British Journal of Radiology (2004) 77, 504-507
© 2004 British Institute of Radiology
doi: 10.1259/bjr/50012454
Fused deposition models from CT scans
J R Meakin, PhD, CPhys, MInstP
1
D E T Shepherd, PhD, CEng, MIMechE
2 and
D W L Hukins, DSc, FRSE
2
1 Department of Bio-Medical Physics & Bio-Engineering, University of Aberdeen, Foresterhill, Aberdeen, AB25 2ZD and 2 School of Engineering, Mechanical Engineering, University of Birmingham, Edgbaston, Birmingham B15 2TT, UK
 |
Abstract
|
|---|
Fused deposition modelling (FDM) is a new method for rapid prototyping, a technique that produces models of objects from computer files. The most commonly used rapid prototyping technique for medical applications is stereolithography, but FDM has several potential advantages. This paper is concerned with the accuracy of an FDM model of a sheep lumbar vertebra using data from a CT scan. The model and the original vertebra were compared by making measurements with vernier callipers and by laser scanning. Visually, the model reproduced the features of the original object; this conclusion was supported by a comparison of the laser scans. Discrepancies in measurements were comparable with those of models produced using other rapid prototyping techniques, demonstrating that FDM is a viable method for making models for clinical use.
|
Introduction
|
|---|
Rapid prototyping (RP) is a technology in which a solid model can be created from a computer file. The computer file may be generated using modelling software or from medical images such as those obtained using CT. Various applications of rapid prototyping in medicine have already been identified. One particular application is in the production of anatomical models [1–6] to aid diagnosis and assist in the planning of surgery. Experience from several branches of surgery suggests that the resulting benefits include a reduction in wound size, the elimination of additional surgery [3, 6, 7], the potential to perform more complicated surgery, the potential to create customized implants [3, 7], and a reduction in operating time estimated to be between 17% and 60% [3, 7]. Anatomical models can also be used to explain pathology and surgical procedures to patients and their next of kin [3, 7]. This has benefits for both the patient and the health service, as studies have shown that patients who receive pre-operative education tend to recover more quickly post-operatively, have less pain and anxiety, and are more satisfied with the outcome of their surgery [8].
Most medical models to date have used stereolithography (SL); a technique where a liquid resin is polymerized by laser light to form a solid material of the required shape [9]. Fused deposition modelling (FDM) is a newer RP method in which a solid model is produced by controlled deposition of a molten polymer monofilament [9]. One of the advantages of FDM, over SL, is that the model is created in a single processing step. SL models require additional cleaning and curing under ultraviolet light [7], which increases the time to produce a model. Furthermore, the resin is toxic and expensive [9]. It has been suggested that the advantages of FDM make it more suitable for a hospital environment than SL [10]. The materials available for use in an FDM machine include a medical grade of acrylonitrile butadiene styrene (ABS) that can be sterilized using gamma radiation, and an investment cast wax. In principle, any thermoplastic with suitable physical properties could be used and recently models have been successfully created using poly(
-caprolactone), a biodegradable polymer [11].
In this paper we are concerned with evaluating the accuracy of bone models made by the FDM technique from computer files generated from CT-scans. If FDM can produce reliable anatomical models, it has the potential to be used for the applications described above.
|
Methods
|
|---|
Materials
A sheep lumbar vertebra was dissected to remove all soft tissues. Sheep lumbar vertebrae are a similar size to human cervical vertebrae [12] and so provide a model for one of the most intricate bones in the skeleton. This specimen was then wrapped in tissue paper soaked in physiological saline, placed in a sealed plastic bag and stored at –40°C until required.
CT scanning
A CT scan of the vertebra was acquired using a Somatom Plus 4 (Siemens, Erlangen, Germany) at Aberdeen Royal Infirmary. Before scanning, the vertebra was defrosted and placed inside a sealed plastic bag with no tissue paper. It was then positioned on a block of foam in the scanner to mimic the position of a cervical vertebra in the neck of a patient in the prone position. The scanning protocol used was that typically used at Aberdeen for the cervical spine. 33 slices were acquired using a spiral scan corresponding to a 2 mm reconstructed slice thickness. Each slice consisted of a 512 by 512 array with a pixel size of 0.12 mm. After scanning, the vertebra was re-wrapped in physiological saline wetted paper, placed in a sealed plastic bag and returned to the –40°C freezer.
