Cases
Data-transfer from CT into CAD - Investigation of Accuracy
Kozak J., Stern A., Aesculap AG& Co. KG, Tuttlingen, Germany
The accuracy of transfering computer-tomographical (CT) data into a computer-aided-design (CAD) system was investigated. A test phantom was used bearing markers with Reference Points. The markers were reconstructed out of CT-data of the test phantom and the Reference Points were determined. Subsequently the spatial distance between reconstructed and corresponding real Reference Points was calculated. With the procedure used, the mean spatial distance was found to be 0.17 mm (MV) +/- 0.07 mm (SD) which is considered to be sufficiently accurate for the intended use in the development of implants.
Introduction
In the development of endoprostheses anatomical data plays an important role for design and dimensioning of the implants. One way to obtain such data is the use of x-rays however the information that is obtained is limited due to it's 2-dimensional nature. Using 3-dimensional data is more promising as it provides a complete geometrical description of the respective bone geometry.
As the data formats of data-acquisition (e.g. CT) and data-processing (e.g. CAD) are distinct it is necessary to convert the data. This data transfer has to be accurate in terms of anatomical information and therefore needs to be investigated. In this study the transfer of CT-data (DICOM) into CAD (Catia) was examined using a commercially available software package and IGES as an intermediate format.
Material and Method
The idea in this study is the reconstruction of markers from CT-scans and the comparison between the positions of Reference Points on these markers with their corresponding real positions.
Therefore a test phantom was used which is made of a plastic skull with 11 markers attached (1). These markers are made of titanium and show a clear contrast to the plastic skull on the CT-Images. Each marker is shaped cylindrically with a conical bore at one end. The point of this conical bore was designed to be identical to the center of gravity of the marker and was used as a Reference Point. The test phantom was scanned in a Picker CT-Scanner (Pixelsize= 0.43 mm, Slice Distance= 1.5 mm, Gantry Tilt = 0) and the CT-data was processed with a commercial software package (Mimics/ MedCAD 7.2, Materialise, Belgium).
A threshold range of 1345-3071 HU was applied to detect the contours of the markers and polylines were derived from these contours. Afterwards the polylines (contour lines) were exported with the IGES-export option of Mimics/ MedCAD (Filter Value=1) and provided the input data for the CAD-system (Catia V4.1). Therefore each marker is built-up of a stack of planar contour lines, where the contour lines are spaced 1.5 mm apart.
In Catia each contour-line was used to create a prism with a thickness of 1.5 mm with the contour line lying in the middle of the prism. At last the reconstruction was finished by assembling each marker out of the corresponding prisms. In the course of this one to two small gaps per Marker 0.070 mm wide evolved.
The centers of gravity of the markers then could be analyzed and so the reconstructed coordinates of the Reference Points could be determined.
The real coordinates of the Reference Points were obtained by measuring the center of gravity (point of conical bore) with a 3-dimensional measuring device (UMM 850, Zeiss, Oberkochen, Germany)
For matching the real coordinates with the reconstructed coordinates a least-square best-fit algorithm was used (1). This algorithm aligns the coordinate systems of measurement device and CAD-system by minimizing the distance between correlating points. The result is the absolute spatial distance between these points.
Result
The mean spatial distance between the real and reconstructed positions of the Reference Points was found to be 0.17 mm (MV)+/- 0.07 mm (SD).
Discussion
The result shows a very good correlation between the positions of the reconstructed points and the measured points.
The gaps that occurred can be neglected due to their small dimension in relation to pixel size and slice thickness.
In conclusion the derivation and transfer of contours from CT-data into CAD as described above using Mimics/ MedCAD can be considered reliable and sufficiently accurate for the intended use in the development of implants.
References
(1) Kozak et al., Semiautomated Registration Using New Markers for Assessing the Accuracy of a Navigation System, Computer Aided Surgery 7:11-24 (2002)