RE: I beg to differ, or Processing scan data into NURBS surfaces. ..

From: Blasch, Larry (
Date: Fri Mar 22 2002 - 18:43:21 EET

Dear Mr.Le Chi Hieu and Listers

I tend to approach all problems from the mechanical Engineers point of view
(that's how I was trained) so forgive me if I misunderstand the problem you
are dealing with.

I understand that the STL file allows you to visualize the scan data in an
easier to understand format, but the scan data that the STL was created from
only had 2D sections in it. The MIMICS software uses a routine to connect
the 2D sections (polyline data) by what I would imagine is a relatively
simple process. My sincere apologies to MIMICS software developers if I make
it sound too simple, but you could:

1) Extract a (uniform number) point set from the polylines of each section.
2) Connect the points in numeric order to the equivalent number point in
each adjacent section.
3) Map every 3 of the resulting connections as triangles.
4) Determine the face vector direction of the triangle.

I'm sure that I am over simplifying the process, but you get my meaning.

The STL was created from the same 2D sections that could be used to create
the NURBS sections. By using the STL vector data to determine inside and
outside of the sections, you could automatically process the NURBS and break
them in the correct place. Once that is accomplished, NURBS surfaces can be
constructed that represent the shape of the inside and outside surface of
the bone structure. You would have editable surfaces that can be split and
mirrored just like the STL file, but would be much smoother than the STL.
(And would be able to be imported into most CAD/CAM programs.)

Talk to the MIMICS people and see what they can do. If there is a market for
this type of data output, I'm sure that they would be happy to work with


Larry Blasch

Lawrence R. Blasch
Design Engineer
CAE Systems Administrator
OPW Fueling Components
P.O. Box 405003
Cincinnati, OH 45240-5003 USA
Voice: (513) 870-3356
Fax: (513) 870-3338

* "Always remember you're unique,*
* just like everyone else." *

-----Original Message-----
From: Mr.Le Chi Hieu []
Sent: Thursday, March 21, 2002 11:42 PM
To: Blasch, Larry
Subject: RE: I beg to differ Re: Stupid STL Question

> By creating a NURBS surface directly from those IGES
> sections (You would probably need to first
> convert the
> lines and curves into NURBS entities,
> You should be able to produce a much cleaner,
> smooth, editable file in any
> competent CAD surfacing program. MIMICS must be
> doing something like this to
> create the "faceted" STL file that it currently
> creates. You don't want the
> facets, so why create them in the first place.

You are right in this modeling original aspect. We did
follow that way in the first attempt. We currently
also use this approach in some applications.However,
for the biomedical modeling, the following
disadvantage exist when we use the IGES bonny contours
as the starting design data:

(i) Mirrored imaging techniques are normally used for
reconstruct the anatomical structures for biomodeling.
For example in the field of cranioplasty, in order to
generate a mirrored intact part, efforts and time are
needed to cut and separating the inner and outer wall
contours. It was very difficult to manipulate these
CAD operations in the interactive way, or complicated
algorithms must be used for automatically solving
these problems. The main cause is that the inner
contour connects with the outer one at slices in the
defect window area (for example, the skull defect). In
addition, geometrical entities are contours
(polylines), which makes difficult for CAD
manipulations. In fact, when we applied the methods
for automatically separating the inner and outer wall
contours, it was not much efficient for modeling the
implant, especially for the case of large defects and
cases in which mirrored imaging techniques could not
be applied.

(ii) The IGES bonny contours file is large, around 20
- 40 MB for the skull (depending on the slice scanning
thickness). That was why the simplification of the
geometry was normally applied to reduce the large
amount of data. This is another reason, which leads to
difficulty for manipulating the design data (The STL
is also a large file but we can view in 3D), and a
powerful PC is required for the design process.

(iii) In the case we would like to generate bonny
contours in the intended or to avoid the unexpected CT
scanning direction (patient's anatomy is not
symmetrical or the patient is not scanned in a neutral
position), we cannot change the slicing direction of
contours because it depends on the scanning direction
of the CT data.

That can be solved by using STL files outputted from
MIMICS as the starting design data, and using NURB
based modeling method for biomodeling.

> Don't assume that the only way to get a 3D model is
> to start with an STL
> file from MIMICS.
> Although I don't suggest that it would be the best
> way to achieve the type of NURBS surface
> file that you are interested in.

The reason for using STL files as the starting design
data is as follows:

(i) It is convenient for preoperative planning as well
as checking the implant design in the same MIP package
used; and therefore the better communication between
the designer and surgeon is established (this is very
important in the field of biomedical engineering,
working with surgeons is not always smooth as
engineers have done). Because handling the solid
entity is easier compared to handling contours. In
addition, the mirroring axis of the anatomical
structures is quite easily determined when
manipulating STL skull models in Rapid Prototyping
packages such as Magics RP.

(ii) By changing the slicing direction, we can
"reconstruct" bonny contours in the direction we want.
That creates more options for the modeling process of
the anatomical structures and implant design. This
cannot be obtained if we use IGES contours as the
starting design data, although it is the best way for
getting more accuracy for 3D reconstruction. In
addition, in the field of biomodeling, the "accuracy"
constraints in some applications can be more flexible
than the ones in the field of engineering, some times
0.5 mm can be acceptable.

Le Chi Hieu

Do You Yahoo!?
Yahoo! Movies - coverage of the 74th Academy Awards®

For more information about the rp-ml, see

This archive was generated by hypermail 2.1.4 : Tue Jan 21 2003 - 20:13:38 EET