**Faculty of Technology, Buckinghamshire University

Key words: Data acquisition, image processing, Rapid Prototyping, Prosthetic design, and anatomic structures

Several research institutions and commercial organisations have integrated Computer-aided Design (CAD) and Rapid Prototyping (RP) systems with medical imaging systems to fabricate medical devices or generate 3D hard copy of these objects for use in surgical rehearsal, custom implant design and casting [1-2]. However, working with RP technologies in the medical field differs radically from using them in manufacturing environments. In manufacturing, models are planned and conceived entirely on the computer screen, then converted to physical reality. In bio-medical applications, the objects normally already exist physically. Therefore building medical models essentially involves reverse engineering, starting with acquiring data such as a stack of Computed Tomography (CT) cross-sectional images. Prior to building, this highly complex data needs extensive pre-processing to provide a format that a CAD program can utilise, before transferring to an RP system.

Successful integration of imaging and rapid prototyping technologies therefore depends on the ability to provide special purpose computer graphics software tools for efficient handling and modification of 2D and 3D data. Here we report on the design and implementation of tools for computer assisted RP medical applications. We demonstrate the application of the RP technologies to general medical practice.

Before fabricating medical devices from medical data one has to ask the question "How can scanned data and processing technology be linked RP technologies to obtain the physical models?" The data has to undergo a number processes otherwise known as the medical modelling process. The medical modelling process is broadly split into three areas :- data acquisition, image processing and model production.

In medical imaging, the two most common systems used in acquiring detailed anatomical information are Computed Tomography (CT), and Magnetic Resonance Imaging (MRI). Other systems used include Ultrasound System, Mammography and X-ray. The key feature of these imaging technologies is their ability to provide detailed information about the anatomical structure and abnormalities.

CT uses radiation in the form of a highly collimated X-ray fan beam to slice a two-dimensional image or slice plane. Standard CT scanners achieve a resolution of 512 X 512 elements within a layer.

On the other hand, MRI, images are obtained based on different tissue characteristics by varying the number and sequence of pulsed radio frequency fields in order to take advantage of magnetic relaxation properties of the tissues. MRI differs from CT in at least two key aspects: (1) MRI measures the density of a specific nucleus, and (2) the MRI measurement system is volumetric (interrogation of the entire body, within the measurement volume, is done all at one time).

CT and MRI represent the finest resolution capability available in diagnostic systems achieving volumetric resolutions. During the scanning process, the patient is stepped through the measurement plane 2-3mm at a time. The information from each plane can be put together to provide a volumetric image of the structure as well as the size and location of anatomical structures. The scanned model becomes a virtual volume that resides in a computer, representing the real volumes of the patient’s bone(s). The virtual volume is displayed on-screen by reformatting the data to create orthographic projections, or by a creating a pseudo 3D representation using surface-rendering algorithms [3-4].

When a series of CT images is reassembled to illustrate a 3D presentation of an anatomic structure, the medical practitioner or prosthetic designer can use this information directly and the overall shape of body structures is more clearly understood or visualised.

However, visualisation requires good visualisation software and so a number of dedicated software packages have been developed to enhance the visualisation of such 3D computer models and enable the surgeon to grasp the particular details of individual cases. Some of these software packages include Analyze biomedical image processing package [5], Surgicad Template from Surgicad Corp. and Intergraph corp.[6] and Mimics from Materialise Software Company and Katholieke University, Leueven, Division PMA [7].

These software packages take anatomical data from CT and MRI scans and create computer models of anatomical structures. A user can modify the image by defining various tissue densities for display. This allows separation of data of interest from the general information available from the scanner. By combining the data generated with a traditional CAD system, design of new parts can be undertaken by comparison with the reconstructed 3-D anatomical shape. When segmentation and visualisation is completed the data can be translated into instructions for manufacture of parts often by RP. The de facto standard interface from CAD to RP is the Standard Triangulation Language (STL), though other transfer formats such as the Initial Graphic Exchange Specification (IGES), Standard for the Exchange of Product Model Data (STEP), Common Layer Interface (CLI) and Virtual Reality Modelling Language (VRML) are also possible [8].

RP is a relatively new technique that was invented over a decade ago to rapidly produce solid 3-D objects of complex shapes directly from CAD files. RP constructs solid physical models from 3D computer data by the addition of layers of material [8,9]. These techniques provide ways for making a variety of complex shaped parts which are difficult, costly or sometimes impossible to make by conventional methods of material removal. However, milling techniques are still important in the production of "less complex" models [11].

