Preliminary study of rapid prototype medical models
The Authors
Bal Sanghera, Research Associate at Imperial College of Science, Technology and Medicine, London, UK
Satyajit Naique, Trauma Fellow, at Imperial College of Science, Technology and Medicine, London, UK
Yannis Papaharilaou, Research Assistant at Imperial College of Science, Technology and Medicine, London, UK
Andrew Amis, Head of Biomechanics Group, at Imperial College of Science, Technology and Medicine, London, UK
Abstract
Rapid prototype models are directly integrated into non-engineering applications such as medicine. Medical models are used to plan complex procedures prior to surgery with potential to optimise patient treatment in the operating theatre. This paper presents results following a 12 month National Health Service Executive research project to assess the feasibility of using rapid prototype medical models. A total of 16 medical models were created. Nine anatomical sites were reconstructed from patient data acquired from five London hospitals. The purpose of the models is described and the commissioning surgeons as part of a questionnaire assessed their usefulness. Future developments are discussed and conclusions about the use of medical models are made.
Article Type:
General review
Keyword(s):
Rapid phototyping; Modelling; National Health Service; United Kingdom.
Journal:
Rapid Prototyping Journal
Volume:
7
Number:
5
Year:
2001
pp:
275-284
Copyright ©
MCB UP Ltd
ISSN:
1355-2546
Introduction
This paper reports the outcome of a 12 month National Health Service (NHS) Executive funded research study into the feasibility of applying rapid prototype (RP) 3D models in the medical field. Orthopaedic and maxillofacial/dental hospital departments were selected as the most favourable for model making applications.
The real cost to the NHS from inaccurate/incomplete treatment, especially in complex surgical procedures, is large but is not directly observed. The cost issue is convoluted into many factors. These include revision procedures to rectify problems with ineffective primary surgery, as well as the costs of using hospital facilities – bed occupancy, operating theatres, staff-time, etc. Implants for orthopaedic surgery can routinely cost up to £2,000 alone. Ill fitting implants cause discomfort and need replacing more frequently (Wallace, 1998; Morrey and Adams, 1992; Trail et al., 1999). Litigation may result from ineffectual, complicated secondary orthopaedic surgery. In difficult cases this may stem from uncertainty in the efficacy of treatment plans based on X-ray images alone, and surgery may not be performed under such conditions. Therefore, it is concluded that overall costs to the NHS are relatively high for complex surgical procedures that are not effective.
Use of physical models for treatment planning/visualisation instead of solely using computer software generated display images based on X-ray computed tomography (CT) or magnetic resonance imaging (MRI) data has a number of distinct advantages. Software methods giving the illusion of 3D volumes on a 2D screen can cause problems through view angle, depth, transparency and lighting anomalies that manifest as viewing orientation uncertainties. Other features include difficulty in accurately observing complex anatomy among bone fragments in the vicinity of fracture sites. Newer display systems adopt a different 3D approach utilising holographic technology (Robertson et al., 1995) and head up displays (Enislidis et al., 1997). However such technology is complex and expensive, requiring training and experience. In addition the result is not always easy to interpret for the surgeon. Some surgeons prefer a physical model for visualisation and for basing their treatment plans (D’Urso et al., 2000; Webb, 2000; Petzold et al., 1999; Winder et al., 1999). This approach is more intuitive for the surgeon and, therefore, may lead to more efficient treatment.
The aim of this work was to investigate the potential of RP medical models in surgical diagnosis and planning. Objectives were to introduce the concept of RP medical models to surgeons, develop protocols for acquiring and transferring hospital scanner data to Imperial College and to assess the usefulness of models by the relevant surgeon’s response to a detailed questionnaire.
Method
Front-end commercial software, MIMICSTM (Materialise NV, Technologielaan 15, 3001 Leuven, Belgium) was used to segment CT patient data to reveal anatomy of interest and convert this to a mesh based surface representation, i.e. stereolithography (STL) format. STL data was required by the RP controlling commercial software, QuicksliceTM (Stratasys Inc., 14950 Martin Drive, Eden Prairie, MN 55344-2020, USA). The models were built on a Stratasys model 1560 Fused Deposition Modeller (FDM) at Imperial College, London using ABS polymer. After trying out the model the surgeon’s response to a detailed questionnaire was used to determine the model’s effectiveness.
