Hello,
I am trying to simulate a Ultrasound Sweep using the “Ultrasound simulation Algorithm” hopping to have as good simulation as you’ve managed to do in this paper. But as you can see in the image bellow all I can see is noise… I have well segmented the organ from the DICOM CT image and tried many calibrations for the virtual probe but still can’t see anything can you please help ??
Best,
Sidaty,
Hello,
the hybrid ultrasound simulation requires the input label map to use certain pixel values. In particular, they must match with the values in the “AC.Prop” tab of the simulation controller, so 1 for background and 2 to 17 for the various materials.
This has the unfortunate side effect that the simulation does not work with regular label maps, which typically use 0 for background and 1 for the segmented object.
You can use the “Replace Values” algorithm from the “Image Processing” category to convert your label map to the required format.
For an example dataset which should work with the default simulation settings, see here.
Cheers,
Simon
Hi Simon,
I have changed the background to 1 and the organ to another value between 2 and 17, it’s much better ! But what about the internal structures of the organ(vessels) should they be labeled such as a background ? or as another organ?
Also, how can we label automatically based on the pixel intensity ?
Thank you,
Best,
Hello,
there should be a material preset for vessels in that dropdown list.
For other cases where your data contains some structure not included in that list, you can also use one of the offered material types you do not need, change its acoustic properties to match whatever it is you are missing and then use that entry.
Cheers,
Simon
Hi Simon,
Thank you very much for your help. I have labeled the internal structures manually using the checkbox “Avoid other labels”.
Would have another(maybe the last) question, I would like to reconstruct a 3D vessels volume from the Ultrasound Sweep that I have simulated, Is there a way to do in Imfusion ?
Best,
You can have a look at the Intensity Clustering algorithms in the Segmentation sub-menu.
For instance the standard one does some kind of k-means on the image intensities, you just have to specify a list of initial intensity values and standard deviations, for instance: clustersMeans: -1000 0 1000, clustersStd: 500 100 300
(Don’t forget to either set the Label map modality to the output to visualize it correctly)
There is also another version using the image histogram that you can try out.