A host of layer-manufacturing processes exist commercially in the market today and many other processes are under development. The selection of a particular process will depend on the medical model application. Some of the most commonly available systems are: Fused Deposition Modelling (FDM) [12], Stereolithography (SLA) [13], Selective Laser Sintering (SLS) [14], Sanders Prototyping Technology and Z Corporation Fabrication Machine.

Given the visualisation provided by sophisticated software packages, the fabrication of physical models may at first seem superfluous; however, this is not the case, as the depiction of a 3D volume on a 2D screen does not provide surgeons with a complete understanding of the patient’s anatomy. The relationship between the patient and what is on the screen is not intuitive; surgeons must learn to interpret the visual information, in order to reconstruct mentally the 3D geometry. To that end, head-mounted displays, stereoscopic glasses and holograms are beginning to complement the 2D screen in an effort to provide more realistic representation of 3D volumes models [15, 16]. The use of 2D or pseudo-3D electronics display systems can be augmented by overlaying images, graphics, text, and so on to aid the user as training guidance. However, current implementations do not feel natural or intuitive. Surgeons must learn to interpret screen displays, particularly when extracting 3D information from them.

It is therefore, clear that several visualisation issues are addressed but not yet resolved by virtual computer models:

• 2D screen displays do not provide an intuitive representation of 3D geometry.
• Unusual or deformed bone geometry may be hard to comprehend on-screen.
• The integration of multiple bone fragments is hard to visualise
• Planning of complex 3D manipulation on the basis of 2D images is difficult.

This provides the motivation for the construction of physical models of bones. The use of rapid prototyping for 3D physical models simplifies communication between surgeons, nurses and even immediate family of the patients. For example the extent of the damage to the bone can be demonstrated to the patients in a clear manner, so that they may evaluate the reconstruction options presented to them.

3D models are useful for diagnostic and treatment planning, more so, in cases of complex surgical procedures [17]. This reduces the risk to the patient owing to the shortened time of surgical procedures and is less expensive than the alternatives. The model accuracy and material properties provide the basis for simulating and evaluating alternative scenarios, so that optimisation can be made prior to the actual surgery.

A physical model derived from CT or MRI data can be held and felt, offering surgeons a direct, intuitive understanding of complex anatomical details which cannot be obtained from imaging on the screen. Allowing surgical dry runs as well as marking out the vital areas to be avoided and predicting complications that might arise. Hence it increases the surgeon’s confidence in the operation, and reduces the search time for the correct entry of surgical tools. This will allow a surgeon, from the outset, to know what to expect when a certain surgical route is adopted. Thus it reduces the duration of the procedure and greatly reduces the risk of infection and the problems of uncertainty.

Complex and sensitive surgery requires extensive planning. In a surgery as delicate and complex as a cranial osteotomy, for example, the displacement of bone segments can be more accurately evaluated. Surgical instruments identical to those used in the actual procedure can be employed on the models to determine the most conservative strategy. The model can be referred to during an operation for guidance during mock training procedures or for academic teaching of surgeons and young doctors. Three-dimensional bio-models have greatly enhanced the design and fabrication of medical implants. The RP models can also used as a 'negative' on which the implant is formed, or a 'master' from which the implant is duplicated.

In order to evaluate the application of rapid prototyping technology in medical situation, a skull of a child diagnosed as having craniosyntosis and a section of a pelvis were chosen. All scans were transverse. Image segmentation was performed using tools from the Materialise software package and the DeskArtes CAD modelling software was used to design the femur implants. The Materialise package has two modules: MIMICS and CTM suites.

Mimics is a software suite that performs the segmentation of the anatomy through sophisticated three-dimensional selection and editing tools. The method adopted for visualisation is the conversion of 2D image slice data, as grey value images. The resolution can vary from 0.2 to 1 mm. The program also generates high-resolution 3D renderings in different colours directly from the slice information, as shown in Figure 1. Contrast enhancement can be carried out interactively to improve the model. The segmentation mask can be displayed in a different colour on top of the image. After visualisation, the data can be interfaced to CTM [7].

CTM is a software suite that interpolates the medical slice in very thin layers, and interfaces directly with most RP systems. Because of this direct interface and the use of higher-order interpolation mathematical algorithms such as Bilinear and C-Spline functions, it produces very accurate models in a very short time. As the pixel size of the input images can vary from only 0.2 mm to more than 1 mm, a resolution enhancement technique is necessary when creating the RP models so as to minimise the effect of stair-stepping, and to retain the natural curvature of the surface. Two techniques are used to increase the resolution of the contour [7].

