View Full Version : summary CT-Scan

Arno Grunendahl
02-06-2003, 06:09 PM
Thanks to everyone for the answers.



there are many papers published regarding your question. IT is not as
as you thought to build up a useful model directly from CT images in

If you know the cortical thickness, you can use a cad program like
Unigraphics to create a cortical shell with a universal thickness.


>As far as I understood the method, the meshing tools build a tetraedric

>mesh on the surface and somehow "deepen" the mesh to a volumetric

This is not entirely accurate. Volume meshing of a closed triangular
surface mesh is one way to create a 3-D model. There are others which
may be better for your particular situation. The "best" method really
depends on a number of factors -- including the shape of the bone,
the spatial resolution of the CT scan, the preferred type and number
of elements desired, and the method intended to assign material

For example, if the topology of your geometry is generally similar
from one CT image to the next, it may be relatively easy to generate
a model by stacking up sequential contours and generating a volume
based on these contours, which can then be meshed (perhaps even with
hexahedral elements). If the topology changes more dramatically
(e.g., due to branching or the presence of interior voids), then the
problem is more difficult and may require a more complicated approach.

>My question is, how can I distinguish between e.g. cortical and
>cancellous bone in a vertebrae with this precedure?

It would be helpful to have a bit more information. Do you hope to
distinguish cortical bone from cancellous bone in a geometric sense
-- i.e., having a mesh in which elements are physically either a
member of a "cortical bone" volume or a "cancellous bone" volume? As
described by an earlier poster, this process of identifying two
tissues into separate volumes is known as segmentation. Or do you
hope to differentiate between the types of bone by assigning
site-specific materials properties (based on CT density) to the
elements of a non-segmented mesh?

During my time in graduate school at Tulane University, several
graduate students developed procedures to generate FE models from CT
scans. Our basic approach was to generate the 3-D geometry from
sequential contours, mesh it, and then assign material properties to
the elements based upon CT density. (Details available at
Sometimes we simply used a threshold value of CT density to segment
the elements into cortical and cancellous bone, which were assigned
"average" material properties for those tissues. On occasion, we used
custom Fortran code to develop a completely heterogeneous FE mesh, in
which each element was assigned a unique value for Young's modulus
which was derived from the CT density.

As evidenced by the many responses to this thread, the generation of
FE models from CT scans can be a complex task. Good luck!

John C. Coleman, Ph.D.

Hi Arno,

I work pelvic CT data and create both cortical and
cancellous surfaces using a program called SEGMENT.
The process is tedoious but works well. Basically the
2D raw CT surfaces (no headers in the file) are
imported into the program on a slice by slice basis.
The mouse is then used to draw dots around the surface
of cortex. Next a spline is created using the
user-inputted dots. The splines are stacked upon one
another using NUAGES to generate polygonal surfaces
that represent the cortical and cancellous bones.

As I mentioned before the process is tedious. There
are other commercial segmentation programs such as
MIMICS that automate the process. However these
programs have a hard time computing in the areas where
the cortex is thin or non-existent. Often you will
have to make an educated guess at the boundary of the
cortical and cancellous bone and thus the automatic
programs will create voids in these areas whereas the
manual process will not.

The program SEGMENT was done as part of an
individual's Ph.D. thesis. I would be happy to let
you know how and where to get it. I hope this helps.


I believe that the Simpleware software (www.simpleware.co.uk)can
between cortical and cancellous bone by according mechanical properties
the mesh elements depending on the measured density from the CT data.

I've not used it myself, but remember chatting to the developers at a
meeting last year sometime....


> My question is, how can I distinguish between e.g. cortical and
> cancellous bone in a vertebrae with this precedure?

The process of separating regions within a CT scan dataset (usually
based on
Hounsfield number) is known as
segmentation, and applies to any separation of adjacent tissues based on
density, not just cancellous and cortical

As noted by some of the responders to the original posting, a software
package named
Mimics is well suited to the task
of taking you from CT data to meshable solid. Mimics includes

Perhaps a search of the web, using a phrase such as "CT scan
segmentation" will lead
you to other software that does
this. I'm not ware of any freeware or lower cost software that does
this. Perhaps
one of the other BIOMCH-L members
would care to chime in...


Danny L. Levine, Ph.D., P.E.
Zimmer, Inc. - Warsaw, Indiana, USA

To unsubscribe send SIGNOFF BIOMCH-L to LISTSERV@nic.surfnet.nl
For information and archives: http://isb.ri.ccf.org/biomch-l