Unable to Automatically Execute Ultrasound Simulation from Saved Workspace

ImFusion Suite Ultrasound
1

Hi,

When I load a saved workspace (.iws), the ultrasound simulation doesn’t automatically run. I tried adding 1 inside the simulation algorithm block, but then it throws an error saying the transducer spline isn’t defined.

I think the spline annotations (center and direction) are created after the simulation algorithm gets executed, so the simulation fails because the required inputs aren’t ready yet. (I also tried reordering the .iws file so that comes before , but it didn’t change anything)

In older versions of ImFusion, it seems there was no issue since the splines were defined directly inside the UltrasoundSimulationHybrid algorithm. For example: ImFusionHybridUSSimulator/USHybridSim_auto_generate_imgs_masks.iws at main · danivelikova/ImFusionHybridUSSimulator · GitHub

I also tried executing multiple iws files sequentially to define splines before, but I was not able to reference a data object that created from the previous script.

reproducing steps:

  1. Load a label map (vertebra.nii.gz)
  2. Ultrasound->Advanced->Hybrid Ultrasound Simulation
  3. Define center and direction splines
  4. Simulate Sweep
  5. Save workspace (ultrasound_simulation.iws) (log_ultrasound_simulation_before_saving.log)
  6. Load workspace ultrasound_simulation.iws (thats where it does not automatically execute simulation) (log_ultrasound_simulation_after_loading_script.log)
  7. add 1 to in the iws file (ultrasound_simulation2.iws)
  8. Load workspace (it gives error saying that: [SyntheticUltrasoundSweepAlgorithm] Transducer center spline is not defined yet.) (log_ultrasound_simulation2_after_loading_script.log)

Example files: ImFusion_Ultrasound_Simulation – Google Drive

Version:
[Base.Framework] Available Plugins: ImFusionAS, ImFusionCT, ImFusionDicom, ImFusionDicomGui, ImFusionImageMath, ImFusionLiveUS, ImFusionML, ImFusionPython, ImFusionReg, ImFusionSeg, ImFusionStream, ImFusionUS, ImFusionVision, TorchPlugin.
ImFusionLib Version 3.13.8 built on 2025-3-7.
ImFusion Suite for Academia - not for clinical use - Version 3.13.8
(C) 2012-2025 ImFusion GmbH, Munich, Germany

System Specs:
Python
[Base.Framework] Available Plugins: ImFusionAS, ImFusionCT, ImFusionDicom, ImFusionDicomGui, ImFusionImageMath, ImFusionLiveUS, ImFusionML, ImFusionPython, ImFusionReg, ImFusionSeg, ImFusionStream, ImFusionUS, ImFusionVision, TorchPlugin.
[GUI.MainWindowBase] OpenGL: 4.1 Metal - 89.4
[GUI.MainWindowBase] Vendor: Apple
[GUI.MainWindowBase] GPU: Apple M1
[GUI.MainWindowBase] Memory: 10.67 GB of 16 GB available

2

Hi Burak,

Thanks for the detailed description of the problem and the steps to reproduce it!
I was able to reproduce the problem and we have a fix for it in progress.
So this will be fixed with the 4.1 release that will arrive in a week or two.

For when you have that new version, may I suggest you start with the simulation parametrization from the demo workspaces? HybridSimulation Demo workspace
You can simply open the workspace and copy the configuration:

image
image792×649 21.6 KB

3

Hi Rüdiger,

Thanks for the update. I’m looking forward to the new release and will check the demo workspace.

I also have a question too, not directly connected to this one.
Is it possible to automate the following action within the workspace file?

Ultrasound Sweep (right click) → Ultrasound → Sweep Properties → Registration → Align to Main Axis

Since this isn’t count as an algorithm, saving the workspace seems to store only the result, not the method itself. Is there a way to include this operation as part of the pipeline?

Thanks again.

4

Hi Burak,

You can find the new academic ImFusion SDK installers (4.1) here:
Ubuntu 24.04
Windows
macOS-ARM

To your question around “Align to Main Axis”: Unfortunately this is not possible right now. But you could either apply it once manually and then save the sweep again (if you only have a limited number of sweeps), or you can replicate the functionality of the button with the follwoing C++:

UltrasoundSweep sweep = …;

mat4 mat;
vec3 extent;
sweep->localBoundingBox(true, true, mat, extent);
mat4 newRegistration = mat.inverse() * sweep->tracking()->registration();
sweep->setRegistration(newRegistration);

I hope that helps.

5

Hi again,

Thanks for the new release and your answer. I tried the new release, and automatically execution works, but I think there is a problem with linear frame geometry. When I simulate a convex sweep, the output looks okey, but when I simulate with linear sweep, the output is just empty and black.

I also tried with the this demo workspace too, but when I create a new linear sweep or change the frame geometry of the sweep to linear, it just computes a black image.

6

I can confirm that!
I added it to our tracker.

As a workaround, you can simulate with a Convex geometry with very low “angle” in the “Ultrasound Sweep Definition”:

image
image631×450 11.8 KB