I don't think it should be a problem to run on skull stripped data. It is ok to run the default recon-all on the submillimeter data as it will just downsample it to 1mm. But using the -hires option is good to get any hires info. It does take a lot longer and sometimes it does funny things, but if the final images and surfaces look ok, I think it is good. You might try version 8 on mp2rage (without skull stripping) as I think it works well (but correct me if you find otherwise)
On 4/16/2025 8:54 AM, Sabrina Turker wrote:
External Email - Use Caution
Dear all,
I'm new to freesurfer and I'm trying to preprocess a 7T mp2rage dataset (0.6mm³ isotropic, TR=6000ms, TE=2.05ms) with freesurfer 7.2.0. The images have been corrected for B1, for intensity inhomogeneity with ANTs N4 bias field removal and they were skull striped with FSL based on intensity. This was done bya colleague who analyzed this dataset in 2017 and I'm hoping to run our updated toolboxes on the preprocessed data. However, he ran into several problems while first analyzing the data: freesurfer did not like the noisy background of the mp2rage, which is why he skullstripped the data before preprocessing it with freesurfer. He then ran recon-all once with the default pipeline and once with skull strip deactivated. According to him, the default pipeline yielded better results. I'm using his skull-stripped data now for the preprocessing and was wondering whether not deactivating the skull strip will lead to inaccurate measures for cortical thickness and gray matter volume. Since this is crucial for our toolboxes, we really want to be sure that we're going for the best solution. A second problem is the high resolution of the data - it's not recommended to run the default recon-all on anything lower than 1mm³ (*MailScanner has detected a possible fraud attempt from "secure-web.cisco.com" claiming to be* https://surfer.nmr.mgh.harvard.edu/fswiki/SubmillimeterRecon), but most of the information I could find is from 2016 or 2017. At the time, he used the -hires flag but got very bad results, which made him use the default pipeline even if the mp2rage data got downsampled. I was wondering if anybody has feedback on how they analyzed a similar dataset (high resolution, problems with skullstripping etc.), general opinions on the best way to analyze this dataset and/or if there have been any developments (e.g., new flags for recon-all flags?) that I'm not aware of. We got feedback from one colleague that they did analyze their dataset successfully with the -hires flag and did not encounter any problems at the time. Any advice would be much appreciated!
Best and thanks in advance,
Sabrina
-- Dr. Sabrina Turker (she/her) Postdoctoral researcher Brain and Language Lab Department of Behavioral & Cognitive Biology University of Vienna
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer