FrangiFilter

In this repository you can find an implementation of Frangi method for computing a probability of vessel appearance from 3D CT DICOM image.

The main steps of app execution:

• reading 3D CT image in DICOM format by ITK (Insight ToolKit) librar
• image noise is smoothed by gaussian filter from ITK
• in parallel loop (OpenMP) it initializes:
  1. a central voxel
  2. its backward, forward and central nearest neighbours (26 ones)
  3. data structures for Hessian matrix (that corresponds the central voxel), eigenvector matrix and eigenvalue vector.
• solving the eigen problem
• computing vesselness function (see Frangi paper), i.e. a probability of that central voxel corresponds to a vessel
• writing an image with vesselness

Here is an image with vesselness. You may notice that hear is also detected.

Links:

• original paper