How to reconstruct ADNIMERGE seg volumes from freesurfer output

[Mathilde Bateson]

Hello

I have re-run Freesurfer on some ADNI subjects and want to compare the volumes I obtain from what is in ADNImerge.csv to see if I get the same results, the % diffs... 

Specifically the Hippocampus volume but other volumes as well which are columns in the file ADNImerge.csv : Ventricles, WholeBrain,  ICV.

To reconstruct the volumes from the output of freesurfer (aseg.stats file) I did : 

Hippocampus = Left-Hippocampus+ Right-Hippocampus

Ventricle = Left-Lateral-Ventricle+Left-Inf-Lat-Vent+Right-Lateral-Ventricle+Right-Inf-Lat-Vent

WholeBrain = BrainSeg

ICV = EstimatedTotalIntraCranialVol

What's weird is I get very different % differences between structures, leading me to think my reconstruction "FreeSurfer volumes to ADNImerge volumes" above are wrong.

For a same patient_visit I get 

0.45% difference between hippocampus values

and 18% difference for ICV values !

Here are my questions:

 1. what am I doing wrong

2. what do you reckon is the precision tolerance (% difference acceptable between my freesurfer output volumes and the ADNImerge values above which I say there is a pb )

I attach the detail of the comparison if its useful.

Cheers !

Mathilde

[Danielle J Harvey]

Hi Mathilde,

 

In the future, please do not include any of the actual data in the thread. I realize you have only included some for 1 participant, but we still want to be sure that only those with approved access to the data can actually work with the data. You could just give a participant ID with the values you obtained yourself with a summary of the differences and we could then look at the ADNI data.

 

ADNIMERGE contains Freesurfer values from a number of different versions. I wouldn’t expect versions to have huge differences, but there may be some. It is also including data from the “cross-sectional” pipeline, rather than the longitudinal pipeline.

 

That being said, the ICV measure in ADNIMERGE corresponds to ST10CV (or Volume (aparc.stat) of Icv or Cortical Volume (aparc.stats) of Icv depending on the version of FreeSurfer). Hippocampus is the sum of Rt and Lt Hippocampus (ST29SV+ST88SV; Volume (WM Parcellation) of LeftHippocampus+Volume (WM Parcellation) of RightHippocampus). Ventricles is the sum of ST30SV, ST37SV, ST89SV and ST96SV, which appear to be the same variables you are using (lt/rt lateral ventricle and lt/rt inferior lateral ventricle). Whole brain is a sum of a bunch of variables ('ST128SV', 'ST17SV', 'ST18SV', 'ST61SV', 'ST16SV', 'ST53SV', 'ST42SV', 'ST29SV',  'ST12SV', 'ST11SV', 'ST65SV', 'ST76SV', 'ST77SV', 'ST120SV', 'ST75SV', 'ST112SV', 'ST101SV', 'ST88SV', 'ST71SV', 'ST70SV', 'ST124SV', 'ST147SV', 'ST148SV', 'ST150SV', 'ST151SV'). You can refer to the ADNI data dictionary to see which regions were included in Whole brain.

 

Danielle

Mathilde Bateson​​​​ |  Senior Data Scientist Confidentiality – Restricted Dissemination This email and any attachments are confidential and shall be processed by the recipient with the same level of protection as receiving recipient confidential information. If received in error, please do not disclose the contents to anyone, but notify the sender by return email and delete this email (and any attachments) from your email inbox. ​ ​Data privacy Owkin processes personal data, including data relating to email communications, in accordance with the terms of its Privacy Policy available by clicking to the link in the footer of Owkin's website.

--
You received this message because you are subscribed to the Google Groups "Alzheimer's Disease Neuroimaging Initiative (ADNI) Data" group.
To unsubscribe from this group and stop receiving emails from it, send an email to adni-data+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/adni-data/ec3e2575-9010-422b-a8c3-98e81c746dden%40googlegroups.com.

**CONFIDENTIALITY NOTICE** This e-mail communication and any attachments are for the sole use of the intended recipient and may contain information that is confidential and privileged under state and federal privacy laws. If you received this e-mail in error, be aware that any unauthorized use, disclosure, copying, or distribution is strictly prohibited. If you received this e-mail in error, please contact the sender immediately and destroy/delete all copies of this message.

[Mathilde Bateson]

Hello and thanks a lot for the anwser above :)

I would like to reproduce the pipeline that goes from the images in MRI data to MRI_preprocessed data.
I believe this is mainly doing the preprocessing steps : (Gradwarp, B1 correction, N3 correction, dicom -> nifti , maybe skull stripping also ?)
 Is the code available to do this ?
is this pipeline similar to what is happening in -autorecon1 of recon-all from FreeSurfer for instance ? Or does this pipeline have to be done before running FreeSurfer recon-all to expect accurate values for brain structure volume? thanks ! 

[Mathilde Bateson]

Hello again ADNI expert ! 

I have a question regarding the T1 processing : I am looking at the ADNI_merge.csv and seeing that the column FSVERSION says that the cross-sectional freesurfer was used :
ex = Cross-Sectional FreeSurfer (FreeSurfer Version 4.3)

-> why wasn't the *longitudinal* pipeline of freesurfer used for better results on accuracy of volumes etc?

See below:

"This pipeline processes a series of images acquired at different time points for the same subject with the longitudinal FreeSurfer stream [Reuter et al., 2012] to increase the accuracy of volume and thickness estimates. "

Thanks !

Mathilde

[Danielle J Harvey]

Hi Mathilde,

 

There are some files available for download on the LONI ADNI website utilizing the longitudinal Freesurfer pipeline. This pipeline is typically run at some point after a phase of ADNI is complete, since the pipeline requires all images for all participants.

 

To view this discussion visit https://groups.google.com/d/msgid/adni-data/0e7659f1-34e1-42ca-a370-cc1c964f9998n%40googlegroups.com.

[Mathilde Bateson]

Hello again Danielle, 

I am still focusing on reproducing the Freesurfer pipeline. I am using Freesurfer 6 so I have focused on patients that were also processed by Tosun group with  Freesurfer 6 (as seen in ADNIMERGE file).
I reconstruct the Hippocampus, Ventricles, and ICV volumes. 
For Hippocampus and Ventricles, I get what I believe is an acceptable discrepancy (average <1%) between the expected volume (i.e., the one found for that patient_month in ADNIMERGE file) and my output (the volume reconstructed from aseg).

However, for ICV, I get a systematic overaluation of +4% in my output versus what is found in the ADNIMERGE file. Could you help me find a possible reason / next step ?
Thanks !