External Email - Use Caution

Hi,

I am currently trying the FreeSurfer 'optimized' pipeline for high-resolution (0.6mm)3 7T MPRAGE datasets as described in this link: https://secure-web.cisco.com/10HAtUm2Jw0xOwFBksoxzxFEJ8D6Vv8ds_sG8BE1pIzC3hT... HighFieldRecon - Free Surfer Wikihttps://secure-web.cisco.com/10HAtUm2Jw0xOwFBksoxzxFEJ8D6Vv8ds_sG8BE1pIzC3hTe09ztX-zuEg2RZMzmbnijrKiBusyxDh3pLVmaRgP42jzPqb2vqYCjA6SDqCJsYkIWj1kMSt_w3HNkVf88AAWWZgZ2GgQFjIqcvjEe-JzIMsYORiAedMCje7wt9285kVOYUdEl4GSmP0my7DnXy-dTS_8DJsmDdGayFnkKrAXE0-9UNN8dWelEuNBdWcEgwyZUVIPu3LqerJKsHRrxhqm0qNFsAUckLFJE5aT-mn-dmMkGqyQJbxhuGHKVT5TzrosAkJjJaiiVE9m6_5VrZzFOJybvK7dfYyTKJU6dqFg/https%3A%2F%2Fsurfer.nmr.mgh.harvard.edu%2Ffswiki%2FHighFieldRecon Running recon-all on data collected at 7T. See also: SubmillimeterRecon Bias correction prior to recon-all. Compared to 3T MPRAGE and MEMPRAGE scans collected at 7T have stronger intensity inhomogeneity, which won't be fully corrected in the default recon-all. surfer.nmr.mgh.harvard.edu

However, I am facing issues with the inhomogeneity correction done during autorecon1. In fact, the rescaling step between orig.mgz -> orig_nu.mgz AND the '-nuintensitycorr' step between orig.mgz -> nu.mgz completely destroys the homogeneity gained from the SPM12 BiasField Correction applied beforehand.

The exact command line I'm using is:

$ recon-all -all -s P7T_020_001bc -hires -i BC_MPRAGE_iPat2.nii -expert expert-file.opts

- where BC_MPRAGE_iPat2.nii is the output of SPM12 BiasField Correction (i.e., the bias field corrected MPRAGE image) and expert-file.opts the file containing ''mris_inflate -n 100".

I am using FS version 7.2.0 (freesurfer-linux-ubuntu18_x86_64-7.2.0-20210721-aa8f76b) on Linux Ubuntu 18.04 bionic (Linux mrphys-gpu1 4.15.0-180-generic #189-Ubuntu SMP Wed May 18 14:13:57 UTC 2022 x86_64 x86_64 x86_64 GNU/Linux).

The whole recon-all runs properly without errors, but the homogeneity is worse after the NU intensity correction... See screenshots in attachments (orig.mgz VS nu.mgz). I have also attached the recon-all.log file.

I am currently testing the '-nonuintensitycor' flag, but it exited WITH errors (not the recon-all.log file attached).

In addition, I tried the exact same pipeline but without passing the data through SPM12 beforehand. It still did the same thing of destroying the small amount of homogeneity in the image to an even worst level of inhomogeneity.

On another matter, this part in the link I provided above is quite unclear:

"""

Recommended bias-correction parameters:

* Bias FWHM = 18 * Sampling distance = 2

"""

A Bias FWHM of 18 mm is not possible with SPM12 (minimum being 30 mm). So, is it for SPM12 or something as part of recon-all? For the sampling distance, it's fine with the 2 mm in SPM12.

Have a great day! 🙂

Marc-Antoine

_______________________________ Marc-Antoine Fortin +47 45 84 99 61 Stipendiat / PhD Candidate Department of Physics Norwegian University of Science and Technology (NTNU)

[cid:95742765-3552-411c-86af-f0811654e0ac] _______________________________