Results
In the course of this project nine anatomical sites were used to produce models from patient data acquired from five London hospitals (St Thomas’s, Guy’s, Charing Cross, St Mary’s and Great Ormond Street). A total of 16 medical models were created for respective commissioning medical consultants and university research personnel, Table I. The time quoted for each model does not include data processing or time for finishing models by hand.
The list below gives a short description of the models and is categorised by site:
- St Mary’s Hospital. A shoulder model (Plate 1) was created, taking three days to build using the RP technique. It revealed a complex fracture of the proximal humerus. Such images were difficult to interpret due to bone debris and anatomical distortion arising from the fracture. The model is now used for teaching purposes.
- St Mary’s Hospital. A complex hip model (see Plate 2) was made following severe loosening of the existing implant. The model took three days to build using the RP technique. Identification of anatomy and implant position from X-ray images was confusing due to the superposition of bone debris with the dislodged/damaged implant. The model assisted the surgeon and contributed towards developing a successful treatment regime for the patient.
- St Thomas’s Hospital. A knee model (Plate3) took four days to build using the RP technique. The femur was built separately to the tibia/fibula for simplicity. Correctly positioning the two sections together after construction was elementary as they fitted precisely in one orientation (i.e. the actual scan position). The patient had a history of knee surgery. On this occasion the knee had suffered a gross deformity following collapse of a previous osteotomy of the proximal tibia.
- Guy’s Hospital. Mandibular jaw distractor models (Plate 4). Created four skull models for preoperative treatment planning of mandible surgery. Each model took two days to construct using the RP technique. The models allowed a close physical matching of the distractor (McGurk et al., 1997) to the irregular shape of each mandible for the best fit and were also used to accurately locate the position of bone cuts on either sides of the mandible away from important nerves. Two further models (i.e. a total of six) have also been constructed which were beyond the original one year scope of the NHS funded project described in this paper.
- Guy’s Hospital. Radiotherapy facemask (Plate 5). A proof of principle concept for constructing a prototype, radiotherapy treatment, fixation facemask using RP and surface scanning technologies is shown here. A test subject had his face scanned using a CCD based 3D optical surface scanner (Tricorder Technology plc, Summerhouse Lane, Harefield, Middlesex, UK) to create the initial 3D face surface data. A mask thickness (for strength and durability) of 3mm was introduced through software. A simple prototype facemask was successfully constructed using this technique. The mask took five days to create.
- Guy’s Hospital. Eye socket reconstruction. This involved creating a template of one half of a face, especially the eye socket region. Using symmetry this model (based on the undamaged side) could then be used to aid reconstruction of a badly damaged eye socket on the opposite side of the face, following serious head trauma. Two models were made, but insufficient detail was achieved due to the very thin bone structure in the orbital regions. A different scan protocol and a more selective method of image pixel thresholding may be required. Further funding is required to investigate these issues.
- Guy’s Hospital. Wrist models. Two models were made for treatment planning in common distal radial wrist fractures. It was concluded from these that a different scan regime was necessary to reveal intricate damaged and compressed anatomy consistent in such fractures. More work is required to assess the potential of RP models for this application. This is a common clinical problem that may lead to routine use of RP models.
- Great Ormond Street Hospital. Craniofacial deformity models (Plate 6). The potential of RP generated models for treatment planning of gross facial deformity correction in children is being investigated with X-ray CT generated data. The first model was received with great enthusiasm by the plastic surgeon/craniofacial head consultant for treatment planning prior to complex surgery of congenital skull deformity. The model revealed important 3D information about the precise planning of complex facial osteotomies and related treatment regimes not available in the X-ray images alone. A second model was created and is being used for similar purposes.
- Charing Cross Hospital. Sections of orangutan spine models (Plate 7). An adult male orangutan from London Zoo came under scrutiny following severe illness. After a period of time it was humanely put down due to its acute condition and suffering. Following post mortem the primate was discovered to be suffering from a form of encephalopathy. It was also observed that the animal appeared to have a distinct scoliosis. The animal’s spine was studied in detail for this unexpected phenomenon (Naique et al., 2000). X-ray CT images of the orangutan were taken and RP models of sections of the spine constructed for closer inspection and analysis. The models confirmed that the observed spine deformity was consistent with idiopathic scoliosis. This was corroborated by dissection of the spine and appears to be the first recorded case of idiopathic scoliosis in a non-human primate.
