43
The CADCAM contribution to customized orthopaedic implants H V Crawford, BSC, CEng, MIMechE, P S Unwin, MSc and P S Walker, PhD, CEng, MIMechE Department of Biomedical Engineering, Institute of Orthopaedics, Stanmore, Middlesex CADCAM (computer aided designlmanufacture) production methods are often associated with mass production; working in the medical field at the Department of Biomedical Engineering, the requirement is for one-of, individualized implants. Using a knowledge-based system, implant designs are produced from X-ray data. Assembly from modular components has greatly reduced the production time of implants for bone tumour cases. CADCAM techniques are also used in the production of custom-made hip replacements using digitized data gathered from radiographs. Femoral canal shape is calculated and the optimal implant designed and manufactured from titanium alloy on the Department4 C N C (computer numerically controlled) machines. 1 RNOH AND THE DEPARTMENT OF BIOMEDICAL ENGINEERING
The Royal National Orthopaedic Hospital (RNOH) in Stanmore, Middlesex, has been prominent in the field of orthopaedic surgery for many years. Since the late 1940s it has been the home of the Institute of Orthopaedics (part of the University College and Middlesex School of Medicine) which is involved in postgraduate teaching and research. Within the Institute, the Department of Biomedical Engineering (BME) carries out research and development, primarily in joint replacement. Part of this work covers new designs for standard hip and knee replacements for primary and revision treatment of osteoarthritis. The Department also provides a unique service in a field known as limb salvage. As an alternative to amputation in bone tumour cases or in revision cases with substantial loss of bone, a major implant can be used. Typical applications are the proximal or distal part of the femur and the proximal tibia. The supply of these major implants which are produced mainly from sections of titanium alloys and cobalt-chrome castings has risen steadily since the first one was produced in 1946, to over 200 per year, with a total of over 2000 implants having been made to date.
required to determine the design features required for a particular case will range from the straightforward and common knee and distal femur replacements to the rarer and more complex hemi-pelvic replacements. 3 NEW TECHNOLOGY
The decision to implement a computerized knowledgebased system to aid the design of the custom implants was prompted by the desire to increase the efficiency and output of the Department and allow the skilled specialists more time to concentrate on the complex cases. The computing facilities within the department already ranged from a MicroVAX mini to a network of Apple Macintosh computers and a few isolated PC (personal computer) compatibles. Since the system was intended to be used by relatively unskilled operators and to link with CAD (computer aided design) and database systems running on the Appletalk local area network, the whole system has been developed on that platform. Proprietary expert system shells were considered but the system is currently being developed under the Hypercard programming language. 3.1 Digitizing X-rays
2 TRADITIONAL METHODS OF IMPLANT DESIGN
When a surgeon requires a major custom implant fulllength measurement X-rays of the limb are provided showing the frontal and side views and incorporating a measurement scale to allow the magnification factor to be calculated. The resection level is marked and a series of precise measurements taken of key anatomical points. From these, the critical dimensions for the custom implant can be calculated and passed to a draughtsman who produces the engineering drawings required by the workshop. These and other dimensions are also used to produce a schematic drawing for the surgeon’s reference during the operation to insert the implant. The skills This paper was presented at Effective CADCAM 91 conference held in Coventry on 6 7 November 1991. The M S was received on 4 February 1992 and was accepted for publication on 13 M a y 1992.
H00392 Q IMechE 1992
Earlier studies have concluded that the use of computerized analysis systems directly to read and interpret medical images [X-rays and CT (computerized tomography) scans] is fraught with problems caused by poor and unreliable image quality and the variation in techniques used by different centres (1, 2). For this reason the decision has been made to continue to employ the expert assistance of human intervention in entering the patient’s bone dimensions into the computer program. The precise marking up required in the past on the measurement X-rays has been reduced to the indication of the bone’s major centre-line axis and the transection line at which the surgeon wishes to cut the bone. The X-rays are then placed on a large, backlit digitizing tablet and, in response to computer prompts, the major anatomical landmarks are idcntified, returning the relative X and Y coordinates to the
0954-4119/92 $3.00
+ .05
Downloaded from pih.sagepub.com at RMIT UNIVERSITY on July 20, 2015
Proc Instn Mech Engrs Vol 206
44
H V CRAWFORD, P S UNWIN AND P S WALKER
6months enables the implant to be extended by the surgeon to match the natural growth of the opposite limb.