1. Bilinear interpolation which increases the in-plate resolution of the contours by a factor of up to 100.
2. Cubic spline convolution, which decreases the effective slice thickness of the data set. Slice thickness of 1 to 3 mm are very common for the input images, whereas the slice thickness used on the RP machines is normally less than 0.25 mm.

Before performing segmentation of the anatomy, the data is loaded into Mimics by a front-end data input software, CT-convert. This software translates the data from the CT scanner into CT-Modeller’s own proprietary data format. The data in Figure 2 shows data in mimics own proprietary format. The image is processed by use of threshold value to differentiate regions of interest. The bone can be separated from soft tissue by setting a fixed threshold value. Typically, algorithms evaluate neighbouring voxels to determine whether the differences in their intensities are within specific threshold values. An area of interest is selected by defining sets of voxels whose intensity is above the selected thresholds. Culling of the ‘unnecessary’ parts or unwanted noise is achieved with sophisticated dimensional selection and editing tools.

Once visualised, the file was interfaced with CTM. CTM uses a higher-order interpolation mathematical algorithm to interpolate the medical slices in every layer generating vertices and triangular surfaces from the 3D models. CTM generates a high-resolution contour model suitable for rapid prototyping machine. The format is specific and dependent on the RP system. In this case the STL format was chosen.

There are two options of interfacing CTM to RP system as illustrated in Figure 3. Approach A interfaces CTM directly to the RP system. In approach B the STL is first imported into a CAD environment, which allows for further manipulation, triangulation correction and designs of medical devices.

Two types of RP system that were considered for production of models were the Sanders ModelMaker machine and ZCorp machine. This choice was based on availability of the machines to the University. The Sanders’ ModelMaker is a high-precision 3D modelling systems that build parts from a wax material. It is suitable for small and most detailed, intricate, and complex parts. While the Z402 3D Printer is a design-based RP system that builds by binding together a fine powder. It provides a much shorter build time and is its models are good as a visualisation tools.

Pictures in Figures 4 and 5 show prototype models from RP systems. Figure 4 is a femur model fabricated using the Sander Prototype machine. Staircasing effect is highly evidenced in this model and this could be explained by the fact large layer thickness required to reduce the time for such a large model. The skull shown in Figure 5 was made by the ZCorp machine and has a rough surface and lacks tiny bone details particularly around the maxilla region. These details were poorly connected to the main model and had to be removed. For instance on the hip it was observed that the femur came away from the acetabalum due to weaknesses in the joint between these parts. These observations serve to emphasis the fact that the Z-Corp. machine is mainly a design based system intended for visualisation purposes.

The opportunity to hold the model in the hand and view it from various angles in a natural fashion offers immediate ease of diagnosis and treatment in medicine. Having seen the accuracy of these models, which though not very high, was quite enough for orthopaedic application, we performed an implant design.

Biomechanical design work is closely related to sculptural work. The human body. does not have sharp corners or edges, thus it was necessary to select CAD software that is versatile enough to give the model irregular shape. The wide variety of modelling capability offered by the DeskArtes CAD system make it suitable for implant design. This software accepts data in neutral formats such as STL and gave us the opportunity to interface CTM to CAD via approach B illustrated in Figure 3. Segmented data was translated into STL file format and imported into the CAD environment. Including the STL file in DeskArtes environment also offered another advantage. It allows the file to be checked and repaired if the conversion software had made any omissions. The STL file was checked for any defects using DeskArtes and it was found to be error free. The CAD environment also allows for both the surgeon and the designer to determine critical dimension and mass properties from the CAD model to aid surgeons in their assessment.

Figure 6, shows the process of designing a hip implant, carried out using DeskArtes software. The model was designed to the nearest/exact shape of the imported femur STL file. Designing basically is a function of experience and skill of using the DeskArtes CAD software. The 3D geometrical shape was then tessellated, and sent to for fabrication using their Z402 machine. The result is shown Figure 7, which shows a picture of a fabricated hip implant model.

Depending on the intended use of the model it can act as either a master pattern for the purposes of rubber moulding, or can be sterilised for assistance in operating theatre.

One limitation of the procedure is the fact that there is a trade off between the file size and the tolerances of the triangles, more so with visualisation software. Taking a higher tolerance leads to a bigger file, slowing down the computer. While, lower tolerance reduces the overall file size but leads to a simplified irregular model.