- Imperial College. Coronary artery bypass graft model (Plate 8). RP technology was used to create an accurate model of an animal heart coronary artery bypass graft using data acquired with MRI imaging. The RP model of the bypass graft is entirely solid and is being used as a cast from which a hollow compliant silicone rubber bypass graft model is created. In the future it is hoped to create an accurate detailed hollow bypass graft model directly if a suitably compliant build medium becomes available for RP technology. MRI flow quantification experiments and numerical simulations will be used to obtain a detailed description of the flow field in the distal section of the end-to-side aorto-coronary anastomosis.
Discussion
The FDM RP machine at Imperial College was chosen over other RP techniques for a combination of reasons. The FDM machine has white ABS thermoplastic build material that is robust, visually resembles bone and can be cut with operating theatre tools. The option of an investment casting wax head in the RP machine was seen as a bonus for creating patient specific implants and any 3D models created could be used as a master for casting. Use of clinical grade ABS offered the potential of RP models for in vivo application, e.g. drill-guides for screw insertion, etc. The RP machine is approved for office environments with little waste product generation. The FDM machine was intended predominantly for research and routine engineering/design projects at Imperial College with a low model yield in comparison to commercial manufacturers. When purchased it was seen as a simple, adaptable cost-effective solution in comparison to other more complex laser based RP techniques, e.g. selective laser sintering (SLS), laminated object manufacturing (LOM) and stereolithography (SLA).
Surgeons ordered models as the need arose and finances permitted. This uncertainty made it difficult to plan ahead for a truly quantitative study so the research presented here is predominately qualitative in nature.
The problem of transferring electronic data from hospital scanners to Imperial College was addressed for each site. In some cases this was a complex and considerable task. It was quite common to have hospital scanners configured on separate internal NHS hospital networks that did not communicate with external wide area networks for security reasons. Different scanner output media and formats (e.g. optical disks and their formats, etc.) were also negotiated successfully, usually in collaboration with local medical physics staff. Issues relating to the Caldicott Guardian (Caldicott, 1997) security of patient data protocols have also been solved in many cases by anonymising patient data. The Digital Imaging and Communications in Medicine (DICOM) standard is a widely approved format for allowing medical information to be transferred between devices along computer networks, e.g. X-ray images from scanner to image viewing PC or back-up media. In some cases scanner manufacturers modified the DICOM standard format and this introduced additional difficulties in reading and anonymising medical images for that hospital. Once the protocol for each site was established it was then used routinely to transfer patient data for future models.
Medical model outcome
Shoulder model
Further shoulder models are being discussed for treatment plan/teaching purposes and to characterise the range of fractures involved in common shoulder surgery. Once this range of models has been built they will be used to help design implants to correct fractures of the head of the humerus. Future studies are dependent on availability of funds.
Hip model
In general a very limited number of hip replacement procedures can be performed on each acetabulum due to loss and weakening of existing bone. RP medical models can act as important treatment planning tools in complex cases to aid the chances of successful surgical outcome and long term implant stability. The models help with the design of custom-made polyethylene acetabular sockets that fit into the remaining bone stock.
Knee model
This model was used for the design of a custom-made knee prosthesis that required fixation stems in an unusual configuration relative to the displaced tibial plateau. The model was successful but the company responsible for producing the custom-made implant made a manufacturing error and a conventional “off the shelf” implant was used instead. The model is used for teaching purposes now.
Mandible distractor models
Once attached, a small screw thread on the distractor was used to slowly slide the surgically modified section of mandible in a particular direction. This process of “distraction osteogenesis” allowed the deformed bone to grow by approximately 15mm in the desired direction. This movement of the mandible corrected the facial deformity of patients over a period of several weeks. The distractor was removed after the bone had healed. This medical application appears to be highly successful for continued and routine use of medical models in the future. The potentially superior treatment plan regime exercised here is expected to result in improved surgery leading to fewer complications. A more quantitative study of the precise improvements is required in future.
Facemask model
Current technology requires the creation of a positive shell from a plaster of paris mould as it dries on a patient’s face. The facemask (i.e. negative) is formed by vacuum forming a heated clear plastic sheet to the positive moulding. This process is crude and unpleasant for the patient, producing much waste products and mess. Dedicated facilities for material storage and disposal are required and face movement during formation of the positive is a recognised problem, especially in the case of children. RP technology offers improvements to this existing regime in an office environment. Additional CCD cameras for scanning would improve definition of face edges (e.g. chin) allowing better facemask design and fit. Additional radiotherapy treatment couch fixation accessories could be integrated as computer aided design (CAD) data with the patient facemask data prior to RP model production. The potential exists for future improved versions of the facemask to be used as a mould or applied directly to the patient. There is scope for much work here if future funding is available to investigate the potential of this elegant application.
