- Email: [email protected]
Development of a cancellous bone structural model by stereolithography for ultrasound characterisation of the calcaneus
PII: S1350-4533(97)00027-1
Med. Eng. Phys. Vol. 19, No. 7, pp. 599–604, 1997 1997 IPEM. Published by Elsevier Science Ltd Printed in Great Britain 1350–4533/97 $17.00 + 0.00
Development of a cancellous bone structural model by stereolithography for ultrasound characterisation of the calcaneus C. M. Langton*†, M. A. Whitehead‡, D. K. Langton† and G. Langley† *Department of Medical Physics, University of Hull, Hull, U.K.; †Department of Medical Physics, Royal Hull Hospitals NHS Trust, Hull, U.K.; ‡Health Research Institute, Sheffield Hallam University, Sheffield, U.K. Received 21 August 1996, accepted 28 May 1997
ABSTRACT A novel method for the development of a user-defined structural model simulating cancellous bone of the human calcaneus is described using stereolithography (SL). The digital image of a cancellous bone section was modified by skeletonisation and dilation to produce a structural model of uniform wall thickness, determined by the resolution of the stereolithography system. Six SL models were produced using the same data file. The SL models were assessed using the McCue CUBAclinical ultrasound bone densitometer. The broadband ultrasound attenuation (BUA) and velocity (VOS) values obtained were commensurate with the commercial phantom provided with the CUBAclinical system. The intra- and inter-sample variability for the six SL models were similar at 5% for BUA and 2.5% for VOS. Stereolithography offers the potential to firstly, simulate perforation and thinning of cancellous bone associated with osteoporosis; and secondly, to evaluate the dependence of ultrasonic and mechanical parameters upon cancellous bone structure. 1997 IPEM. Published by Elsevier Science Ltd Keywords: Bone structure, ultrasound, stereolithography Med. Eng. Phys., 1997, Vol. 19, 599–604, October
1. INTRODUCTION Ultrasound measurements of velocity and broadband ultrasound attenuation (BUA) of the human calcaneus are being widely assessed for their ability to identify osteoporotic changes and ultimately, identify the risk of fracture within an individual subject. A major attraction of ultrasound measurements is the dependence upon bone structure in addition to bone mineral density (BMD), as measured by the established ionising radiation techniques such as dual energy X-ray absorptiometry (DXA) and quantitative computed tomography (QCT). The evidence for the structural dependence is based upon in vivo sitematched studies of the calcaneus1,2 and in vitro studies of equine3 and bovine4 bone samples exhibiting significant anisotropy in cubic samples. In both in vivo and in vitro studies, the statistical variance explained by bone mineral density has been of the order of only 50%, the remaining variance attributed to structural dependence. A significant limitation of the in vivo site-matched studies is that they have related ultrasound measurements, predominantly BUA, to BMD and have not incorporated a direct assessment of bone structure. In vitro studies on human cancellous bone5,6 have been
undertaken on an elderly range of cadaveric specimens. In all cases, a significant difficulty in analysing in vitro bone samples is that the structural parameters have to be elucidated by destructive means, typically in the form of slice histomorphometry. In addition, the human in vitro samples tend to be from an elderly population and therefore may be of limited structural variation compared with the full population age range. The development of a physical model of cancellous bone whose structure may be controlled would provide significant advantages over the study of in vitro samples and would also enable us to define more exactly the relationship between ultrasound parameters of velocity and broadband attenuation with cancellous bone structure. Previous studies have investigated the potential of developing physical models of cancellous bone. A porous mimic of cancellous bone has been produced7 consisting of roughly cubic gelatine granules of approximately 1 mm dimension dispersed in an epoxy resin. The gelatine was subsequently removed and replaced by a vegetable oil, thus providing a porous material. The epoxy resin and oil were chosen to closely match the mechanical and ultrasonic properties of bone and marrow, respectively. However, it is not possible to control
Development of a cancellous bone structural model: C. M. Langton et al.
the exact structure of this model, only the average pore size and porosity. An alternative approach8 considered an array of holes drilled into a perspex block. By enlarging the diameters of the holes, the porosity of the cancellous bone mimic was increased. Limitations of this approach are, that the pore size is limited to those available by mechanical drilling (typically 1 mm minimum), the process of drilling a large number of holes is timeconsuming, and the samples produced consisted of an array of parallel capillary tubes with only two potential degrees of anisotropy. 2. STEREOLITHOGRAPHY TECHNIQUE Stereolithography (SL) is a form of rapid prototyping that allows complex solid objects to be manufactured directly from 3D computer models in the form of successive layers of light-cured resin. There are essentially two stages to SL, design and manufacture, illustrated in Figure 1. During the design stage the required object is initially created using standard 3D solid modelling techniques and then converted into the SL format
consisting of a series of thin slices. The SL manufacturing system consists of a vat of light-sensitive resin with an elevator and computer-controlled scanning laser. At the start of the process, the elevator is positioned just below the surface (typically 0.1 mm) of the liquid SL resin. The laser scanner “prints” the bottom layer onto the resin surface, which solidifies upon exposure to the laser beam. The elevator then moves down by an incremental distance and the SL resin is respread over the surface of the vat prior to the next scan. As successive solid layers are formed, they bond to produce a single solid object. When the model is completed the elevator is raised and the unused resin allowed to drain. The laser cures the SL resin to approximately 60%, the curing process is completed in an ultraviolet oven. The resolution of the SL process is governed by the laser spot size and the vertical movement of the elevator. Typically, the laser spot diameter will be better than 0.3 mm and the elevator movement resolution will be about 2.5 m. The recommended minimum layer thickness is 0.1 mm.