Image analysis
The CT data were transferred from the scanner in ACR-NEMA 2.0 format using a DEC-702 magneto optical disc. The images were then processed using MIMICS version 7.1 software (Materialise NV, Dinnington, Sheffield). Images were segmented by firstly altering the contrast and then selecting a suitable threshold value to isolate the bony structure. The resulting data was then exported in STL format, which is the standard format used by RP machines.
Creation of model vertebra
The STL file was used to create a model from the polymer ABS using a fused deposition modeller (FDM3000; Stratasys Inc., Eden Prairie, MN). Soluble support material was also laid down so that the model vertebra was stable as it was built. The thickness of the ABS slices laid down by the FDM process was 0.254 mm. The total build time for the model was just under 5 h and the cost of the materials (modelling and support material) was £5. After the model was built, the soluble supports were removed using an ultrasound bath containing WaterWorks Soluble Release solution supplied by Stratasys.
Measurements
Both the model and the vertebra were measured in six places using vernier callipers. Measurements were taken of: (1) the vertebral body height; (2) the distance between the ends of the transverse processes; (3) the distance between the tips of the inferior and superior left hand articular processes; (4) the width of the spinous process; (5) the lateral diameter of the spinal canal at its superior entrance, and (6) the anterior–posterior diameter of the spinal canal at its superior entrance. Each measurement was performed by one of the authors (JRM) five times, and a mean and standard deviation calculated.
Laser scanning
Both the original bone and the model were scanned using a three-dimensional laser scanner (VI-900; Minolta (UK) Limited, Industrial Instruments, Milton Keynes, UK). Scanning was performed to show any overall differences between the surface of the bone and model. The scanning and post-processing of the scans was performed by Archaeoptics Ltd (Archaeoptics Ltd, Glasgow, UK).
|
Results
|
|---|
Figure 1
shows that the model, produced by FDM, closely resembles the original sheep vertebra. Table 1 shows that the measurements made from the original vertebra and the model differ by no more than a few millimetres. The only percentage difference that is not less than 10% is for the width of the spinous process; the high percentage difference here arises because the process is very thin (less than 4 mm wide). Figure 2 is a comparison of the surfaces of the vertebra and the model obtained by laser scanning. It shows the difference between the two surfaces with blue being zero difference. The regions that are most different (red regions) are in bands on the end-plate and at the tips of the vertebral processes.
View larger version (38K):
[in this window]
[in a new window]
|
Figure 1. Photograph of the rapid prototype model (left) together with the sheep vertebra (right) viewed from (a) the front and (b) above.
|
|
View larger version (52K):
[in this window]
[in a new window]
|
Figure 2. Plot of difference between the model and the vertebra viewed from (a) the front and (b) above. The surfaces of each object were obtained using a laser scanner, overlaid, and the different between them calculated. Blue shows regions of no difference and red shows regions of difference up to 3 mm.
|
|
|
Discussion
|
|---|
The two main sources of error in SL models are the thickness of the CT slices and the value used for thresholding the CT images [13]. The first source of error leads to partial volume averaging where sharp corners or edges falling between adjacent slices are smoothed out. This would account for the differences seen in our measurements of the vertebral body height and the articular process separation. The smaller differences in the other measurements are consistent with the in-slice pixel size (0.12 mm) being much smaller than the slice thickness (2 mm).
The second source of error can make the model either larger or smaller than the original object, depending on whether a lower or higher threshold value has been applied [13]. The change in size expands the surface of the model and has the effect of increasing external measurements and decreasing internal measurements [13]. The results from our study (Table 1
and Figure 2) are consistent with a threshold being applied that was too low. However, we only know it was too low because we can measure the dimensions of the original vertebra. Thus, in a real application, the threshold could not be optimized.
The differences in the distance measurements between the bone and the model in this study are of similar magnitude to those found in other studies that investigated accuracy of RP models [13]. Although these differences may make the model unsuitable for implant fitting, the majority of uses identified for anatomical models (described in the introduction) are still viable.