Craniofacial deformity models
This medical application appears very well suited to RP model use. The first model provided more valuable information about complex treatment planning than was expected. As a result surgery was rescheduled later to accommodate the improved plan. The second model was successfully used clinically for a child with existing facial bone grafts who required further facial reconstruction. A new grant proposal with Great Ormond Street is under discussion to promote such research in this field. The study is likely to take a significantly more quantitative approach towards the benefits of RP technology here.
Orangutan spine models
It has been assumed that biplanar asymmetry is a major determinant contributing towards idiopathic scoliosis. Distinct human posture and an erect bipedal stance were believed to be important factors for causing a lordotic spine to buckle resulting in rotational scoliotic deformity. Idiopathic scoliosis is not associated with primates primarily for these reasons. Orangutans have predominantly quadrupedal stance, large forelimbs and long thoracic-lumbar kyphosis. This may be the first recorded case of idiopathic scoliosis in a non-human primate. If so, the causes of idiopathic scoliosis may need refining as orangutans have substantially different form and gait compared with humans. This conclusion is supported by work from other groups researching non-bipedal aetiology of idiopathic scoliosis (Porter, 2000; Pincott and Taffs, 1982). It is hoped that similar models can be used to plan human spinal surgery in the future.
Arterial bypass graft model
Arterial bypass grafting is a commonly performed operation unfortunately having a 50 percent failure rate over a ten-year period (Bryan and Angelini, 1994). Determination of accurate local flow field in bypass grafts is an important tool for understanding mechanisms behind graft failure. The small diameter of coronary vessels and the complex pattern of movement caused by heart/respiratory motion makes it extremely difficult to obtain accurate detailed in vivo measurements of complex flow in aorto-coronary bypass grafts, using the flow measurement technology currently available. Therefore, in vitro studies in realistic models of vessel anatomy are necessary to understand the haemodynamics in coronary artery bypass grafts and their implications on vascular biology. RP models offer the potential to achieve representative measurements in this difficult area.
Medical model clinical assessment
The models were evaluated by posing a set of 16 questions to the consultant responsible for ordering the model. Where applicable the consultant was asked to grade their response between 1 (for poor) to 10 (for excellent) for each question. The questions asked were:
- (1) How do you rate the model for treatment planning?
- (2) How do you rate the model as a surgical visual aid?
- (3) How do you rate the model for training purposes?
- (4) How do you rate the time taken to create the model?
- (5) How do you rate the cost of the model?
- (6) How do you rate the model accuracy?
- (7) How do you rate any potential timesaving in surgery using the model?
- (8) How do you rate any potential cost saving in surgery using the model?
- (9) What is the approximate cost of the surgical procedure including implants?
- (10) What would you estimate as the predicted cost of secondary treatment due to inaccurate primary treatment?
- (11) How do you assess the level of suing and how RP models can reduce this with better primary treatment?
- (12) How do you rate using the models as proof of the treatment plan?
- (13) How do you rate the implications of surgery not done due for fear of malpractice?
- (14) How do you rate the overall score of using the model?
- (15) Will you have other models made in future?
- (16) Do you think this facility should be available routinely as part of the NHS?
Table II shows examples of feedback to individual questions for three models. Models and consultants are not identified here, in accordance with the general wish of surgeons for anonymity regarding their response to individual questions. A (–) signifies that the surgeon was unable to assign a score rating to that question.
Consultants were also asked to rate the “overall score” of using the RP model facility, similarly between 1 and 10, in the context of their clinical work. This score was used to summarise their response to the questionnaire as a whole and applied especially when surgeons preferred not to assign scores to specific questions but preferred to give an epitomised feedback (Table III).
Wrist and eye socket models required further study and were omitted from Table III as they did not contribute clinically. In the case of the mandible models a single questionnaire was used as models were employed for the same surgical procedure and a similar clinical outcome resulted. The score presented for the craniofacial model is an average between the score for the second model used clinically (score 6) and the score for the first model as a tool to plan complex surgery (score 9).
This feedback suggested that the models were an effective, in some cases invaluable, tool for the surgeons. The demand for further models from the same consultant are seen as proof that once the usefulness of RP models has been recognised by a surgeon, that surgeon is likely to want to use them routinely. Thus, they may come to play an important role in patient care in the NHS. It is hoped to obtain funding for future work with individual consultants and specific RP projects based on the findings of this paper.
