Freesurfer February 2021

freesurfer@nmr.mgh.harvard.edu

98 participants
118 discussions

Postdoctoral fellowship opportunity in diffusion MRI and functional vision
by Corinna Bauer 01 Feb '21

01 Feb '21

External Email - Use Caution *Thalamocortical connectivity and multicomponent T2 relaxometry in youths with neonatal periventricular white matter injury* NIH-funded Postdoctoral fellowship at the Schepens Eye Research Institute of Massachusetts Eye and Ear, Harvard Medical School in Boston, MA We are seeking a full-time postdoctoral fellow to work with Dr. Corinna Bauer on an NIH-funded project investigating the impact of neonatal thalamic injury on the development of vision-processing networks in the brain and corresponding visuo-perceptual dysfunctions. The project employs structural, diffusion, and relaxometry based MRI methods to target thalamo-cortical changes in morphometry, neural networks, and MRI-based markers of myelin. The position is available beginning Spring/Summer 2021. The initial time frame will be for 18 months with the possibility of extension. *Required Qualifications:* • Doctoral degree required with 0-1 years of related work experience • Strong coding skills in python, MATLAB, bash, etc. • Advanced skills in the processing and analysis of diffusion (including connectivity, DKI, and NODDI approaches) data, particularly focusing on network-based approaches • Strong publication record in neuroimaging, with a preference for a focus on multimodal techniques in neurodevelopmental disorders • Good command of the oral and written English language • Strong organizational and data management skills • Prior experience reviewing, analyzing, and summarizing scientific literature • Excellent attention to detail and interpersonal, organizational, writing, and project management skills *Preferred Qualifications:* • Experience with multicomponent T2 relaxometry • Experience working with clinical pediatric populations • Experience with psychophysical assessments of functional vision • Experience with fMRI design and analysis Interested candidates should send a cover letter including, a brief summary of their research interests and career goals, their C.V., and the names and phone numbers of three references to Corinna Bauer ( Corinna_Bauer(a)meei.harvard.edu) Review of applications will begin immediately and continue until the position is filled. Additional details can be found at: https://secure-web.cisco.com/1aWE_5wlqXvxDWmzy4gy3dtCobaxYZF4YKpgZiy1P01DGE… *Massachusetts Eye and Ear is proud to be an equal-opportunity employer and is committed to providing a workplace free from harassment or discrimination. All employment decisions are made without regard to race, color, age, gender, gender identification, sexual orientation, religion, marital status, sex, pregnancy or conditions related to pregnancy, national origin/ancestry, citizenship, disability, military status, genetic information, or any other basis prohibited by law. These protections extend to all management practices and decisions, including recruitment and hiring practices, appraisal systems, promotions, training, and career development programs*

1 0

White matter Parcellation
by Jose Graterol 01 Feb '21

01 Feb '21

External Email - Use Caution Dear FreeSurfer experts, I need help trying to understand how the white matter gets parcellated. I have a clinical dataset, which means that the resolution is not as research images, but it ran without a problem. I would like to measure the volume of the basal ganglia and looking at wmparc.mgz I am not sure how Freesurfer decides where a structure ends and the other one begins. I am attaching two images to show what I mean. I would really appreciate some clarification. I am using FreeSurfer 7.1.1 Thanks in advance. José [image: whiteparc2.png] [image: whiteparc1.png]

1 0

(no subject)
by Rahul Sharma 01 Feb '21

01 Feb '21

External Email - Use Caution Hi, Hope all are safe and well. Can anyone help me in finding any biomarker for Alzheimer's Disease or any guidance related to that? Waiting for response. Regards: Rahul Sharma Research Scholar 9827834360 *"Manners Maketh Man"*

1 0

recon all
by 郑凤莲 01 Feb '21

01 Feb '21

External Email - Use Caution Hi experts, We have segmented hippocampus using Freesurfer 6.0. I have a question that is there "head motion correct" in this step? If yes, how does it examined? I am looking forward to your reply. Thanks very much. Sincerely, Zheng | | 郑凤莲 | | 邮箱：zhengfenglian0105(a)163.com | 签名由网易邮箱大师定制

2 3

the T2.mgz in subject/mri should be skull stripped or not?
by shan chen 01 Feb '21