3.3 SMILES
Fig. 1 The points on a frontal X-ray which would need to be digitized to provide data for the prosthesis design software for a proximal femur replacement prosthesis
computer. In order to allow the operator to concentrate on the X-ray the textual prompts on the computer screen are supported by synthesized spoken instructions and feed-back. Figure 1 shows the points that need to be digitized for the design of a distal femur and knee replacement implant.
The task of the design and production of some of the more common major implants has been simplified and speeded by the implementation of the Stanmore modular individualized lower extremity system (SMILES). This is a range of stock modular components which a statistical analysis of past implant designs has shown can be easily tailored to build into the majority of implant designs. Attempts have been made to develop modular sets of components that the surgeon can assemble into the required configuration at the time of the operation (3, 4), but these cannot be tailored quite as precisely for a particular patient and make no allowance for bone curvature or other special factors. The manufacture of the SMILES modular components is both sub-contracted and made in-house and in either case the savings in assembly time for the modular prosthesis is a considerable advantage. Where possible, the expert system software will base a prosthesis design on the availability of standard components, but will indicate the contra-indications where this is not appropriate.
3.4 CAD links
3.2 Design choices The full digitized coordinate data from the X-ray films are reduced to the salient measurements for processing within the system. Using rules provided by the experience of the human experts and deduced from our experience in the fields of bone growth and remodelling, strain distribution, surgical technique and implant design it may prompt for further information which may be based on surgeon preferences, the patient’s age and condition or other peripheral factors. If the system encountcrs any inconsistencies in the data or unusual anatomical conditions it will make recommendations but allow modification to its design decisions. The design rules currently entered into the knowledge base allow it to produce design recommendations for three major classes of implant which account for almost 60 per cent of the output: (a) proximal femoral replacements (top of femur and hip joint), (b) distal femora1 replacements (bottom of femur and knee joint), (c) proximal tibia replacements (top of tibia and knee joint). Proximal humeral replacements, which account for a further 8 per cent of the output, are obvious candidates for inclusion in the system. Work is also continuing on establishing the rules for design of a further category of implants (5 per cent) which are variants of these designs and are used for the treatment of young patients who require ‘growing’ or ‘extendible’ prostheses. Following the initial major surgery required to implant the prosthesis in the patient, a more minor operation every
The use of the modular components in the fabrication of the custom implants, where appropriate, has simplified the requirements of the engineering drawings supplied to the workshops. Where each element had to be precisely dimensioned and toleranced before, the prefinished items now only need to be identified by their critical dimensions and any custom modification required for a particular implant indicated. This is achieved by a customized program running within Schlumberger’s MacBravo CAD package which uses the implant design parameters generated by the expert system to construct the required drawings (Fig. 2).
3.5 Op drawings Similar data from the expert system detailing the implant parameters is fed to a database package for storage and for the generation of suitable annotated surgical operation drawings (Fig. 3). Further links are envisaged which will allow the automated production of packing lists and box labels in addition to adding the information to the existing unique database resource. This stores the details of patients and implants from 20 years of orthopaedic experience together with analysis of retrieved implants which records failure modes and wear patterns.
4 CUSTOM HIPS
A further application of computer technology to the production of orthopaedic implants is the hip design workstation. Developed in conjunction with Zimmer,
Part H: Journal of Engineering in Medicine
Q IMechE 1992 Downloaded from pih.sagepub.com at RMIT UNIVERSITY on July 20, 2015
THE CADCAM CONTRIBUTION TO CUSTOMIZED ORTHOPAEDIC IMPLANTS
I
28.00
45
Stem diameter tapering from 13.0 down to 11.5
Fig. 2 CAD image of a distal femur and knee replacement produced to show dimensions required to build the implant from modular components