Figure 1 Diagrammatic representation of the stereolithography process consisting of design (a) and manufacture (b). The design stage consists of formation of a 3D image using CAD and conversion into the STL slice format. For manufacture, a cured resin replication of the 3D model is achieved by successive “printing” of the individual STL slices onto the light-sensitive resin by a computer-controlled scanning laser.
600
Development of a cancellous bone structural model: C. M. Langton et al.
3. METHODS AND RESULTS 3.1. Feasibility of creating stereolithographic structural models The density of SL resins available within the manufacturing industry range from 1.12 to 1.14 g cm⫺3, the most popular being XB5154 supplied by Ciba-Geigy (Duxford, UK) having a density of 1.12 g cm⫺3. A 20 mm cubic sample was initially manufactured by the SL process (SLA 250, 3D Systems, CA) using the XB5154 resin. The density of the cubic sample was calculated from its mass and dimensions. The elasticity (Young’s modulus) of the sample was measured using an Instron 1122 testing machine at a nominal strain rate of 0.0017 s⫺1. The elasticity was defined as the maximum slope of the stress–strain curve in the most linear portion of the pre-yield region. Ultrasound measurements of velocity and BUA were recorded using 12.5 mm diameter 1 MHz ultrasound transducers connected to the CUBAresearch system (McCue, Winchester, UK). Mechanical and ultrasonic data for human cortical bone and the XB5154 resin are shown in Table 1, noting that cancellous bone may be considered as a porous cortical bone structure interspaced with bone marrow. Although the mechanical properties of density and elasticity for the resin are approximately 60% of those for cortical bone, the ultrasound velocity and broadband attenuation are similar and, therefore, it is considered that resin XB5154 may be used as an ultrasonic surrogate for cortical bone. The temperature coefficient for BUA and velocity of the solid SL resin was not measured since it is widely accepted that the dominant temperature sensitivity would be caused by the liquid marrow mimic. The potential for stereolithography to be used as a structurally controlled cancellous bone mimic was initially investigated by producing a 3D rod lattice of 3 mm centres with rod diameters of 1 mm (70% porosity) representing healthy bone and 0.4 mm (95% porosity) representing osteoporotic bone, illustrated in Figure 2. 3.2. Development of the CAD model A computer-aided design (CAD) model of a calcaneal bone structure was created by digitising photographs of a cadaveric specimen embedded in a casting resin. A 20 mm diameter cylindrical core was removed Table 1 Comparison of mechanical and ultrasonic properties of cortical bone and SL resin. The cortical bone data are typical values from the literature, the SL resin data being obtained experimentally.
Density (g cm⫺3) Elasticity (MPa) Velocity (m s⫺1) BUA (dB MHz⫺1cm⫺1)
Human cortical bone
Resin XB 5154
1.8–2.0 17 3000–4000 2.8
1.12 9.91 2800 2.5
Figure 2 Simple rod lattice stereolithography models with rod diameters of 1 mm (70% porosity) and 0.4 mm (95% porosity) representing healthy and osteoporotic bone.
from the left calcaneus of an 80-year-old female cadaver, obtained from the anatomical region, measured in vivo by ultrasound bone densitometers. The core was cut into two semi-cylindrical samples and defatted. Soaking in a sodium hydroxide solution bleached the bone tissue to aid photographic contrast. One of the semi-cylindrical samples was then embedded in a casting resin. A wide range of casting resins are available with various numbers of components, viscosity, curing time, hardness etc. Our priorities were sufficient curing time, viscosity low enough to completely fill the voids within the cancellous bone under vacuum and the ability to be coloured black to aid photographic contrast. Acrulite casting resin (Rubert and Co, Cheadle, UK) was chosen consisting of a liquid and powder. The mixing ratio (not being critical, affecting only the curing time) was 2 parts powder by weight to 4 parts liquid with 0.3 parts of black graphite powder for colouring. An aluminium mould was constructed into which the semi-cylindrical sample was inserted with the cut surface exposed. The casting resin was added with sufficient vacuum applied to remove any air bubbles, with care taken not to have too high a vacuum, which would have resulted in boiling of the resin mixture. A microtome was used to provide a clean surface of the resin-impregnated cancellous bone sample. Photographs of the surface were taken with a 35 mm SLR camera positioned directly over the sample with a small lamp attached to provide an even light distribution. The calcaneal section photograph was scanned into a bitmap format which was processed and analysed (Imagery Software, University of Leeds, UK) with calibration performed using horizontal and vertical scales. The bitmap image was first converted to binary format from grey level by thresholding, and then smoothed and skeletonised. Skeletonisation thins any line to 1 pixel thickness (0.05 mm in this case) without breaking. Additional lines were added to the image manually to join up the skeleton to form a complete structure. This was then dilated to a uniform thickness of 0.35 mm, which is greater than the
601
Development of a cancellous bone structural model: C. M. Langton et al.
