Dear Reo,

Thanks for the interest in our atlas, and apologies for the slow response (I was travelling).

Most important: thanks for raising this issue. I’m fairly confident it’s due to some labeling / co-registration issues. For the time being: you can get a binary mask for this region by keeping the largest connected component. I’ll try to fix the atlas as soon as I can.

Best wishes,

/Eugenio

Juan Eugenio Iglesias

http://www.jeiglesias.com

From: asada.reo.42z@st.kyoto-u.ac.jp <asada.reo.42z@st.kyoto-u.ac.jp>
Date: Wednesday, June 24, 2026 at 12:28 AM
To: freesurfer@nmr.mgh.harvard.edu <freesurfer@nmr.mgh.harvard.edu>
Cc: Takahiko Inagaki <taka_inagaki@kuhp.kyoto-u.ac.jp>
Subject: [Freesurfer] Inquiry about the segmentation of deep-white structure (stria medullaris of thalamus)

External Email - Use Caution

Dear Professor Iglesias and the FreeSurfer Development Team,

I hope this email finds you well.

My name is Reo Asada, and I am a sixth-year medical student at the Department of Psychiatry, Kyoto University. I am currently working on a neuroimaging research investigating the subcortical structures associated with Major Depressive Disorder (MDD) led by my senior collaborator, Dr. Takahiko Inagaki, who will be the first author and is CC'd on this email.

For our structural analysis, we are utilizing your NextBrain atlas.I would like to thank you for providing such an exceptionally powerful resource to the imaging community.

I am writing to kindly seek your guidance regarding an issue I encountered with the segmentation of the stria medullaris of the thalamus (SMe) (Label 201).

When running the NextBrain pipeline on our dataset, the resulting SMe segmentations appear significantly fragmented and frequently mislocalize into the adjacent fornix area.

My environment is Freesurfer 8.2.0 with the latest fix applied on Ubuntu 22.04

For reference, I have attached a screenshot demonstrating this issue. The background image is T1w_acpc_dc_restore.nii (processed via the HCP pipeline in native ACPC space, 0.8 x 0.8 x 0.8 mm resolution, image from HC). The overlays shown are:

white_matter_of_forebrain (Label 7, light green)
pons (Label 580, deep green)
stria_medullaris_of_thalamus (Label 201, red)

The command used to create the segmentation is :

mri_histo_atlas_segment_fireants --i /mnt/h/Habenula_Test_Retest/HCPPipeline/Nextbrain/Images/NLH121/run01/T1w/T1w_acpc_dc_restore.nii.gz --o /mnt/h/Habenula_Test_Retest/HCPPipeline/Nextbrain/Images/NLH121/run01/T1w/NextBrain --device cuda --mode invivo --threads 20 --side left

(I think that this follows the standard command for mri_histo_altas_segment_fireants)

Given the highly delicate nature of the SMe, I would deeply appreciate your advice on the following points:

Are there specific command-line flags, resolution adjustments, or preprocessing steps you recommend to improve the continuity and boundaries of such small tracts in NextBrain?
Is this level of fragmentation expected when operating on standard isotropic T1w resolutions, or could this point to an issue with the underlying registration/priors for this specific structure?

Thank you very much for your time, consideration, and continued support of the community.

Sincerely,

Reo Asada

Sixth-year Medical Student

Faculty of Medicine, Kyoto University