one of the large multi-national orthopaedic suppliers, this VAX-based system also uses data digitized from measurement X-rays, but for the production of custommade, non-cemented hip replacements. The majority of total hip replacements (THRs) are fitted by opening out an oversize cavity down the femoral canal and fitting in a n off-the-shelf implant held in place with acrylic cement. While these have offered good medium-term fixation of the implant with rapid mobilization of the patient, the long-term performance is compromised by fatigue failure of the cement followed by loosening, especially in young patients. Many modern designs of THR rely on a closer fit between stem and bone and are used without cement. Early indications are that, if well fitted, these provide a better long-term prognosis (5). Any mass-produced press-fit THR will clearly be based on assumptions about average anatomy, but a better fit should be achieved with a custom implant. Custom implants are also indicated in cases of congenital deformity where unusual anatomy can be corrected by, for example, adjusting femoral head positions. Figure 4 shows two views of a standard implant together with
the medial view of one designed for a patient requiring correction of the femoral head position. 4.1 Hip design workstation
The Stanmore hip design workstation has, at its core, a datafile representing the canal profiles of a n average femur, based on measurements taken from sectioned femurs and CAT (computerized axial tomography) scan images (6). Data collected by digitizing the canal profiles of a particular patient’s measurement X-rays (frontal and lateral) are used to scale the average profiles in order to predict the full three-dimensional canal shape for the patient. The specially written stem design program then simulates the surgical preparation of the femur with reamers and rasps, to match key geometrical parameters of the femur. The program then generates a stem which is a close proximal fit, a clearance fit in the mid region and a sliding fit distally. User-selected variables allow for appropriate neck cut, stem length, closeness of fit, collar type and anteversion angle. The final implant shape is held as an array of 100 sections
% Bone replaced: 57 % Stem length/bone length: 308
(a)
Fig. 3 Drawing showing information required by a surgeon at operation to insert a proximal femoral replacement prosthesis. This is created by the database program
which stores implant data
Fig. 4
(b)
(C)
Showing (a) front and (b) medial views of a standard implant together with (c) the medial view of one designed for a patient requiring correction of the femoral head position Proc Instn Mech Engrs Vol 206
Q IMechE 1992 Downloaded from pih.sagepub.com at RMIT UNIVERSITY on July 20, 2015
H V CRAWFORD, P S UNWIN AND P S WALKER
46
through the implant, with 40 points around each section defined by their Cartesian coordinates. New algorithms are currently being written for more accurate canal shape prediction by using data from selected CT scan slices. 4.2 CNC
A CNC tool path file to machine the implant shape from titanium 318 alloy (Ti, A1 6%, V 4%) is generated by another custom FORTRAN program. Running on a four-axis Fadal VMC40 mill the machining process is carried out in two stages: an initial helical roughing cut followed by a finishing cut with a ball-ended cutter. Some hand finishing and polishing is required and a separate spherical head is fitted to provide the articulation with the acetabular cup in the pelvis. One advantage of the standard off-the-shelf implant designs is the wide range of tooling available to the surgeon to assist in fitting the implant. With custommade implants, custom-made tooling is required, and work is continuing to develop a simple but effective method for producing rasps to match the implants at a cost that is not prohibitive. With recent improvements in the cutting algorithm the time to machine a typical hip has been reduced from 6 to under 3 hours. 5 CONCLUSIONS
Computer technology is coming to the aid of the medical profession in many of its disciplines and this includes the field of orthopaedics. Accurately matching the prosthesis design to the patient increases the probability of survival, while the reductions in design and
manufacture time for custom prostheses make them viable alternatives to the modular kits. The prototype ‘expert system’ has proved the principles. Conversion to a language better suited to decision making and deductive reasoning will allow more power to be added without sacrificing speed of response. The use of computer-designed implants is improving the quality of life for many patients.
ACKNOWLEDGEMENT
The application of the hip design workstation to CADCAM hips for juvenile rheumatoid arthritis cases is supported by the Arthritis and Rheumatism Council.
REFERENCES 1 Ellam, S. V. and Maisey, M. N. Knowledge based system to assist in medical image interpretation: design and evaluation methodology. Sixth Technical Conference of the British Computer Society, Brighton, 1988. 2 Garland, L. H. Studies on the accuracy of diagnostic procedures. Am. J. Roentgenology, 1959,97 (4), 901-905. 3 Bos, G., Sim, F. H., Pritcbard, D. J., Sbives, T. C., Rock, M., Askew, L. and Chao, E. Y. S. Prosthetic proximal humeral replacement: the Mayo Clinic experience. In Limb salvage in musculoskeletal oncology, Orlando, 1985, pp. 61-72 (Churchill Livingstone). 4 Ritschl, P., Braun, O., Pongracz, N., Eyb, R., Ramach, W. and Kotz, R. Modular reconstruction system for the lower extremity. In Limb suluuye in musculoskeletal oncology, Orlando, 1985, pp. 237-243 (Churchill Livingstone). 5 Bargar, W. Shape the implant to fit the patient. Clin. Orthop. Related Res., December 1989,249, 73-78. 6 Walker, P. S., Poss, R., Robertson, D. D., et al. Design analysis of press fit stems. Orthop. Related Sci., 1990,1, 75-85.