stereolithography resolution. The image analysis process from the cancellous bone image through skeletonisation, modification and dilation is illustrated in Figure 3. The bitmap image was converted into a vector format (Encapsulated Postscript, EPS), using “Coral Trace” (Corel Corporation, Canada). “Coral Draw” was then used to convert the EPS image into an AutoCad.DXF format which could then be imported into most CAD packages. The DXF format was imported into Unigraphics, a 3D modelling package. A difficulty was encountered where the conversion into vector format created a very high number of entities for each object, requiring high memory allocation. The number of entities was reduced manually in AutoCad, although time-consuming, it made modelling much more efficient. The entity count was reduced from approximately
1200 to 200 with no noticeable loss of resolution. The reduced DXF format image was mirrored about both axes and extruded to produce a 3D model 34 mm × 34 mm × 32 mm in dimension. This included an outer jacket around the four sides to strengthen the structure and to mimic a cortical shell. This was initially 2 mm in thickness, subsequently reduced to 0.5 mm to prevent the occurrence of multiple echo artefacts when measured with ultrasound. The DXF files were finally converted into STL format ready for production. The porosity of the final structure was analysed within the Imagery software and calculated to be 70%, this is slightly lower than typical human cancellous bone values, ranging from 75 to 85%. Ten models of the same design were produced from resin XB5154, taking about 40 h of uninterrupted scanning with the SL system. Once built,
Figure 3 The image analysis process based upon the calcaneal cancellous bone image. The bitmap image of the cut calcaneal section (a) was initially skeletonised to a single pixel thickness (b) which was manually modified to provide a complete structure with no free segments and dilated to a uniform thickness of 0.35 mm (c). This was converted into vector format, mirrored about both axes and extruded to form a solid model with a cortical shell (d).
602
Development of a cancellous bone structural model: C. M. Langton et al.
the models were drained and any excess resin removed by careful use of an airjet and then placed in an ultraviolet oven to complete the curing process. The final stereolithography model is illustrated in Figure 4. 3.3. Bone marrow mimic The material used as the marrow mimic will affect the overall ultrasonic properties of the cancellous bone model and several materials were considered, namely, silicone rubber, water, synthetic oil and castor oil. Silicone rubber has the advantage of forming a very stable container not requiring an additional support cell. Two SL models were vacuum cast into 10 mm oversized blocks using a clear silicone casting rubber (KE1300T). The two models were tested ultrasonically and exhibited a very high attenuation at 0.2 MHz of 40 dB. This was surprising, as previous tests carried out on pure samples of SL resin and silicone rubber had not shown these high attenuation values, indicating the presence of air pockets within the model. Water provides similar BUA and velocity values to bone marrow. Two further SL models were immersed in water but were found to be hygroscopic, resulting in a softening and distortion of the resin material. Castor and synthetic oils also have similar BUA and velocity values to bone marrow but have an inherent temperature sensitivity. In order to investigate the magnitude of this, samples of each liquid were placed in a perspex cell (with 1 mm thickness “ultrasound windows” incorporated) between two 19 mm diameter 1 MHz broadband ultrasound transducers. This arrangement was placed in a temperature controlled water bath. Ultrasound measurements of velocity and BUA were recorded using the CUBAresearch system (McCue plc) over a temperature range of 20–45°C in increments of 5°C. The BUA temperature sensitivity was ⫺ 0.4 dB MHz⫺1°C⫺1 for castor oil (range 75.0–64.6 dB MHz⫺1 ) and + 0.71 dB MHz⫺1°C⫺1
Figure 4 Photograph of the final stereolithography model based upon the cancellous bone structure of the calcaneus.
