Error in TI tutorial

[abbytang]

Hello! I am new to using Sim4Life and am currently working through the tutorial titled tutorial_TI_|X|025_head_model.ipynb. However, I encountered an error while running the code, specifically in the "Extract Surfaces" part. The error message states: "invalid input labeled."
Could you kindly assist me in resolving this issue? Thank you for your help.

[abbytang]

I used the Controller to run the auto-label function (head modeling in the Ribbon), but encountered the same error as in the Python script. The error log is shown in the attached image.

[bryn]

The error looks like something is wrong with the matplotlib package in your installation (the MPLBACKEND environment variable). Did you possibly install any python packages in Sim4Life's Python with 'pip'?

Can you run any other notebooks with matplotlib plots?

[abbytang]

I didn’t install any Python packages in Sim4Life’s Python environment using 'pip.' I installed this software by selecting all components during setup.

I’m not sure if my computer’s GPU (NVIDIA GeForce MX250) influenced this process. However, I chose "install deprecated NVIDIA driver" when installed this software.

[abbytang]

When I restart the software and only import IXI025-Guys-0852-T1.nii.gz, the auto-label function works correctly.

However, when I run tutorial_TI_|X|025_head_model.ipynb, the same error occurs again.

After the error occurs, if I only load IXI025-Guys-0852-T1.nii.gz and run the 'auto-label function' (head modeling in the Ribbon), the same error now appears at this step. The only way to resolve it is to restart the software and reload IXI025-Guys-0852-T1.nii.gz—then it works correctly again.

However, if I try to run the TI tutorial, the error persists even after restarting the software.

![0BB7M7UAW`1BYC63FR)U]26.png](/assets/uploads/files/1748234490112-b021ff28-75cd-48f4-a814-49a0e4042b11-0bb7m7uaw-96-1byc63fr-u-26.png)

[abbytang]

I installed the software on a computer with an NVIDIA GTX 1060 GPU, but the errors still persist.

[bryn]

@abbytang Thanks for the update. We will try to reproduce the issue, and if possible either provide a workaround or contact you to understand why it happens.

Can you run other tutorial notebooks, e.g. the mesh extrusion tutorial?

[abbytang]

Thank you for your reply. The mesh extrusion tutorial runs correctly. Since my research focuses on TI, I need tutorial_TI_|X|025_head_model to run correctly. I would greatly appreciate any assistance with resolving those issues.

[bryn]

Are you using the latest Sim4Life, v8.2.2.18061?

[abbytang]

Yes, I'm currently using the latest Sim4Life version (v8.2.2.18061). I will also install version 8.2.0.16876 and test it again for comparison.

[abbytang]

When I copy the file tutorial_TI_|X|025_head_model.ipynb into the CAD files folder of Sim4Life version 8.2.0.16876 and run it in this older version, most sections execute correctly, including the "Auto Label Tissues and Extract Surfaces" part. However, the script fails at the "Computing the Maximum Modulation Using the Masked Electric Fields" section with the following error:
AttributeError: module 's4l_v1.analysis.core' has no attribute 'MaxModAmpEvaluator'.
Could you please advise on how to resolve this issue? Thank you for your assistance.

[bryn]

the MaxModAmpEvaluator was exposed in s4l_v1 6 months ago, so after 8.2.0.16876. You could probably use it though via the 'raw api': XPostProcessor.MaxModAmpEvaluator.

[lucky_lin]

@bryn Hello, I am using a script to run TI simulations. I noticed that when I change the maximum grid step size for the electrodes and tissues from [2.0, 2.0, 2.0] to [3.125, 3.125, 3.125] in each simulation, the peak value of the calculated electric field (E-field) changes from 205 V/m to 445 V/m. Why does the grid step size affect the final calculated E-field value? Additionally, I found that changing the step size also significantly alters the Scale Factor used to normalize the results in each simulation. Here is the relevant code :
manual_grid_settings.MaxStep = numpy.array([3.125, 3.125, 3.125]), units.MilliMeters
manual_grid_settings.Resolution = numpy.array([1.0, 1.0, 1.0]), units.MilliMeter

[bryn]

@lucky_lin Please post your question as a separate thread. This is unrelated to the topic of this thread. Thanks

[bryn]

@abbytang I can reproduce your issue in v8.2.2.18061. Investigating what is causing it and what workaround i can offer (except for reverting to the previous version). Sorry for the inconvenience.

[abbytang]

Thank you for your help. I will run it in a previous version.

[halder]

Hi, @abbytang this is happening because in Sim4Life 8.2.2, due to internal environment changes the matplotlib default backend is set to: matplotlib_inline. The fix is to override the backend manually to a compatible one (Agg) early in the script. The fixed tutorial will be included in the v9 which will release soon. However, for now you can add this to the beginning of the script
"import os
os.environ["MPLBACKEND"] = "Agg"

import matplotlib
matplotlib.use("Agg")"

Please send an email to s4l-support@zmt.swiss and I can send you the updated jupyter notebook.

[abbytang]

Thank you. It can work right now.

[bryn]

BTW. this is a temporary fix, which breaks inline plotting with matplotlib. We have fixed it in the release branch. A slightly improved workaround may be to temporarily unset the MPLBACKEND environment variable before the call to HeadModelGeneration, e.g. something like this

import os

original_backend = os.environ.pop("MPLBACKEND", None)
labelfield = HeadModelGeneration(
    images=image_list,
    add_dura=True,
    close_skin=True,
    close_csf=True,
    close_skull=True,
    version=ImageML.eHeadModel.head16,
)
if original_backend is not None:
    os.environ["MPLBACKEND"] = original_backend