The animation below shows the evolution of the level set surface from its initial state to the final segmentation of the thrombus. Notice how the lumen segmentation is used as the initial level set in this stage.
The second step is to segment the outer aortic wall around the stent graft, using what we found out about the location of the aorta in the previous step. The segmentation was performed in 3D using the level set method.
 |
 |
The first picture again shows the unfiltered slices of the initial volume. In the right image, the dilated aorta can be seen as a big round shape around the the bright lumen. |
 |
 |
First, we blur the image we got from the fast marching filter in the first step using a gaussian filter (variance 0.1). We then create a binary image of it in which all intensity values greater than 0 are thresholded to 255. The reason we do this is that we don't want edges inside the region we are going to segment, and we can get rid of the edge lines of the stent graft by covering those bright pixels with pixels of intensity similar to those in the surrounding region. |
 |
 |
In this image, the highest pixel values have been thesholded off. Reducing the dynamic range of the image this way affects the eventual potential image in that the edges delineating the outer aortic wall will be more discernable. |
 |
 |
Gaussian filtered image. (variance = 1.0). |
 |
 |
Median filtered image (radius = 3). |
 |
 |
Optimization of dynamic range. This image is converted to a potential image for the level set filter. |
 |
 |
Optimization of dynamic range. |
 |
 |
Level set filtered image. |
 |
 |
This is the postprocessed median filtered image, which is used in the final model. |
The final rendered model:
