Modeling chambers of the heart

[Justin Tso]

Hi, the user guide documents how to model blood vessels, but how would you go about modeling the atria and ventricles, as the shape is very different? Namely, what path would you choose (or how would you choose that path) that would enable you to create a contour group for the shapes of the atria and ventricles?

[David Parker]

Hi Justin,

Modeling anatomical structures that are not tubular using paths could be a bit tricky. You could try the SV MITKSegmentation Tool (right click on Tools in the top SV menu bar). You can find some documentation here http://docs.mitk.org/2016.11/org_mitk_v ... ation.html. There is a section there on Interpolation which I think could be useful for segmenting the heart chambers.

Cheers,
Dave

[Justin Tso]

Hi, the MITKSegmentation tool appears to work well, but I appear to be running into problems copying and pasting the 3d segmentation into the segmentations folder to form the model: when I try to form the model, the option to select the 3D segmentation doesn't appear to be selected (I suspect this might be because my computer is underpowered, as when I try to open the 3D segmentation in the folder, the program crashes). Alternatively, for the left ventricle, I've tried a different approach of plotting a straight line path from the mitral valve to the apex and making cross sections down this axis. It works well, but it runs into the problem of there being a small hole at the end of the segmentation, since all contour sets appear to be tubular. Is there a way to make this hole solid, or in other words, turn one of the caps of the model into a wall so that there's only an inlet, and the outlet is obtained by the union with another contour group? Thanks again for your help, I really appreciate it.

[David Parker]

Hi Justin,

I didn't quite understand what 3D Segmentation problem you are having.

I created a model using 3D Region Growing and then select Create a smoothed polygon model which writes the polygonal surface to a .vtp file. This file can then be read in from the Model tool although it will need to be preprocessed somehow to create planar inlet/outlet faces needed by SV.

Using path lines to segment an image is probably the best way right now since the surface it creates will be compatible with the way SV creates models. When creating a model from the segmentations the Model tool will try to fill any holes with a planar surface and set the type to cap. You can then change the type of the cap to wall.

We are also interested in creating heart models. We have students in our Lab working on mitral valve simulation, are collaborating with the Kuhl Lab (Stanford) on the Living Heart Model project and with the Shadden Lab (Berkeley) who are heart CFD and machine learning for heart segmentation.

Cheers,
Dave

[Justin Tso]

Hi, regarding the MITK segmentation, I'm able to create the segmentation using interpolation, but when I try to create a model, I don't see the options for "Create a smoothed polygon model," or "Create a polygon model." Also, I've found that these segmentations are never saved, even as other components of the project done afterwards are. Do you know what might be wrong? Attached is an image of what I'm seeing (no further options appear when right clicking any of the other nodes as well).

[David Parker]

Hi Justin,

From the picture you attached it looks like you are doing a 2D segmentation. Here's an example of using Fast Marching 3D to create a 3D segmentation. Right clicking on the seg1 segmentation brings up the menu allowing you to create models.

When you save your project SV writes .s3d and .vtp files to your projects Segmentations directory, the VTK .vtp file contains the segmentation geometry.

Cheers,
Dave

[Justin Tso]

Hi, once again, thank you for your advice, it's been extremely helpful. Earlier in this thread, you mentioned that the model created using MITK segmentation would have to be pre-processed somehow in order to create the caps that SV uses. Do you have an idea of how this pre-processing would take place, or using what tool? I've found that SV is pretty robust for most of the features of CFD but can be a little clunky during the segmentation phase, which is understandable as it's primarily used for vessels. If there's some way I could use another tool to segment, pre-process that segmentation/model, and import it into SV for the rest of the steps, that would be ideal.

[David Parker]

Hi Justin,

SV is not ideal for creating heart models. You could try 3D Slicer to create your models. The model would need to have smooth surfaces and well-defined inlet/outlet geometry that are planar surfaces so SV can classify them as caps. I'm not sure how much functionality 3D Slicer has for smoothing surfaces, you could try Mesh Lab for that.

Cheers,
Dave