Re: Minimum thickness of CT/MRI data

From: Andy Christensen (chefandy@ix.netcom.com)
Date: Fri Jul 11 1997 - 00:46:34 EEST


On July 10th, Junghoon Hur wrote:

> I'm working and researching on the medical RP model using CT or MRI data.
> Generally, we take a set of CT/MRI slice data with a thickness of 5 mm in
> our medical center.
> I wonder if the number of CT/MRI data is enough to reconstruct the surface
> model or STL data.
> You know, because of the maximum permissible exposure to radiation or
> something, we cannot take a thinner slice of MRI/CT data for a patient.
> I need your advice or experiences of surface/STL reconstruction job using
> MRI/CT data.

And on July 10th Bob Yancey replied:

> Our experience has been that with smoothly varying objects (such as with
> the human anatomy), the accuracy of the surface/STL file is more dependent
> on the slice spacing rather than the slice thickness. A CT data set with a
> 5 mm slice thickness and a slice spacing of 1 mm is as good or better than
> a data set with 1 mm slice thickness and 1 mm slice spacing. The reason
> that it might be better is that your will have a better signal-to-noise
> ratio in the image with a 5 mm slice thickness versus a 1 mm slice
> thickness. This may not solve your limitations with radiation exposure,
> however, since both data sets will contain the same number of slices and
> therefore the same radiation exposure to the patient. In general, the
> accuracy of your surface/STL model will only be as good as the accuracy of
> your data. If your slices are spaced every 5 mm, then your surface/STL
> model will only be acccurate to within 5 mm in the slice direction.

Dear Junghoon and Bob,

I think that the basics of this question were answered by Mr. Yancey.
We would, however, like to add our two cents...

For obvious reasons, a thinner slice thickness/spacing would give you
greater resolution in the model. Since the layer of an rp model is
typically in the 5-10mils range (0.125mm to 0.250mm) additional layers
must always be interpolated between the CT/MRI slice data. Specialized
software (eg: Materialise) handles this task well.

In general, any thickness/spacing can be used, but larger increments
will not provide the detail of small anatomical features. For MRI
scans, large (5mm+) slices are used because it takes TIME to gather all
the data in a slice. By keeping the number of slices down, you keep
overall scan time down, and, therefore minimize the chances of patient
movement during the scan. As of yet, there are no known side effects to
long MRI scans. 1mm slicing/spacing will produce a great looking model
if you can get the patient to stay still.

For CT scans there is a definite side effect: x-ray radiation exposure.
CT's are normally only recommended when the risk of gathering the data
is balanced by the benefits of having the scan done (ie: diagnosis, SL
model creation, etc.). The balance that seems to work best ia a 1mm
slice/spacing for anatomy with plenty of detail (eg: craniofacial area)
and 3mm slice/spacing for areas that are smooth and continuous like long
bones.

There is still quite a bit of art to creating anatomical models from
scans utilizing rapid prototyping. A skilled biomedical technician is a
must for creating accurate models!

In reply to Mr. Yancey: I would have to disagree with the statement
about a 5mm slices spaced 1mm apart. This would produce a
smoother-looking model, but in doing so you would introduce small
anatomy/model discrepancies such as pseudo-foramina and loss of details
less than 5mm in vertical size. This would show up in areas of very
thin bone and finer features such as in the craniofacial region.

Just our thoughts!

Todd Magnusen & Andy Christensen
 
----------------------------------------------------------
Medical Modeling Corporation
17301 W. Colfax Ave., Ste. 300
Golden, CO 80401 USA
888/273-5344
303/273-5344
303/277-9472 fax
chefandy@ix.netcom.com

"SLA Anatomical Models . . . The Ultimate Analysis Tool."
----------------------------------------------------------



This archive was generated by hypermail 2.1.2 : Tue Jun 05 2001 - 22:39:50 EEST