Part H: Journal of Engineering in Medicine
@ IMechE 1992 Downloaded from pih.sagepub.com at RMIT UNIVERSITY on July 20, 2015
The CADCAM contribution to customized orthopaedic implants H V Crawford, BSC, CEng, MIMechE, P S Unwin, MSc and P S Walker, PhD, CEng, MIMechE Department of Biomedical Engineering, Institute of Orthopaedics, Stanmore, Middlesex CADCAM (computer aided designlmanufacture) production methods are often associated with mass production; working in the medical field at the Department of Biomedical Engineering, the requirement is for one-of, individualized implants. Using a knowledge-based system, implant designs are produced from X-ray data. Assembly from modular components has greatly reduced the production time of implants for bone tumour cases. CADCAM techniques are also used in the production of custom-made hip replacements using digitized data gathered from radiographs. Femoral canal shape is calculated and the optimal implant designed and manufactured from titanium alloy on the Department4 C N C (computer numerically controlled) machines. 1 RNOH AND THE DEPARTMENT OF BIOMEDICAL ENGINEERING
The Royal National Orthopaedic Hospital (RNOH) in Stanmore, Middlesex, has been prominent in the field of orthopaedic surgery for many years. Since the late 1940s it has been the home of the Institute of Orthopaedics (part of the University College and Middlesex School of Medicine) which is involved in postgraduate teaching and research. Within the Institute, the Department of Biomedical Engineering (BME) carries out research and development, primarily in joint replacement. Part of this work covers new designs for standard hip and knee replacements for primary and revision treatment of osteoarthritis. The Department also provides a unique service in a field known as limb salvage. As an alternative to amputation in bone tumour cases or in revision cases with substantial loss of bone, a major implant can be used. Typical applications are the proximal or distal part of the femur and the proximal tibia. The supply of these major implants which are produced mainly from sections of titanium alloys and cobalt-chrome castings has risen steadily since the first one was produced in 1946, to over 200 per year, with a total of over 2000 implants having been made to date.
required to determine the design features required for a particular case will range from the straightforward and common knee and distal femur replacements to the rarer and more complex hemi-pelvic replacements. 3 NEW TECHNOLOGY
The decision to implement a computerized knowledgebased system to aid the design of the custom implants was prompted by the desire to increase the efficiency and output of the Department and allow the skilled specialists more time to concentrate on the complex cases. The computing facilities within the department already ranged from a MicroVAX mini to a network of Apple Macintosh computers and a few isolated PC (personal computer) compatibles. Since the system was intended to be used by relatively unskilled operators and to link with CAD (computer aided design) and database systems running on the Appletalk local area network, the whole system has been developed on that platform. Proprietary expert system shells were considered but the system is currently being developed under the Hypercard programming language. 3.1 Digitizing X-rays
2 TRADITIONAL METHODS OF IMPLANT DESIGN
When a surgeon requires a major custom implant fulllength measurement X-rays of the limb are provided showing the frontal and side views and incorporating a measurement scale to allow the magnification factor to be calculated. The resection level is marked and a series of precise measurements taken of key anatomical points. From these, the critical dimensions for the custom implant can be calculated and passed to a draughtsman who produces the engineering drawings required by the workshop. These and other dimensions are also used to produce a schematic drawing for the surgeon’s reference during the operation to insert the implant. The skills This paper was presented at Effective CADCAM 91 conference held in Coventry on 6 7 November 1991. The M S was received on 4 February 1992 and was accepted for publication on 13 M a y 1992.
H00392 Q IMechE 1992
Earlier studies have concluded that the use of computerized analysis systems directly to read and interpret medical images [X-rays and CT (computerized tomography) scans] is fraught with problems caused by poor and unreliable image quality and the variation in techniques used by different centres (1, 2). For this reason the decision has been made to continue to employ the expert assistance of human intervention in entering the patient’s bone dimensions into the computer program. The precise marking up required in the past on the measurement X-rays has been reduced to the indication of the bone’s major centre-line axis and the transection line at which the surgeon wishes to cut the bone. The X-rays are then placed on a large, backlit digitizing tablet and, in response to computer prompts, the major anatomical landmarks are idcntified, returning the relative X and Y coordinates to the
0954-4119/92 $3.00
+ .05
Downloaded from pih.sagepub.com at RMIT UNIVERSITY on July 20, 2015
Proc Instn Mech Engrs Vol 206
44
H V CRAWFORD, P S UNWIN AND P S WALKER
6months enables the implant to be extended by the surgeon to match the natural growth of the opposite limb.