for synthetic oil ( range 69.4–88.0 dB MHz ⫺1 ). The velocity temperature sensitivity was ⫺ 4.8 m s⫺1°C ⫺1 for castor oil (range 1792– 1666 m s⫺1) and ⫺ 4.3 m s⫺1°C⫺1 for synthetic oil (range 1715–1594 m s⫺1). Since castor oil has the smallest temperature coefficient for BUA, being almost half that of the synthetic oil, with approximately equal velocity coefficient, it was chosen as the marrow mimic material. 3.4. Ultrasonic assessment The remaining six stereolithography models were housed in a purpose-built cell consisting of a perspex ring with two polyethylene sheet windows covering the two surfaces of the model interrogated by the ultrasound beam. The ring had two access holes to enable the castor oil to be inserted and air removed by degassing. The models were measured using the CUBAclinical II (McCue plc) ultrasound bone densitometer, providing BUA and velocity (VOS) data of the calcaneus. This incorporates two 19 mm diameter 1 MHz centre frequency transducers. Both transducers incorporate a 10 mm thickness silicone pad to accommodate variation in the surface shape of the heel. In order to position the stereolithography models co-axially between the ultrasound transducers, a cell holder was designed and manufactured (Figure 5), whose external shape conformed with the footplate of the CUBAclinical system. BUA and VOS measurements were recorded for the six SL models (SL1-6), along with the standard phantom supplied with the CUBAclinical system, consisting of a series of porous polyethylene disks filled with water. Thirty-one measurements of each phantom were recorded over a period of 108 days, shown in Table 2. 4. DISCUSSION The BUA and VOS mean values obtained for the SL models are well within the range of human calcaneal data (BUA range 20–120 dB MHz⫺1,
Figure 5 Photograph of the stereolithography model encapsulated in castor oil and the sample holder for accommodation in the McCue CUBAclinical ultrasound bone densitometer.
603
Development of a cancellous bone structural model: C. M. Langton et al.
VOS range 1300–1900 m s⫺1) obtained with the CUBAclinical system. The SL model data are also similar to the results obtained using the CUBAclinical commercial phantom. The BUA precision values (CV%) for the individual SL models are similar to the commercial phantom, although the velocity CV% values are higher by a factor of almost 5. Of note, CUBAclinical in vivo precision values 9 correspond closely to those obtained with the commercial phantom. The inter-sample precision for the six SL models is 5% for BUA and 2.5% for velocity, both obtained using the rms (root mean square) value for the standard deviation. The inter-sample variability for the SL models for both BUA and VOS is similar to that obtained within a single SL model, indicating the high degree of reproducibility achievable with stereolithography. We note, however, that the results obtained with model SL4 deviated significantly from those of the other SL models. Upon visual inspection of the SL models, air-filled pockets were observed in some of the narrow diameter capillaries. We propose that these were trapped when undrained liquid resin solidified during the ultraviolet oven curing. Sample SL4 exhibited the greatest number of these artefacts. Future development with the SL process should consider a minimum pore diameter of 1 mm. The increased VOS precision error for the SL models compared with the CubaClinical commercial phantom may be attributed to the incorporation of oil rather than water for the marrow mimic, resulting in a temperature dependence, also observed in other resin/oil phantoms7. It is feasible that an improvement in velocity precision could be achieved if a temperature-dependent correction factor was incorporated, perhaps aided by a liquid crystal thermometer mounted on the surface of the phantom. 5. CONCLUSION A structural model for cancellous bone has been developed using stereolithography. The SL model has been assessed using the McCue CUBAclinical ultrasound bone densitometer. The BUA and velocity values obtained were commensurate with the commercial phantom. The intra- and intersample variability for the SL model was similar at 5% for BUA and 2.5% for velocity. The model has significant potential for assessing both the ultrasonic and mechanical dependence upon cancellous bone structure. For example, the initial CAD image may be modified to exhibit both perforation and thinning observed in type I and type II osteoporotic changes. The authors know of no other technique which may
604
Table 2 BUA and VOS mean and precision data for the McCue phantom and stereolithography models over a 108-day period Samples
CubaClinical Phantom SL1 SL2 SL3 SL4 SL5 SL6
BUA
VOS
Mean
CV%
Mean
CV%
66.64
4.55
1718
0.50
67.00 59.00 67.94 53.03 65.74 67.29
4.15 5.95 5.76 5.77 4.94 3.37
1855 1853 1840 1690 1843 1864
2.46 2.41 2.54 2.44 2.54 2.58
