From the recon-all.log files that you shared, I noticed that you set SUBJECTS_DIR to the same directory for both 5.3.0 and 7.4.1 runs. I think it is better to keep them separate because there are symbolic links under the SUBJECTS_DIR pointing to release specific files.

The FS5.3.0 was built on centos6. Running it on Rocky8 machine might run into some compatibility issues.

From: Kim, Minhae <EKIM35@mgh.harvard.edu>
Sent: Thursday, February 26, 2026 5:17 PM
To: freesurfer@nmr.mgh.harvard.edu
Subject: [Freesurfer] Truncated T1.mgz after recon-all

Hello FreeSurfer team,

I encountered an error while running recon-all where T1.mgz files for two subjects appear to be cropped posteriorly along the Y dimension (RAS). The cortex is preserved, but portions of the skull and the scalp are removed. The input file ($subj/mri/orig/001.mgz) is intact, suggesting a possible issue during conversion (e.g., mri_convert).

I didn’t find similar reports in the archive, and the problem reproduces identically across both versions of FreeSurfer below. recon-all.log files are attached. 

• FreeSurfer: freesurfer-Linux-centos6_x86_64-stable-pub-v5.3.0 AND freesurfer-linux-centos7_x86_64-7.4.1-20230613-7eb8460

• OS: Rocky 8.10

• uname -a: Linux r440-05.nmr.mgh.harvard.edu 4.18.0-553.79.1.el8_10.x86_64 #1 SMP Wed Oct 15 12:48:32 UTC 2025 x86_64 x86_64 x86_64 GNU/Linux

Any insight would be greatly appreciated. Thank you!

Kindly,

Ellie

Ellie Minhae Kim (she/her/hers)

Program Director | Pain and Neuroinflammation Imaging Lab

Athinoula A. Martinos Center for Biomedical Imaging

Massachusetts General Hospital

149 Thirteenth St, Suite 2316

Charlestown, MA 02129

Phone: (617) 643-0569

https://loggialab.mgh.harvard.edu/