From the physical models fabricated by rapid prototyping system it is clear that this technology has a lot to offer to the medical field. The ability of the systems to identify precisely the relative location, shape and thickness of the damage of the anatomy. For instance the skull in Figure 5 shows a defect on the left side of the skull. Such information assists the surgeon to make the right diagnosis and correct decision before embarking on any operation.

Realistically, RP technology can make significant impact in the field of biomedical engineering and surgery. Physical models enable correct identification of bone abnormality, intuitive understanding of the anatomical issues for a surgeon, implant designers and patients as well. A precise RP model facilitates the pre-operative planning of am optimal surgical approach and enables selection of correct or appropriate implants.

The reliability and the accuracy of an RP model in surgical application allow surgeons to rehearse the re-alignment of bones or fitting of implants on the RP models prior to operating the patient, to evaluate and gain confidence in, the planned approach. Surgical procedures continue to be more effective day by day with reduced risk and expense to both the patient and the hospital.

This integration of technologies such as medical imaging, CAD and RP is important in the medical field. With wider practice by smaller institution, it is possible to reach small, isolated community hospitals. This could help minimise the problem of long waiting list and congestion in ‘big’ hospitals by reducing referral cases.

Reference:

1. Ashley, S., "Rapid Prototyping For Artificial Body Parts", Mechanical Engineering, 155(5), 50-53, May 1993.
2. Mcgurk, M., Aimis, A. A., Potamianos, P., and Goodger, N.M., "Rapid Prototyping Techniques For Anatomical Modelling In Medicine", Ann. Royal Coll. Surgery Engl. 79, 167-174, 1997.
3. Pommert, J. K. et al., "Three Dimensional Imaging In Medicine: Method And Applications", In Computer Integrated Surgery (Eds R.H. Taylor et al.), Ch. 9, 155-174, 1996.
4. Udupa, J. K., and Goncalves, R. J., "Imaging Transforms For Volume Visualisation", In Computer Integrated Surgery (Eds R. H. Taylor et al.), Ch. 3, 33-57, 1996.
5. Puttre, M., "Replacing Parts On Nature’s Machines", Mechanical Engineering, pp. 58-61, May 1993.
6. Mankovich, N. J. et al. "Three Dimensional Anatomy with Stereolithography Models", Journal of Digital Imaging, 3, 200-203, 1990.
7. Swalens, B. and Kruth, J.P., "Medical Application Of Rapid Prototyping Techniques", 4^th International conference on Rapid Prototyping, 14-15^th June, pp 107-120, 1993.

8. Kai, C.C., Jacob, G. G. K., and Mei, T., "Interface Between CAD and Rapid Prototyping Systems. Part 1: A study of Existing Interfaces", International Journal of Advanced Manufacturing Technology, 13, 566-570, 1997.
9. Jacobs, P. F., "Rapid Prototyping And Manufacturing," society of Manufacturing Engineers, 1992.
10. Chua, K. C., and Leong, K. F., "Rapid Prototyping: Principles And Application In Manufacturing", John Wiley, April 1997.
11. Danek Group, "Fused Deposition Modelling Used To Aid In Design Of Spinal Implants", Rapid Prototyping Report, 14(8), 1994.
12. Taha, F. and Wouters, K., "First French Workshop on Medical Application Of Rapid Prototyping Techniques", Phidias, No. 1, 2-3, December, 1998
13. Thomas, S. W., "Optimising Composite Part Design and Manufacturing Using Stereolithography Tooling, E-Systems, Third International Conference on Rapid Prototyping, 267-279, June 7-10, 1992.
14. Jacobs, P. M., "Research Developments In Rapid Prototyping", Proc Inst. Mech. Engrs Vol. 209, 261-266, 1995.
15. Rosen, J. M., Soltanian, H., Radett, R. J. and Laub, D. R., "Evolution of Virtual Reality", IEEE Engng. Med. Biol., 16-22, March-April 1996.
16. Rosen, J. M., Lasko-Harvill, A. and Satava, R., "Virtual Reality and Surgery", In Computed Integrated Surgery (Eds R. H. Taylor et al.), ch.14, pp. 231-243, 1996.
17. Vaandrager, J.M., Zenneveld, F. W., Mothijsen, I. M. J., and Hovious, S. E. R., "First Generation 3D Model Of a Noma Patient", Phidias, No. 1, 6-7, December, 1998.

DESKARTES and the DA symbol are trademarks of DeskArtes Oy.
Copyright 1996-2008 DeskArtes Oy.