enable the structural complexity of a porous sample to be defined. ACKNOWLEDGEMENTS The authors acknowledge the EC BIOMED 1 Concerted Action Assessment of Bone Quality in Osteoporosis for their support and Professor J. D. Currey, University of York, for assistance with the mechanical analysis of the solid stereolithography sample. REFERENCES 1. Gluer, C. C., Vahlensieck, M., Faulkner, K. G., Engelke, K., Black, D. and Genant, H. K., Site-matched calcaneal measurements of broad-band ultrasound attentuation and single x-ray absorptiometry: do they measure different skeleton properties?. J. Bone Miner. Res., 1992, 7, 1071– 1079. 2. Waud, C. E., Lew, R. and Baran, D. T., The relationship between ultrasound and densitometric measurements of bone mass at the calcaneus in women. Calcif. Tissue Int., 1992, 51, 415–418. 3. Langton, C. M., Riggs, C. M. and Evans, G. P., Pathway of ultrasound waves in the equine third metacarpal bone. J. Biomed. Eng., 1990, 13, 113–118. 4. Gluer, C. C., Wu, C. Y. and Genant, H. K., Broadband ultrasound attenuation signals depend on trabecular orientation: an in vitro study. Osteo. Int., 1993, 3, 185–191. 5. McKelvie, M. L., Fordham, J., Clifford, C. and Palmer, S. B., In vitro comparison of quantitative computed tomography and broadband ultrasonic attenuation of trabecular bone. Bone, 1989, 10, 101–104. 6. Langton, C. M., Njeh, F., Hodgskinson, R. and Currey, J. D., Prediction of mechanical properties of the human calcaneus by broadband ultrasonic attenuation. Bone, 1996, 18, 495–503. 7. Clarke, A. J., Evans, J. A., Truscott, J. G., Milner, R. and Smith, M. A., A phantom for quantitative ultrasound of trabecular bone. Phys. Med. Biol., 1994, 39, 1677–1687. 8. Njeh C. F. The dependence of ultrasound, velocity and attenuation on material properties of cancellous bone. Ph.D. Thesis. Sheffield Hallam University, 1995. 9. Langton, C. M., ZSD: a universal parameter for precision in the ultrasonic assessment of osteoporosis. Physiol. Meas., 1997, 18, 67–72.
Med. Eng. Phys. Vol. 19, No. 7, pp. 599–604, 1997 1997 IPEM. Published by Elsevier Science Ltd Printed in Great Britain 1350–4533/97 $17.00 + 0.00
Development of a cancellous bone structural model by stereolithography for ultrasound characterisation of the calcaneus C. M. Langton*†, M. A. Whitehead‡, D. K. Langton† and G. Langley† *Department of Medical Physics, University of Hull, Hull, U.K.; †Department of Medical Physics, Royal Hull Hospitals NHS Trust, Hull, U.K.; ‡Health Research Institute, Sheffield Hallam University, Sheffield, U.K. Received 21 August 1996, accepted 28 May 1997
ABSTRACT A novel method for the development of a user-defined structural model simulating cancellous bone of the human calcaneus is described using stereolithography (SL). The digital image of a cancellous bone section was modified by skeletonisation and dilation to produce a structural model of uniform wall thickness, determined by the resolution of the stereolithography system. Six SL models were produced using the same data file. The SL models were assessed using the McCue CUBAclinical ultrasound bone densitometer. The broadband ultrasound attenuation (BUA) and velocity (VOS) values obtained were commensurate with the commercial phantom provided with the CUBAclinical system. The intra- and inter-sample variability for the six SL models were similar at 5% for BUA and 2.5% for VOS. Stereolithography offers the potential to firstly, simulate perforation and thinning of cancellous bone associated with osteoporosis; and secondly, to evaluate the dependence of ultrasonic and mechanical parameters upon cancellous bone structure. 1997 IPEM. Published by Elsevier Science Ltd Keywords: Bone structure, ultrasound, stereolithography Med. Eng. Phys., 1997, Vol. 19, 599–604, October
1. INTRODUCTION Ultrasound measurements of velocity and broadband ultrasound attenuation (BUA) of the human calcaneus are being widely assessed for their ability to identify osteoporotic changes and ultimately, identify the risk of fracture within an individual subject. A major attraction of ultrasound measurements is the dependence upon bone structure in addition to bone mineral density (BMD), as measured by the established ionising radiation techniques such as dual energy X-ray absorptiometry (DXA) and quantitative computed tomography (QCT). The evidence for the structural dependence is based upon in vivo sitematched studies of the calcaneus1,2 and in vitro studies of equine3 and bovine4 bone samples exhibiting significant anisotropy in cubic samples. In both in vivo and in vitro studies, the statistical variance explained by bone mineral density has been of the order of only 50%, the remaining variance attributed to structural dependence. A significant limitation of the in vivo site-matched studies is that they have related ultrasound measurements, predominantly BUA, to BMD and have not incorporated a direct assessment of bone structure. In vitro studies on human cancellous bone5,6 have been
undertaken on an elderly range of cadaveric specimens. In all cases, a significant difficulty in analysing in vitro bone samples is that the structural parameters have to be elucidated by destructive means, typically in the form of slice histomorphometry. In addition, the human in vitro samples tend to be from an elderly population and therefore may be of limited structural variation compared with the full population age range. The development of a physical model of cancellous bone whose structure may be controlled would provide significant advantages over the study of in vitro samples and would also enable us to define more exactly the relationship between ultrasound parameters of velocity and broadband attenuation with cancellous bone structure. Previous studies have investigated the potential of developing physical models of cancellous bone. A porous mimic of cancellous bone has been produced7 consisting of roughly cubic gelatine granules of approximately 1 mm dimension dispersed in an epoxy resin. The gelatine was subsequently removed and replaced by a vegetable oil, thus providing a porous material. The epoxy resin and oil were chosen to closely match the mechanical and ultrasonic properties of bone and marrow, respectively. However, it is not possible to control
Development of a cancellous bone structural model: C. M. Langton et al.
