codes and provide you some feedback if you think that I can help bit.

3. Future directions mentioned in QA Tools wiki site are interesting and thoughtful. Are other features like 'Detect outliers in aparc and aparc2009 volumes' on the way?

4. Is there a function to summarize all QA measures to csv files, like the quantifyBrainstemStructures.sh in brainstem substructures?

Thank you in advance!

Kind regards,

Yong Li, PhD
_________________________
Department of Neurology,
Klinikum rechts der Isar,
Technische Universität München,
Munich, Germany