3.3 SMILES
Fig. 1 The points on a frontal X-ray which would need to be digitized to provide data for the prosthesis design software for a proximal femur replacement prosthesis
computer. In order to allow the operator to concentrate on the X-ray the textual prompts on the computer screen are supported by synthesized spoken instructions and feed-back. Figure 1 shows the points that need to be digitized for the design of a distal femur and knee replacement implant.
The task of the design and production of some of the more common major implants has been simplified and speeded by the implementation of the Stanmore modular individualized lower extremity system (SMILES). This is a range of stock modular components which a statistical analysis of past implant designs has shown can be easily tailored to build into the majority of implant designs. Attempts have been made to develop modular sets of components that the surgeon can assemble into the required configuration at the time of the operation (3, 4), but these cannot be tailored quite as precisely for a particular patient and make no allowance for bone curvature or other special factors. The manufacture of the SMILES modular components is both sub-contracted and made in-house and in either case the savings in assembly time for the modular prosthesis is a considerable advantage. Where possible, the expert system software will base a prosthesis design on the availability of standard components, but will indicate the contra-indications where this is not appropriate.
3.4 CAD links
3.2 Design choices The full digitized coordinate data from the X-ray films are reduced to the salient measurements for processing within the system. Using rules provided by the experience of the human experts and deduced from our experience in the fields of bone growth and remodelling, strain distribution, surgical technique and implant design it may prompt for further information which may be based on surgeon preferences, the patient’s age and condition or other peripheral factors. If the system encountcrs any inconsistencies in the data or unusual anatomical conditions it will make recommendations but allow modification to its design decisions. The design rules currently entered into the knowledge base allow it to produce design recommendations for three major classes of implant which account for almost 60 per cent of the output: (a) proximal femoral replacements (top of femur and hip joint), (b) distal femora1 replacements (bottom of femur and knee joint), (c) proximal tibia replacements (top of tibia and knee joint). Proximal humeral replacements, which account for a further 8 per cent of the output, are obvious candidates for inclusion in the system. Work is also continuing on establishing the rules for design of a further category of implants (5 per cent) which are variants of these designs and are used for the treatment of young patients who require ‘growing’ or ‘extendible’ prostheses. Following the initial major surgery required to implant the prosthesis in the patient, a more minor operation every
The use of the modular components in the fabrication of the custom implants, where appropriate, has simplified the requirements of the engineering drawings supplied to the workshops. Where each element had to be precisely dimensioned and toleranced before, the prefinished items now only need to be identified by their critical dimensions and any custom modification required for a particular implant indicated. This is achieved by a customized program running within Schlumberger’s MacBravo CAD package which uses the implant design parameters generated by the expert system to construct the required drawings (Fig. 2).
3.5 Op drawings Similar data from the expert system detailing the implant parameters is fed to a database package for storage and for the generation of suitable annotated surgical operation drawings (Fig. 3). Further links are envisaged which will allow the automated production of packing lists and box labels in addition to adding the information to the existing unique database resource. This stores the details of patients and implants from 20 years of orthopaedic experience together with analysis of retrieved implants which records failure modes and wear patterns.
4 CUSTOM HIPS
A further application of computer technology to the production of orthopaedic implants is the hip design workstation. Developed in conjunction with Zimmer,
Part H: Journal of Engineering in Medicine
Q IMechE 1992 Downloaded from pih.sagepub.com at RMIT UNIVERSITY on July 20, 2015
THE CADCAM CONTRIBUTION TO CUSTOMIZED ORTHOPAEDIC IMPLANTS
I
28.00
45
Stem diameter tapering from 13.0 down to 11.5
Fig. 2 CAD image of a distal femur and knee replacement produced to show dimensions required to build the implant from modular components
one of the large multi-national orthopaedic suppliers, this VAX-based system also uses data digitized from measurement X-rays, but for the production of custommade, non-cemented hip replacements. The majority of total hip replacements (THRs) are fitted by opening out an oversize cavity down the femoral canal and fitting in a n off-the-shelf implant held in place with acrylic cement. While these have offered good medium-term fixation of the implant with rapid mobilization of the patient, the long-term performance is compromised by fatigue failure of the cement followed by loosening, especially in young patients. Many modern designs of THR rely on a closer fit between stem and bone and are used without cement. Early indications are that, if well fitted, these provide a better long-term prognosis (5). Any mass-produced press-fit THR will clearly be based on assumptions about average anatomy, but a better fit should be achieved with a custom implant. Custom implants are also indicated in cases of congenital deformity where unusual anatomy can be corrected by, for example, adjusting femoral head positions. Figure 4 shows two views of a standard implant together with