the exact structure of this model, only the average pore size and porosity. An alternative approach8 considered an array of holes drilled into a perspex block. By enlarging the diameters of the holes, the porosity of the cancellous bone mimic was increased. Limitations of this approach are, that the pore size is limited to those available by mechanical drilling (typically 1 mm minimum), the process of drilling a large number of holes is timeconsuming, and the samples produced consisted of an array of parallel capillary tubes with only two potential degrees of anisotropy. 2. STEREOLITHOGRAPHY TECHNIQUE Stereolithography (SL) is a form of rapid prototyping that allows complex solid objects to be manufactured directly from 3D computer models in the form of successive layers of light-cured resin. There are essentially two stages to SL, design and manufacture, illustrated in Figure 1. During the design stage the required object is initially created using standard 3D solid modelling techniques and then converted into the SL format
consisting of a series of thin slices. The SL manufacturing system consists of a vat of light-sensitive resin with an elevator and computer-controlled scanning laser. At the start of the process, the elevator is positioned just below the surface (typically 0.1 mm) of the liquid SL resin. The laser scanner “prints” the bottom layer onto the resin surface, which solidifies upon exposure to the laser beam. The elevator then moves down by an incremental distance and the SL resin is respread over the surface of the vat prior to the next scan. As successive solid layers are formed, they bond to produce a single solid object. When the model is completed the elevator is raised and the unused resin allowed to drain. The laser cures the SL resin to approximately 60%, the curing process is completed in an ultraviolet oven. The resolution of the SL process is governed by the laser spot size and the vertical movement of the elevator. Typically, the laser spot diameter will be better than 0.3 mm and the elevator movement resolution will be about 2.5 m. The recommended minimum layer thickness is 0.1 mm.
Figure 1 Diagrammatic representation of the stereolithography process consisting of design (a) and manufacture (b). The design stage consists of formation of a 3D image using CAD and conversion into the STL slice format. For manufacture, a cured resin replication of the 3D model is achieved by successive “printing” of the individual STL slices onto the light-sensitive resin by a computer-controlled scanning laser.
600
Development of a cancellous bone structural model: C. M. Langton et al.
3. METHODS AND RESULTS 3.1. Feasibility of creating stereolithographic structural models The density of SL resins available within the manufacturing industry range from 1.12 to 1.14 g cm⫺3, the most popular being XB5154 supplied by Ciba-Geigy (Duxford, UK) having a density of 1.12 g cm⫺3. A 20 mm cubic sample was initially manufactured by the SL process (SLA 250, 3D Systems, CA) using the XB5154 resin. The density of the cubic sample was calculated from its mass and dimensions. The elasticity (Young’s modulus) of the sample was measured using an Instron 1122 testing machine at a nominal strain rate of 0.0017 s⫺1. The elasticity was defined as the maximum slope of the stress–strain curve in the most linear portion of the pre-yield region. Ultrasound measurements of velocity and BUA were recorded using 12.5 mm diameter 1 MHz ultrasound transducers connected to the CUBAresearch system (McCue, Winchester, UK). Mechanical and ultrasonic data for human cortical bone and the XB5154 resin are shown in Table 1, noting that cancellous bone may be considered as a porous cortical bone structure interspaced with bone marrow. Although the mechanical properties of density and elasticity for the resin are approximately 60% of those for cortical bone, the ultrasound velocity and broadband attenuation are similar and, therefore, it is considered that resin XB5154 may be used as an ultrasonic surrogate for cortical bone. The temperature coefficient for BUA and velocity of the solid SL resin was not measured since it is widely accepted that the dominant temperature sensitivity would be caused by the liquid marrow mimic. The potential for stereolithography to be used as a structurally controlled cancellous bone mimic was initially investigated by producing a 3D rod lattice of 3 mm centres with rod diameters of 1 mm (70% porosity) representing healthy bone and 0.4 mm (95% porosity) representing osteoporotic bone, illustrated in Figure 2. 3.2. Development of the CAD model A computer-aided design (CAD) model of a calcaneal bone structure was created by digitising photographs of a cadaveric specimen embedded in a casting resin. A 20 mm diameter cylindrical core was removed Table 1 Comparison of mechanical and ultrasonic properties of cortical bone and SL resin. The cortical bone data are typical values from the literature, the SL resin data being obtained experimentally.
