Neumann boundary conditions for 3-D image Volume, need this to continue on thesis!

Hi All,

I am trying to implement Direction dependent regularized diffusion as described here: http://www.ncbi.nlm.nih.gov/pubmed/21764627

To do this i must be able to calculate derivatives only inside and outside an object. I work with MRI volumes which and need a 3-D variant to calculate the the derivative only inside or outside an object using boundary conditions. My current method was calculating derivatives on the image as a whole and multiply with a mask but this is completely wrong off course. Can anybody help me out? I need this really bad for my master thesis. I have included my current code now. I tried to extend the values along the normals of my mask, which are computed with a level set function.Dt is a dipiamge function to create a distance map. The m file can be run with phantom sliding gauss.

What i want is to extend the mask along it's normals and sample that back onto the image and thus creating neumann boundary conditions on the edge pixels of the image.