01 Feb '21

External Email - Use Caution Hi all, I encountered a problem about the T2.mgz in subject/mri. I ran my data using recon-all -all -i T1path -s name -T2 T2path -T2pial(7.1.1 or 7.1.0), but the results were inconsistent, sometimes the mri/T2.mgz had skull sometimes it didn't. That is weird. I read the recon-all.log T2.mgz with skull, the code is: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% writing normalized volume to /xx/subject/mri/T2.norm.mgz #VMPC# mri_normalize VmPeak 978796 mri_normalize done @#@FSTIME 2021:01:31:21:55:47 mri_normalize N 18 e 42.55 S 2.36 U 40.16 P 99% M 723884 F 0 R 925502 W 0 c 78 w 4 I 0 O 78480 L 62.01 61.65 61.48 @#@FSLOADPOST 2021:01:31:21:56:29 mri_normalize N 18 61.84 61.65 61.49 mri_mask -transfer 255 -keep_mask_deletion_edits /xx/subject/mri/T2.norm.mgz /xx/subject/mri/brain.finalsurfs.mgz /xx/subject/mri/T2.mgz transfer mask voxels=255 to dst vol Transferring mask edits ('1' voxels) to dst vol DoAbs = 0 MRImask(): AllowDiffGeom = 1 MRImask(): AllowDiffGeom = 1 Writing masked volume to /xx/subject/mri/T2.mgz...done. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% when it comes to no skull, the code is: writing normalized volume to /xx/subject/mri/T2.norm.mgz mri_mask /xx/subject/mri/T2.norm.mgz /xx/subject/mri/brainmask.mgz /xx/subject/mri/T2.mgz DoAbs = 0 Writing masked volume to /xx/subject/mri/T2.mgz...done. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% it seems that the pial is not refined by the T2 image if the T2.mgz is with-skull. anybody know what's going on here? Thanks, Yijun.

1 0

report results
by Marina Fernández 01 Feb '21

01 Feb '21

External Email - Use Caution Dear experts, In addition to t and p-value, I would like to report the beta coefficient of the fixed effects in the following regression equation for the vertex showing the max statistic within the cluster that survived multiple comparison correction: y = β0 + β1*sex + β2*age + β3*(sex+age) + e I know that beta.mgh contains the beta estimates corresponding to the intercepts of male and female and the slopes of age for each sex. However, what I need is the beta corresponding to the intercept (β0), and the beta for the slope of sex (β1), age (β2) and sex*age interaction (β3). I also need the residual variance of the model (e), but rvar.mgh contains 4 values, 2 for the intercepts and another two for the slopes. I would really appreciate if you could help me to obtain this information. Thanks a lot in advance Marina

1 0

FW: Re: SAMSEG questions
by Stefano Cerri 01 Feb '21

01 Feb '21

External Email - Use Caution Forwarding Koen's and my answers ________________________________ From: Koen Van Leemput <leempko(a)gmail.com> Sent: 01 February 2021 10:22 To: Stefano Cerri Cc: David Kamson Subject: Re: SAMSEG questions Hi Stefano and David, Regarding GPU/tensorflow issues, Andrew recommended running fspython -c 'import tensorflow' to diagnose what the problem is. Koen On Mon, Feb 1, 2021 at 3:04 AM Stefano Cerri <stce(a)dtu.dk<mailto:stce@dtu.dk>> wrote: Hi David, 1a) It depends on what analysis you are doing, but in general I would correct with SBTIV the longitudinal analysis as well. 1b) I assume you are computing volumes on the seg.mgz file with 3D slicer. The samseg.stats file is computed from "soft segmentations", i.e. probabilistic maps of each individual structure. These maps can be obtained by adding the flag --save-posteriors. 1c) This file is a scaling factor to obtain a certain average intensity in a target structure. It is used later on in the FreeSurfer pipeline for creating the wm mask and surfaces. It is unrelated to volume analysis/correction. 2) On this page (http://secure-web.cisco.com/1VcMQfGu6N_o0lPCGE8rKcisZZBWi7SnvF0e_eGMHnEZteu…) there is a link for downloading the development version. Installation is then similar to Freesurfer 7.1, you just need to change the name of the downloaded file. Remember that it is essential to process all your subjects with the same version of FreeSurfer. 3) If you have both Tensorflow versions, the latest installed version is used. If you have the GPU version installed and the CPU is used, it means that Tensorflow has some problems locating your cuda/cudnn libraries. Unfortunately, I can't help much here. The specific cuda/cudnn libraries depend on your specific OS and GPU graphic card. Also note that the GPU is used only in the last part of the algorithm, during the sampling phase. Hope this helps. Best, Stefano ________________________________ From: David Kamson <dkamson1(a)jhmi.edu<mailto:dkamson1@jhmi.edu>> Sent: 30 January 2021 20:45 To: Stefano Cerri; freesurfer(a)nmr.mgh.harvard.edu<mailto:freesurfer@nmr.mgh.harvard.edu> Cc: leempko(a)gmail.com<mailto:leempko@gmail.com> Subject: SAMSEG questions Hey Stefano, Thanks for the quick responses! Your arxiv.org<http://secure-web.cisco.com/1fCw1M5LHKguprGZaKGYWYPrl-nXSWhkOsGqvsvmcZq9aDp…> paper is very helpful! I was wondering if you could help me with some of the follow up questions: 1) SAMSEG long should compute more consistent longitudinal structural volumes than SAMSEG while detecting better volumetric group differences (cf. Table 1 and Fig 3 of https://secure-web.cisco.com/1GQa4CGwElUY2mW4Sq82w19at83-bnSN-rjbzKNFRS6VfQ…) Yes, using SBTIV for correcting volume changes from different scanners would be beneficial and most probably necessary. * Do you recommend using the SBTIV correction for the longitudinal analyses as well? * The volumes I get if I look at the statistics for SEG in 3D slicer, they differ by up to 2%. Is there another scaling factor applied in the samseg.stats file? * What is the the SUBJ_scaling.txt file referring to? Should I use that to make corrections? 2) The lesion version of SAMSEG long is available in the development version. Enabling the lesion extension can be done by adding the same flag(s) as for the cross-sectional lesion version (--lesion, see the Wiki for details: https://secure-web.cisco.com/1zxjznTdMuBEIc1Dx0NZk6wpPdZ8iSQLZ2MiValYSz09WG…) Thank you! Unfortunately the wiki doesn’t specify how I could get access to the development version? 3) We never conducted extensive experiments for the number of burn-in and samples for the MCMC scheme, but we found that 50 burn-in and 50 samples lead to robust and replicable lesion segmentation results. Increasing these two numbers might increase the lesion segmentation performance, although differences might be minimal and the overall computational cost high. We suggest to not decrease these two numbers as lesion performance could potentially decrease. This is very very helpful! Is there a way to force CUDA for the in the lesion segmentation? In order for me to be able to run the lesion segmentation under Ubuntu, I had to install both the CPU and the GPU tensorflow, and it does not seem to use the GPU at all (TensorFlow puts out the CPU specs but not the GPU and GPU utilization % doesn’t change during the processing). Thank you for the help! David Hi FreeSurfers, Thank you for sharing the new SAMSEG tool, which I consider a major breakthrough in image processing! I have a few questions about using SAMSEG vs SAMSEG_Long: 1. What volumetric differences should I expect between running SAMSEG on individual images, vs. using SAMSEG_Long for the longitudinal analysis? Could I use SBTIV to adjust differences between scans? (all images are FLAIRs preprocessed using mrI_robust_template) 2. Is there currently a way to integrate lesion segmentation into the SAMSEG_long?s longitudinal analysis? 3. What is the expected impact of increasing or decreasing -Burn-In and -Sample numbers for lesion segmentation? Thank you for all the help! Best, Dave Kamson

