[Insight-users] Segmentation of microscopic vascular images

[Gib Bogle]

Hi,

Our stack of vascular images was obtained by fluorescent staining and 
multi-photon microscopy (on a mouse lymph node).  The significant difference 
from the CT and MRI data that are the usual starting point is that in our case 
only the inside surface of the vessels is visible.  In other words, the stack 
contains empty tubes with very thin walls - the interior of the tubes having the 
same intensity as the exterior.  This means that the usual level set approach to 
segmentation will not work directly.  If the tubes had no holes the "snake 
evolution" method of ITK-snap would work fine, but variability in stain 
intensity gives rise to many leaks.  What is needed is a "tube healing" step in 
which holes in the tube wall are patched.

I have some ideas about how to approach this, but since it's quite tricky and 
likely to be time-consuming I thought I should check first to see if anyone has 
done something similar, or if there are existing methods in ITK that could be 
employed.  What's needed is a kind of focussed dilation procedure.  The edge of 
a hole needs to grow in the plane tangential to the nearby tube wall.

Any suggestions?

Thanks
Gib