Implications of medical model use
The majority of models created here were used to plan treatment where the physical models revealed 3D details difficult to discern on the X-ray images alone. This study suggests that models should be available routinely on an ad hoc basis for complex orthopaedic surgery. Most interest arose from consultants in maxillofacial/craniofacial surgery, where nine models were created for this application. RP models in this field appear to be highly regarded. In many cases surgery in this field can take place over many years resulting in numerous operations to correct deformity. A superior treatment plan resulting from the use of medical models has the potential to lead to enhanced surgery with the prospect for better primary care and for fewer follow up procedures to correct for irregularities etc.
Many consultants expressed concern about the implications of difficult surgery in complicated procedures that could then lead to legal proceedings against the surgeon. Some consultants were interested in having models for insurance purposes. The model was seen as an extra tool in the surgeon’s armoury against possible legal action. Interest was shown in using the models to validate a treatment plan for complex secondary surgery.
Most of the models made have fulfilled the role that each consultant had for their particular use. The greatest concern from all consultants was the financial cost of the service. Participating consultants paid for models from their own budgets. Imperial College Consultants (ICON) Ltd, a multidisciplinary consultancy company, funded the RP machine at Imperial College for medical research and routine in house engineering applications on a commission basis. The NHS Executive funded a postdoctoral research associate to investigate the application of medical models clinically. Each surgeon would be willing to have more models made routinely if cost was not such a major factor. They all advocated that the service be funded centrally or heavily subsidised by the NHS rather than through each consultant’s own limited budget. It is the opinion of surgeons who participated in this study that hospital financial administrators are more likely to support this approach, leading to increased hospital use of RP models in treatment planning for complex bone surgery. Potential exists for the NHS or a group of hospital trusts/departments to collaborate and purchase a dedicated RP machine (∼£50,000 today) and technicians to run the service for prompt model creation. Alternatively a hospital technician may be funded to process scan data prior to models being created by commercial engineering RP manufacturing companies running a bureau service. The large number of companies now competing to create RP models in the private sector has seen the average price per model drop significantly over the past year. This approach is worthy of investigation as a cost-effective approach to model creation. In the past commercial companies were expensive and did not produce models on a priority basis for patients requiring prompt surgery due to the low numbers of models requested.
To date only London hospitals were targeted due to their proximity, ease of visiting and due to previous contacts that could assist with issues discussed earlier.
Future developments
A bioresorbable medium (Sharifi-Yazdi and Amis, 2000) is a material that can be placed into the location of damaged bone to slowly dissolve and stimulate new bone to grow in its place. In cases of severe damage, e.g. bone-crushing, large fractures, etc., it is proposed to custom design bone replacement implants using X-ray CT images and then create these using RP model making technology. It is anticipated that matching a bioresorbable material to suitable RP model-making technology may revolutionise customised treatment in this field.
It is hoped that the RP model making facility can be extended to cover a prompt, cost-effective NHS London regional and nationwide service. The positive responses from London consultants suggested that such a service would be beneficial to patient care nationally.
Conclusions
Medical models have revealed complex 3D anatomy difficult to identify in routine X-ray CT images alone. Such models, especially for maxillofacial/craniofacial work, have been used to prepare treatment plans with more accuracy leading to improved surgery. This has benefits of enhanced primary treatment, less time in surgery per procedure for patient/clinical staff and the potential for fewer follow up procedures in future. This approach should lead to better patient care regimes, enhanced use of resources and the potential for reduced waiting lists.
Further research is required to explore the range of usage of medical models and to fully assess quantitative potential improvements from their use in surgery. Orthopaedic and maxillofacial/craniofacial hospital departments are most likely to benefit substantially from a centrally funded/subsidised, routine RP medical model-making facility available regionally or to the entire NHS. Proposed new areas of investigation, e.g. bioresorbable materials and RP model manufacture, may significantly benefit NHS patient treatment for the future and are worthy of investigation.
