Q:Computing Tissue Density from CAT Scan Images
Quote:
> I'm trying to estimate the mass center of the head in 3-D and the head
> moment of inertia in 3-D from CAT scan images. To do this, I need a way
> of relating the CAT scan image intensity (grey levels) to the actual
> tissue density (e.g.
> g/cm^3)
> (1) How do you do this?
> (2) What software is avaliable for doing what I just described in the
> preceding paragraph?
> (3) What is a Haumsfeld unit (spelling uncertain) and how does it
> relate to
> CAT scan image intensity and tissue density?
> Thank you very much for your help
> Steve Lutes
> Center for Balance Disorders Rm. NA315
> Dept of Oto
> Baylor College of Medicine
> One Baylor Plaza
> Houston, TX 77030
> Voice: 713-798-6336
> Fax: 713-798-8658
> Check out GeometryID, a package for
> linear or nonlinear system identification at:
> http://www.bcm.tmc.edu/cfbd/geometryID/
Here is my approach for solving this problem.
First of all you do not need know the density of the tissue
for any given gray scale level. Therefore you do not need to know
anything about Hounsfield units or etc. Let's assume you have 8 bit
gray scale from 0 to 255. Typically you will get larger dynamic
range but let's make it simple for the time being. Divide the
spectrum into subranges which may correspond to some specific tissue
types. For example
40- 80 range fat, call this => density=1
81-125 range soft tissue, call this => density=2
126-180 range bone, call this => density=3
181-225 range dense bone, call this => density=4
226-255 range metallic implant in the skull call this => density=5
I am sure medical people will come up with a better division than mine.
Now stack your CT slices on top of each other to form a 3D collection
of cubes and by using density values in 2D (I assumed that patient
did not move his head at all), determine a density
value for each cube [d(x,y,z)]. Then use the equations for center of
mass and
compute the x,y,z giving you the center coordinates. Since your data is
not continous but discrete, you need to resort to summations instead of
integrations. d(x,y,z) is going to give you a discrete value for each
cube
and it is very convenient for summation. Because you have the density
values
of tissues d(x,y,z) both in numerator and denaminator of the equations,
actual
numbers are not important as long as you can distinguish tissue types
from
each other. Densities will be cancelling each other in any condition.
In any approach, you will be introducing systematic errors. Therefore,
I would not recommend going into pains for trying to come up with a
very accurate representation of tissue density values, center
coordinates.
Good Luck!
--
Feyzi INANC
Iowa State University, Center for Nondestructive Evaluation
Applied Sciences Complex II, 1915 Scholl Road, Ames IA 50011
Ph. (515) 294 9738 Fax. (515) 294 6368
http://www.cnde.iastate.edu
