Description:

This example illustrates the process of grid generation starting from medical images. In this case a model of the circle of Willis was reconstructed from contrast-enhanced MRA images using a region growing algorithm.

First the anatomical 3D images are processed and segmented in order to label the voxels that belong to the desired arteries. This process is illustrated in the next two images where the maximum intensity projection (MIP) of the images before and after segmentation are shown.

After segmentation a geometrical model is constructed by tessellating the images. This model is then cut in order to simulate only a desired portion of the arterial tree. These processes are illustrated in the following two images.

This final model is the used to generate finite element grids for the CFD calculations. The following three images show the surface of the finite element grid from different view-points.

Observations:

• The geometry of the anatomical models reconstructed from medical images may be quite complex. Therefore, unstructured grids are ideally suited for this type of calculations.
• Robust surface gridding algorithms are required in order to avoid having to represent the model with analytical surfaces.
• Surface smoothing techniques are necessary in order to get rid of noise present in the original images and due to the voxelization.
• The major drawback of the segmentation procedure is that it may lead to fake intersection of arteries which are close in space but do not cross each other (this is a common problem to most of the segmentation procedures, unless one does it manually).