Inquiry: Flex-Search electrode outputs, Gmsh visualization, and Freeview settings for deep brain targets

[LaixSS]

Dear TI-Toolbox Developers,
Thank you for developing and sharing such an excellent open-source platform for temporal interference research. I am currently using the TI-Toolbox v2.2.4 and have a few questions regarding the extraction and visualization of my simulation results.

1. Identifying selected electrodes: I have successfully run the Flex-Search Optimizer and the Simulator, using the EEG10-10_Cutini_2011 EEG cap. Where or how can I see exactly which electrodes were calculated and selected by the optimizer for my montage?
2. Visualizing electrodes: How can I display these selected electrode positions on the 3D model in Gmsh? Additionally, are there other recommended ways or built-in tools to visualize the electrode placement on the head?
3. Freeview visualization for deep targets: When simulating the electric field and viewing the NIfTI results in Freeview, how can I clearly display the field strength at my specific target? I selected the subthalamic nucleus as my target, but with the default Freeview options, there does not seem to be a noticeable field strength at this deep location. Could you explain what the different visualization options in the Freeview tab mean? Do you have any suggestions or recommended settings for properly visualizing field strengths in deep subcortical targets?  

Thank you very much for your time and assistance. I look forward to hearing from you!
Best regards, LaixS

[idossha]

HI @LaixSS,

I am glad to hear you are finding the software helpful for your research, and thank you for reaching out. Find below answers to your questions:

1. Electrode question.

If you used the mapping approach when simulated your optimized results, you should be able to find the selected electrodes in the following places:

• /{project_name}/derivatives/SimNIBS/sub-{id}/Simulations/{simulation_name}/TI/montage_imgs/{simulation_name}_highlighted_visualization.png

example:

• Also, you have a .json file containing some of those key parameters in /{project_name}/derivatives/SimNIBS/sub-{id}/Simulations/{simulation_name}/documentation/

2. Electrode visualization: Currently the only built in way is via the built-in Blender inregration. Please see:

• wiki page: https://idossha.github.io/TI-Toolbox/wiki/blender/
• Video tutorial: https://idossha.github.io/TI-Toolbox/wiki/video-tutorials/

3. Freeview visualization for deep targets:

To visualize deep targets, please run a voxel level analysis in the analyzer tab, and make sure you select your region
or coordinates if you used a spherical target. Then it should be recognized by the Freeview tab and load it automatically for you.

example:

[LaixSS]

Thank you so much for the guidance! The first two points are perfectly clear now.
However, I still have a few questions regarding the final point. Even after following your suggestions to adjust the settings, I am still not seeing the expected field strength visualization at my target in Freeview. I am wondering if there might be an issue with my upstream pipeline.
Below, I have outlined my general workflow step-by-step, along with a few specific questions where I have some doubts:

1. Optimizer Tab - Basic Parameters: Since my target is a small, deep nucleus, are my selections for the basic parameters correct?

2. Optimizer Tab - ROI Definition: When locating the ROI, my target coordinates in Freeview are as follows:

Based on these coordinates, is there any problem with how I filled out the ROI Definition here?

(💡 A quick feature suggestion: If a user has already provided T2w data in the subject's folder, it would be incredibly helpful if Freeview could automatically import/overlay the T2w data when picking coordinates. Certain deep nuclei are much easier to identify on T2 scans than T1.)
3. Simulator Tab: I kept the Electrode parameters exactly consistent with what was set in the Optimizer tab.
4. Analyzer Tab: I filled out the fields as shown below, ensuring that the RAS coordinates exactly match my target coordinates:

5. NIfTI Viewer Tab: Are my configuration settings correct here?
6. Freeview Overlays: After clicking "Load Subject Data" and opening Freeview, exactly which of these overlay options/checkboxes should I select to properly display the field strength at my target?

Thank you again for your time and patience in helping me troubleshoot this!

[LaixSS]

Please note: You might notice slight inconsistencies in the exact coordinates across some of the screenshots below. A few of them were taken during previous runs, but the general target location remains the same.

[idossha]

Thanks for the suggestion. I will implement this in an up-coming release.

Please make sure to run the analysis in voxel space. That will generate the file you need.

[LaixSS]

Following your suggestion, I have run the analysis using the voxel-based approach in the Analyzer tab. However, I am still not seeing any noticeable field strength at the exact target coordinates. Instead, the electric field seems to be distributed around the target region.

[idossha]

What the color indication here shows is not field strength, but modulation depth. i.e. TImax, or the "depth of the envelope".

Currently, it looks like your hotspot is above your intended target. This is likely because your target is contains mainly WM and not GM. By default, the optimizer will only take into account GM area and not WM. Please Look at the logs of the optimizer to see how successful or not it was based on the objective function. The more negative it is, the better it performed.

If you specifically need to target WM, you may need to do some manual modification to the codebase. However, I will think of the option of including this in the next release.

In any case, it sounds like your project is quite unique, I am happy to schedule a call and see what your best approach should be, and if I can assist you with your project requirements.

I am hoping to finalize the new release within about one week.

Ido

[LaixSS]

Hi Ido,
Thank you so much for the detailed explanation and for your incredibly generous offer to schedule a meeting! I truly appreciate your dedication to helping users and supporting my project.
Regarding the target tissue, I am actually a bit confused and would love your insight. My intended ROI is the subthalamic nucleus (STN). Anatomically, the STN itself is primarily composed of gray matter (GM). However, it is a very small deep-brain structure that is tightly surrounded and enveloped by thick white matter (WM) tracts.
Do you think it is possible that the automated tissue segmentation pipeline (e.g., FreeSurfer/charm) might have misclassified or merged this specific small, deep GM nucleus into the surrounding WM? If the software's GM mask failed to capture the STN properly, that would perfectly explain why the optimizer ignored it and shifted the hotspot to the clear GM regions above it.
Regarding your kind offer to connect, I would absolutely love to discuss the best approach and the upcoming release with you. However, I must admit that I am not very confident in my spoken English, and I worry that a traditional voice or video call might not be very efficient due to the language barrier.
Would it be possible for us to have a live text-based chat instead? We could use any platform that is convenient for you.
Please let me know if a live text chat works for you, and what days/times would suit your schedule.
Thank you again for your amazing support and understanding!
Best regards,LaixS

[idossha]

Hi @LaixSS,

It is okay, we can keep chatting here. Unless you want a more involved help/collaboration, than I would suggest to meet via Zoom.

In the meantime, you bring a very important point, that thus far, we have not considered.

I am working on a solution for it. Currently, FreeSurfer and SimNIS do not allow for automatic segmentation of deep nuclei. However, I may have found a solution that I will integrate in the coming release.

Please refer to this paper: https://elifesciences.org/articles/59430

It gives a probabilistic map of those deep nuclei which we can integrate in the TI-Toolbox. Will that help your project?

It contains the STN.

Ido

[LaixSS]

Hi Ido,
Thank you so much for your understanding and for agreeing to continue our discussion here. I really appreciate your flexibility!
And wow, thank you for looking into the segmentation issue so quickly! Integrating the probabilistic map of deep nuclei from that eLife paper sounds like a fantastic and elegant solution. Yes, since it contains the STN, it would absolutely be a huge help and a game-changer for my project!
Regarding this upcoming integration, I do have one technical/theoretical question: If we successfully integrate this automatic segmentation (so the optimizer perfectly recognizes the STN as a targetable GM region), can it guarantee that the energy (the TImax modulation depth) will be precisely focused directly at the center of the STN?
Given that the STN is a very small deep nucleus and is tightly surrounded by thick white matter tracts (and potentially nearby CSF), might the physical properties of the tissue (like current shunting) or the optimization algorithm still cause the actual hotspot to shift slightly off-center? Or will the optimizer be powerful enough to force the peak field directly into the STN center once it is properly masked?
I am very much looking forward to the next release. If you need anyone to beta-test this specific deep-nuclei feature on real subject data before the official launch, I would be more than happy to help!
Thank you again for your incredible and proactive support.
Best regards,
LaixS

[idossha]

Hi @LaixSS,

Please let us know if the new version answered your needs for analysis and visualization of sub nuclei in both GM and WM.

[idossha]

Hi @LaixSS — sorry for the delayed follow-up. A related set of GUI/Analyzer reliability improvements is queued for v2.3.1.

The upcoming release improves Analyzer/NIfTI Viewer refresh behavior and adds clearer messages when expected TI/mTI mesh or NIfTI post-processing outputs are missing. This should make it easier to distinguish a visualization/settings issue from an upstream simulation or registration/output issue.

The deep-target visualization recommendations remain the same for now: use voxel-space analysis for small deep structures and verify the anatomical/registration quality carefully. The GUI reliability updates should be released with v2.3.1 in the coming days.