The ease and relative low cost of the FDM technique, which make it more useful for a hospital environment than the more commonly used SL technique, mean that it can have many uses in addition to anatomical models. Examples include the creation of patient specific moulds for brachytherapy [14] or immobilization devices during radiotherapy. Furthermore, although many models produced thus far have been of bony structures and have hence utilized CT scans, the application is not limited to this imaging modality and models have been created from MRI [15] and ultrasound images [16]. Recent developments in using poly(-caprolactone) in an FDM machine [11] also mean that patient-specific biodegradable scaffolds may be constructed for repair by in vivo tissue engineering.
|
Acknowledgments
|
|---|
We thank O Robb and P Skinner for the CT scans and McIntosh-Donald, Portlethan, UK for providing a sheep spine.
|
Footnotes
|
|---|
This study was supported by a grant from the Wellcome Trust (FDM equipment and salary for JRM). 
Received for publication July 30, 2003.
Accepted for publication November 19, 2003.
|
References
|
|---|
- Berry E, Brown JM, Connell M, Craven CM, Efford ND, Radjenovic A, et al. Preliminary experience with medical applications of rapid prototyping by selective laser sintering. Med Eng Phys 1997;19:90–6.[CrossRef][Medline]
- D'Urso PS, Askin G, Earwaker JS, Merry GS, Thompson RG, Barker TM, et al. Spinal biomodeling. Spine 1999;24:1247–51.[Medline]
- D'Urso PS, Earwaker WJ, Barker TM, Redmond MJ, Thompson RG, Effeney DJ, et al. Custom cranioplasty using stereolithography and acrylic. Br J Plast Surg 2000;53:200–4.[Medline]
- Mittelmeier W, Peters P, Ascherl R, Gradinger R. Rapid prototyping – the manufacture of models for the preoperative planning of reconstructive pelvic surgery. Orthopäde 1997;26:273–9.[Medline]
- Petzold R, Zeilhofer HF, Kalender WA. Rapid prototyping technology in medicine – basics and applications. Comput Med Imaging Graph 1999;23:277–84.[Medline]
- Potamianos P, Amis AA, Forester AJ, McGurk M, Bircher M. Rapid prototyping for orthopaedic surgery. Proc Inst Mech Eng [H] 1998;212:383–93.[Medline]
- Erickson DM, Chance D, Schmitt S, Mathis J. An opinion survey of reported benefits from the use of stereolithographic models. J Oral Maxillofac Surg 1999;57:1040–3.[Medline]
- Shuldham C. A review of the impact of pre-operative education on recovery from surgery. Int J Nurs Stud 1998;36:171–7.
- Pham DT, Gault RS. A comparison of rapid prototyping technologies. Int J Mach Tools Manufacture 1998;38:1257–87.[CrossRef]
- Popat AH. Rapid prototyping and medical modelling. Phidian Newsletter 1998;1:10–2.
- Zein I, Hutmacher DW, Tan KC, Teoh SH. Fused deposition modeling of novel scaffold architectures for tissue engineering applications. Biomaterials 2002;23:1169–85.[CrossRef][Medline]
- Wilke HJ, Kettler A, Wenger KH, Claes LE. Anatomy of the sheep spine and its comparison to human spine. Anat Rec 1997;247:542–55.[Medline]
- Choi JY, Choi JH, Kim NK, Kim Y, Lee JK, Kim MK, et al. Analysis of errors in medical rapid prototyping models. Int J Oral Maxillofac Surg 2002;31:23–32.[Medline]
- Poulsen M, Lindsay C, Sullivan T, D'Urso P. Stereolithographic modelling as an aid to orbital brachytherapy. Int J Radiat Oncol Biol Phys 1999;44:731–5.[Medline]
- Coward TJ, Watson RM, Wilkinson IC. Fabrication of a wax ear by rapid-process modeling using stereolithography. Int J Prosthodont 1999;12:20–7.[Medline]
- D'Urso PS, Thompson RG. Fetal biomodelling. Aust NZ J Obstet Gynaecol 1998;38:205–7.
Top
Abstract
Introduction
