[Insight-users] Centerline

[Dan Mueller]

Hi,

If you are fine with manually selecting start and end points on the
path, you might be interested in Fast Marching minimization:
    http://www.insight-journal.org/browse/publication/213

AFAIK VMTK uses the same underlying approach, but the speed function
is fixed (based on Voronoi diagrams):
    http://www.vmtk.org/

Attached are some results I got using Fast Marching minimal paths,
with a distance map as the speed image, I only spent about 10 minutes
on this so you may be able to come up with a better speed function for
more central paths (note: I had to pad the segmented image because the
gradient descent back tracing does not appreciate centerlines near the
edge of an image).

HTH

Cheers, Dan

On 28 April 2012 07:05, chans at cs.uh.edu <chans at cs.uh.edu> wrote:
> Hi,
>
> I am a newbie to ITK. I have a confocal image of zebrafish
> vasculature(stack.tif, MIP.tif). My aim is to model zebrafish vessel using
> parametric model such as ellipse.
>
> I have a segmented image(segmentedStack.tif), which I have obtained by doing
> median filtering, followed by thresholding.
>
> I wish to extract the centerline for this data and compute radius in tangent
> and normal direction at each centerline point. With this information, I
> would like to fit an ellipse to each centerline point.
>
> My concerns are following
> 1. I tried to extract centerline using 3d thinning, it gave me very noisy
> centerline.
> 2. Is there a way, I can improve segmentation and fill holes in the vessel,
> to get a better centerline.
> 3. May be there is some other method, which would be more suitable for my
> kind of data.
>
> I am really stuck here, any help would be greatly appreciated.
>
> All the data is here http://www2.cs.uh.edu/~chans/image/data/
>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: zebrafish-fast-marching-paths.png
Type: image/png
Size: 35704 bytes
Desc: not available
URL: <http://www.itk.org/pipermail/insight-users/attachments/20120428/9e66851a/attachment.png>