the medial view of one designed for a patient requiring correction of the femoral head position. 4.1 Hip design workstation
The Stanmore hip design workstation has, at its core, a datafile representing the canal profiles of a n average femur, based on measurements taken from sectioned femurs and CAT (computerized axial tomography) scan images (6). Data collected by digitizing the canal profiles of a particular patient’s measurement X-rays (frontal and lateral) are used to scale the average profiles in order to predict the full three-dimensional canal shape for the patient. The specially written stem design program then simulates the surgical preparation of the femur with reamers and rasps, to match key geometrical parameters of the femur. The program then generates a stem which is a close proximal fit, a clearance fit in the mid region and a sliding fit distally. User-selected variables allow for appropriate neck cut, stem length, closeness of fit, collar type and anteversion angle. The final implant shape is held as an array of 100 sections
% Bone replaced: 57 % Stem length/bone length: 308
(a)
Fig. 3 Drawing showing information required by a surgeon at operation to insert a proximal femoral replacement prosthesis. This is created by the database program
which stores implant data
Fig. 4
(b)
(C)
Showing (a) front and (b) medial views of a standard implant together with (c) the medial view of one designed for a patient requiring correction of the femoral head position Proc Instn Mech Engrs Vol 206
Q IMechE 1992 Downloaded from pih.sagepub.com at RMIT UNIVERSITY on July 20, 2015
H V CRAWFORD, P S UNWIN AND P S WALKER
46
through the implant, with 40 points around each section defined by their Cartesian coordinates. New algorithms are currently being written for more accurate canal shape prediction by using data from selected CT scan slices. 4.2 CNC
A CNC tool path file to machine the implant shape from titanium 318 alloy (Ti, A1 6%, V 4%) is generated by another custom FORTRAN program. Running on a four-axis Fadal VMC40 mill the machining process is carried out in two stages: an initial helical roughing cut followed by a finishing cut with a ball-ended cutter. Some hand finishing and polishing is required and a separate spherical head is fitted to provide the articulation with the acetabular cup in the pelvis. One advantage of the standard off-the-shelf implant designs is the wide range of tooling available to the surgeon to assist in fitting the implant. With custommade implants, custom-made tooling is required, and work is continuing to develop a simple but effective method for producing rasps to match the implants at a cost that is not prohibitive. With recent improvements in the cutting algorithm the time to machine a typical hip has been reduced from 6 to under 3 hours. 5 CONCLUSIONS
Computer technology is coming to the aid of the medical profession in many of its disciplines and this includes the field of orthopaedics. Accurately matching the prosthesis design to the patient increases the probability of survival, while the reductions in design and
manufacture time for custom prostheses make them viable alternatives to the modular kits. The prototype ‘expert system’ has proved the principles. Conversion to a language better suited to decision making and deductive reasoning will allow more power to be added without sacrificing speed of response. The use of computer-designed implants is improving the quality of life for many patients.
ACKNOWLEDGEMENT
The application of the hip design workstation to CADCAM hips for juvenile rheumatoid arthritis cases is supported by the Arthritis and Rheumatism Council.
REFERENCES 1 Ellam, S. V. and Maisey, M. N. Knowledge based system to assist in medical image interpretation: design and evaluation methodology. Sixth Technical Conference of the British Computer Society, Brighton, 1988. 2 Garland, L. H. Studies on the accuracy of diagnostic procedures. Am. J. Roentgenology, 1959,97 (4), 901-905. 3 Bos, G., Sim, F. H., Pritcbard, D. J., Sbives, T. C., Rock, M., Askew, L. and Chao, E. Y. S. Prosthetic proximal humeral replacement: the Mayo Clinic experience. In Limb salvage in musculoskeletal oncology, Orlando, 1985, pp. 61-72 (Churchill Livingstone). 4 Ritschl, P., Braun, O., Pongracz, N., Eyb, R., Ramach, W. and Kotz, R. Modular reconstruction system for the lower extremity. In Limb suluuye in musculoskeletal oncology, Orlando, 1985, pp. 237-243 (Churchill Livingstone). 5 Bargar, W. Shape the implant to fit the patient. Clin. Orthop. Related Res., December 1989,249, 73-78. 6 Walker, P. S., Poss, R., Robertson, D. D., et al. Design analysis of press fit stems. Orthop. Related Sci., 1990,1, 75-85.
Part H: Journal of Engineering in Medicine
@ IMechE 1992 Downloaded from pih.sagepub.com at RMIT UNIVERSITY on July 20, 2015