Density (g cm⫺3) Elasticity (MPa) Velocity (m s⫺1) BUA (dB MHz⫺1cm⫺1)
Human cortical bone
Resin XB 5154
1.8–2.0 17 3000–4000 2.8
1.12 9.91 2800 2.5
Figure 2 Simple rod lattice stereolithography models with rod diameters of 1 mm (70% porosity) and 0.4 mm (95% porosity) representing healthy and osteoporotic bone.
from the left calcaneus of an 80-year-old female cadaver, obtained from the anatomical region, measured in vivo by ultrasound bone densitometers. The core was cut into two semi-cylindrical samples and defatted. Soaking in a sodium hydroxide solution bleached the bone tissue to aid photographic contrast. One of the semi-cylindrical samples was then embedded in a casting resin. A wide range of casting resins are available with various numbers of components, viscosity, curing time, hardness etc. Our priorities were sufficient curing time, viscosity low enough to completely fill the voids within the cancellous bone under vacuum and the ability to be coloured black to aid photographic contrast. Acrulite casting resin (Rubert and Co, Cheadle, UK) was chosen consisting of a liquid and powder. The mixing ratio (not being critical, affecting only the curing time) was 2 parts powder by weight to 4 parts liquid with 0.3 parts of black graphite powder for colouring. An aluminium mould was constructed into which the semi-cylindrical sample was inserted with the cut surface exposed. The casting resin was added with sufficient vacuum applied to remove any air bubbles, with care taken not to have too high a vacuum, which would have resulted in boiling of the resin mixture. A microtome was used to provide a clean surface of the resin-impregnated cancellous bone sample. Photographs of the surface were taken with a 35 mm SLR camera positioned directly over the sample with a small lamp attached to provide an even light distribution. The calcaneal section photograph was scanned into a bitmap format which was processed and analysed (Imagery Software, University of Leeds, UK) with calibration performed using horizontal and vertical scales. The bitmap image was first converted to binary format from grey level by thresholding, and then smoothed and skeletonised. Skeletonisation thins any line to 1 pixel thickness (0.05 mm in this case) without breaking. Additional lines were added to the image manually to join up the skeleton to form a complete structure. This was then dilated to a uniform thickness of 0.35 mm, which is greater than the
601
Development of a cancellous bone structural model: C. M. Langton et al.
stereolithography resolution. The image analysis process from the cancellous bone image through skeletonisation, modification and dilation is illustrated in Figure 3. The bitmap image was converted into a vector format (Encapsulated Postscript, EPS), using “Coral Trace” (Corel Corporation, Canada). “Coral Draw” was then used to convert the EPS image into an AutoCad.DXF format which could then be imported into most CAD packages. The DXF format was imported into Unigraphics, a 3D modelling package. A difficulty was encountered where the conversion into vector format created a very high number of entities for each object, requiring high memory allocation. The number of entities was reduced manually in AutoCad, although time-consuming, it made modelling much more efficient. The entity count was reduced from approximately
1200 to 200 with no noticeable loss of resolution. The reduced DXF format image was mirrored about both axes and extruded to produce a 3D model 34 mm × 34 mm × 32 mm in dimension. This included an outer jacket around the four sides to strengthen the structure and to mimic a cortical shell. This was initially 2 mm in thickness, subsequently reduced to 0.5 mm to prevent the occurrence of multiple echo artefacts when measured with ultrasound. The DXF files were finally converted into STL format ready for production. The porosity of the final structure was analysed within the Imagery software and calculated to be 70%, this is slightly lower than typical human cancellous bone values, ranging from 75 to 85%. Ten models of the same design were produced from resin XB5154, taking about 40 h of uninterrupted scanning with the SL system. Once built,
Figure 3 The image analysis process based upon the calcaneal cancellous bone image. The bitmap image of the cut calcaneal section (a) was initially skeletonised to a single pixel thickness (b) which was manually modified to provide a complete structure with no free segments and dilated to a uniform thickness of 0.35 mm (c). This was converted into vector format, mirrored about both axes and extruded to form a solid model with a cortical shell (d).
602
Development of a cancellous bone structural model: C. M. Langton et al.
