Thank for the comprehensive feedback, Luis!
* ROIs: A lot of small structures were merged with others when building (and then further simplifying) the atlas, including (pro-)subicular labels. Sorry for the inconvenience. Even with histology, it can be tricky to reliably label some of these. I’m CCing our co-author Martina in case she can shed further light on this. * Thanks for detecting the bug. Trying to keep all the versions has led to rather messy code. We are in the process of reorganizing / refactoring the code and hope to have a clean version out soon (2-3 weeks). * The error with the missing atlas should be solved by now in the development version; apologies for the inconvenience.
Again: thanks for you feedback and for your patience with our new software! /E
--
Juan Eugenio Iglesias
From: Luis Alberto Barreiro Araújo pg54004@alunos.uminho.pt Date: Saturday, January 17, 2026 at 14:11 To: Iglesias Gonzalez, Juan E. JIGLESIASGONZALEZ@mgh.harvard.edu Cc: freesurfer@nmr.mgh.harvard.edu freesurfer@nmr.mgh.harvard.edu, Tiago Gil Rodrigues Oliveira tiago@med.uminho.pt, Francisco Almeida franciscojcga@gmail.com Subject: NextBrain atlas: Missing hippocampal subfields in FreeSurfer 8.1
External Email - Use Caution Dear Dr. Iglesias, I'm a researcher at the University of Minho - ICSV Institute, conducting research on hippocampal patterns in Alzheimer's disease. I've used NextBrain for detailed hippocampal and full segmentation on NACC dataset subjects. Issue: I noticed your NextBrain demo website shows hippocampal structures including: prosubiculum (uncal, rostral, caudal) presubiculum (rostral, caudal) indusium_griseum supracallosal_subiculum However, these structures don't appear in either atlas_simplified or atlas_full YAML configuration files in FreeSurfer 8.1. I've verified this by checking: data_simplified/atlas_names_and_labels.yaml data_full/atlas_names_and_labels.yaml Additionally, when attempting to use atlas_full, the segmentation crashes with "IndexError: list index out of range" at label 266, suggesting the data_full/combined_atlas_labels_1mm.yaml config is incomplete (maps only 265 of 334 labels). Questions: Are these additional hippocampal subfields available in FreeSurfer dev or a more recent atlas version? If not yet released, are they planned for a future update? Would you recommend an alternative approach for segmenting these specific structures? Should I post these issues in the official FreeSurfer GitHub repository for broader advice ? Note: I'm running NextBrain via FreeSurfer 8.1 Docker container, due to incompatibility issues in our workstation with the OS. The Docker image was missing both the atlas files and the Python segmentation scripts, so I had to manually download:
* atlas_simplified.zip and atlas_full.zip from the FTP server * ERC_bayesian_segmentation scripts from the GitHub dev branch * SuperSynth_August_2025.pth model (1GB)
After mounting these files into the container, atlas_simplified works perfectly. However, atlas_full crashes as described above, and neither atlas contains the additional hippocampal subfields visible in your demo. From my understanding, recently (in the past week) a new version with the bayesian segmentation folder was put available but the issues I have remain. Another small note: I had to manually correct the mri_histo_atlas_segment_fireants.sh script because some of the parameters for the python scripts where missing. I'm not sure if that's supposed to happen or if it's a version error. Thank you for your attention. I look forward to receiving a response. Best regards, Luís Researcher, Informatics Engineering University of Minho - ICVS
freesurfer@nmr.mgh.harvard.edu
-
Iglesias Gonzalez, Juan E.