1 0

Voxel-wise volumes vs APARC volumes
by Mark Wagshul 31 Jan '21

31 Jan '21

External Email - Use Caution Dear Freesurfer experts, I'm stuck on something which I think has to do with how Freesurfer calculates volumes, but can't quite wrap my head around it, hopefully you can enlighten me. I have already produced robust results for our data using region-based cortical volumes, using aparcstats2table. This is in a group of 55 healthy older adults, modeling is using linear mixed effects models. All volumes were normalized with the eTIV. We were hoping on exploring the same models with voxel-wise analyses, and are getting wildly different results. Data were created using mri_preproc, resampled onto fsaverage, and my assumptions was that because of this resampling, we should NOT be normalizing these values to the eTIV. The same regions which were highly significant on the region-based analyses and now not significant at all, and other regions are now showing up as significant. Just as a sanity check, I attempted to reproduce the region-based analyses with these data using the following simple (I thought!) procedure: 1) Created an mgz mask of the region I'm interested in, using mri_annotation2label and mri_label2label 2) Read my volume data (#subjects x #vertices) into Matlab 3) Read my mask into Matlab (yes, I know I can probably do this in Freesurfer, but I'm more comfortable with Matlab!) 4) Multiply them together and take the sum across vertices. 5) Feed the resultant data back into my LME model. Two things are puzzling to me. 1) The highly significant results from the old LME are no longer significant (at the moment, only checked for two of the cortical regions), and 2) The correlation between these new data and the aparc volume are OK, although far from perfect (R2 ~ 0.86). So, I'm obviously missing something here. What is the difference between the two volume calculation methods which could explain these discrepancies? Any help with this would be MOST appreciated! Best, Mark [cid:image004.jpg@01D6F59B.DA0205F0] ____________________ Mark Wagshul, PhD Associate Professor Gruss Magnetic Resonance Research Center Albert Einstein College of Medicine Bronx, NY 10461 Ph: 718-430-4011 FAX: 718-430-3399 Email: mark.wagshul(a)einsteinmed.org<mailto:mark.wagshul@einsteinmed.org> [einstein-logo-rgb] This email message and any accompanying attachments may contain privileged information intended only for the named recipient(s). If you are not the intended recipient(s), you are hereby notified that the dissemination, distribution, and or copying of this message is strictly prohibited. If you receive this message in error, or are not the named recipient(s), please notify the sender at the email address above, delete this email from your computer, and destroy any copies in any form immediately.

2 4

2026

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

Freesurfer February 2021