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:
- modified codes for generating file order list in recon_checker;
- 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:
- 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?
- 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.
- 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?
- 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