Table ISummary of the models made for each site
Plate 1Shoulder model. The head of the humerus (at centre, left) has been fractured. The fragments have burst apart and slipped down alongside the bone shaft
Plate 2Hip model. The loosened hip prosthesis eroded the bone of the acetabulum, causing the large hole at the centre of the picture
Plate 3Knee model. A surgically-created bone cut (osteotomy) across the tibia (at centre, right side of picture) has slipped out of place before healing, so that the femur (top) is no longer in line with the tibia
Plate 4Mandible model. The bone was cut in a nearly vertical plane, starting near the bottom, right corner of the picture. By sliding the cut surfaces apart, the asymmetry was corrected
Plate 5Facemask model. A test subject’s face was scanned using an optical scanner. The data generated formed the basis of this rudimentary radiotherapy fixation mask as proof of principle concept
Plate 6Craniofacial model. This was used to plan the cuts and movements of sections of the skull to correct the abnormal growth
Plate 7Spine section model. The model was used to confirm the presence of scoliosis in the spine of a primate
Plate 8Artery model. The artery RP model was used as a cast to create a hollow compliant silicone model for use in flow quantification measurements
Table IISpecific responses of surgeons to 16 questions for three RP models
Table IIISurgeon’s response to overall score of using RP models
References
Bryan, A, Angelini, G. (1994), "The biology of saphenous vein graft occlusion: aetiology and strategies for prevention", Current Opinions in Cardiology, No.9, pp.641-9.
Caldicott, F (1997), "The Caldicott committee: report on the review of patient-identifiable information", Caldicott Report, National Health Service Executive, Department of Health, London, .
D’Urso, P.S, Earwaker, W.J, Barker, T.M, Redmond, M.J, Thompson, R.G., Effeney, D.J, Tomlinson, F.H (2000), "Custom cranioplasty using stereolithography and acrylic", British Journal of Plastic Surgery, Vol. 53 No.3, pp.200-4.
Enislidis, G, Wagner, A, Ploder, O, Ewers, R. (1997), "Computed intraoperative navigation guidance – a preliminary report on a new technique", British Journal of Oral and Maxillofacial Surgery, Vol. 35 No.4, pp.271-4.
McGurk, M, Slabberti, H.D.V, Amis, A.A, McDonald, F (1997), "A preliminary report of a case using an intra-oral distraction device", British Dental Journal, Vol. 183 No.2, pp.63-6.
Morrey, B.F, Adams, R.A (1992), "Semiconstrained arthroplasty for the treatment of rheumatoid arthritis of the elbow", The Journal of Bone and Joint Surgery, Vol. 74-A No.4, pp.479-90.
Naique, S, Porter, R, Cunningham, A., Sanghera, B, Hughes, H, Amis, A.A (2000), "Scoliosis in an orangutan", Spine, .
Petzold, R, Zeilhofer, H.-F., Kalender, W.A (1999), "Rapid prototyping technology in medicine – basics and applications", Computerized Medical Imaging and Graphics, Vol. 23 pp.277-84.
Pincott, J.R., Taffs, L.F (1982), "Experimental scoliosis in primates: a neurological cause", The Journal of Bone and Joint Surgery (British), Vol. 64 No.4, pp.503-7.
Porter, R.W (2000), "Idiopathic scoliosis: the relation between the vertebral canal and the vertebral bodies", Spine, Vol. 25 No.11, pp.1360-6.
Robertson, D.D, Sutherland, C.J, Chan, B.W, Hodge, J.C, Scott, W.W, Fishman, E.K (1995), "Depiction of pelvic fractures using 3D volumetric holography: comparison of plain X-ray and CT", Journal of Computer Assisted Tomography, Vol. 19 No.6, pp.967-74.
Sharifi-Yazdi, F, Amis, A.A (2000), "Bioresorbable Poly-DL-Lactide-co-Glycolide/Bioglass 45S5 biocomposite for tissue engineering", Proceedings of the Second Smith & Nephew International Symposium, Tissue Engineering, Advances in Tissue Engineering, Biomaterials and Cell Signalling, York, .
Trail, I.A, Nuttall, D, Stanley, J.K (1999), "Survivorship and radiological analysis of the standard Souter-Strathclyde total elbow arthroplasty’", The Journal of Bone and Joint Surgery (British), Vol. 81-B pp.80-4.
Wallace, G (1998), Joint Replacement in the Shoulder and Elbow, Butterworth-Heinemann, Oxford, .
Webb, P.A (2000), "A review of rapid prototyping (RP) techniques in the medical and biomedical sector", The Journal of Medical Engineering and Technology, Vol. 24 No.4, pp.149-53.
Winder, J, Cooke, R.S, Gray, J., Fannin, T, Fegan, T (1999), "Medical rapid prototyping and 3D CT in the manufacture of custom made cranial titanium plates", The Journal of Medical Engineering and Technology, Vol. 23 No.1, pp.26-8.