the models were drained and any excess resin removed by careful use of an airjet and then placed in an ultraviolet oven to complete the curing process. The final stereolithography model is illustrated in Figure 4. 3.3. Bone marrow mimic The material used as the marrow mimic will affect the overall ultrasonic properties of the cancellous bone model and several materials were considered, namely, silicone rubber, water, synthetic oil and castor oil. Silicone rubber has the advantage of forming a very stable container not requiring an additional support cell. Two SL models were vacuum cast into 10 mm oversized blocks using a clear silicone casting rubber (KE1300T). The two models were tested ultrasonically and exhibited a very high attenuation at 0.2 MHz of 40 dB. This was surprising, as previous tests carried out on pure samples of SL resin and silicone rubber had not shown these high attenuation values, indicating the presence of air pockets within the model. Water provides similar BUA and velocity values to bone marrow. Two further SL models were immersed in water but were found to be hygroscopic, resulting in a softening and distortion of the resin material. Castor and synthetic oils also have similar BUA and velocity values to bone marrow but have an inherent temperature sensitivity. In order to investigate the magnitude of this, samples of each liquid were placed in a perspex cell (with 1 mm thickness “ultrasound windows” incorporated) between two 19 mm diameter 1 MHz broadband ultrasound transducers. This arrangement was placed in a temperature controlled water bath. Ultrasound measurements of velocity and BUA were recorded using the CUBAresearch system (McCue plc) over a temperature range of 20–45°C in increments of 5°C. The BUA temperature sensitivity was ⫺ 0.4 dB MHz⫺1°C⫺1 for castor oil (range 75.0–64.6 dB MHz⫺1 ) and + 0.71 dB MHz⫺1°C⫺1
Figure 4 Photograph of the final stereolithography model based upon the cancellous bone structure of the calcaneus.
for synthetic oil ( range 69.4–88.0 dB MHz ⫺1 ). The velocity temperature sensitivity was ⫺ 4.8 m s⫺1°C ⫺1 for castor oil (range 1792– 1666 m s⫺1) and ⫺ 4.3 m s⫺1°C⫺1 for synthetic oil (range 1715–1594 m s⫺1). Since castor oil has the smallest temperature coefficient for BUA, being almost half that of the synthetic oil, with approximately equal velocity coefficient, it was chosen as the marrow mimic material. 3.4. Ultrasonic assessment The remaining six stereolithography models were housed in a purpose-built cell consisting of a perspex ring with two polyethylene sheet windows covering the two surfaces of the model interrogated by the ultrasound beam. The ring had two access holes to enable the castor oil to be inserted and air removed by degassing. The models were measured using the CUBAclinical II (McCue plc) ultrasound bone densitometer, providing BUA and velocity (VOS) data of the calcaneus. This incorporates two 19 mm diameter 1 MHz centre frequency transducers. Both transducers incorporate a 10 mm thickness silicone pad to accommodate variation in the surface shape of the heel. In order to position the stereolithography models co-axially between the ultrasound transducers, a cell holder was designed and manufactured (Figure 5), whose external shape conformed with the footplate of the CUBAclinical system. BUA and VOS measurements were recorded for the six SL models (SL1-6), along with the standard phantom supplied with the CUBAclinical system, consisting of a series of porous polyethylene disks filled with water. Thirty-one measurements of each phantom were recorded over a period of 108 days, shown in Table 2. 4. DISCUSSION The BUA and VOS mean values obtained for the SL models are well within the range of human calcaneal data (BUA range 20–120 dB MHz⫺1,
Figure 5 Photograph of the stereolithography model encapsulated in castor oil and the sample holder for accommodation in the McCue CUBAclinical ultrasound bone densitometer.
603
Development of a cancellous bone structural model: C. M. Langton et al.
VOS range 1300–1900 m s⫺1) obtained with the CUBAclinical system. The SL model data are also similar to the results obtained using the CUBAclinical commercial phantom. The BUA precision values (CV%) for the individual SL models are similar to the commercial phantom, although the velocity CV% values are higher by a factor of almost 5. Of note, CUBAclinical in vivo precision values 9 correspond closely to those obtained with the commercial phantom. The inter-sample precision for the six SL models is 5% for BUA and 2.5% for velocity, both obtained using the rms (root mean square) value for the standard deviation. The inter-sample variability for the SL models for both BUA and VOS is similar to that obtained within a single SL model, indicating the high degree of reproducibility achievable with stereolithography. We note, however, that the results obtained with model SL4 deviated significantly from those of the other SL models. Upon visual inspection of the SL models, air-filled pockets were observed in some of the narrow diameter capillaries. We propose that these were trapped when undrained liquid resin solidified during the ultraviolet oven curing. Sample SL4 exhibited the greatest number of these artefacts. Future development with the SL process should consider a minimum pore diameter of 1 mm. The increased VOS precision error for the SL models compared with the CubaClinical commercial phantom may be attributed to the incorporation of oil rather than water for the marrow mimic, resulting in a temperature dependence, also observed in other resin/oil phantoms7. It is feasible that an improvement in velocity precision could be achieved if a temperature-dependent correction factor was incorporated, perhaps aided by a liquid crystal thermometer mounted on the surface of the phantom. 5. CONCLUSION A structural model for cancellous bone has been developed using stereolithography. The SL model has been assessed using the McCue CUBAclinical ultrasound bone densitometer. The BUA and velocity values obtained were commensurate with the commercial phantom. The intra- and intersample variability for the SL model was similar at 5% for BUA and 2.5% for velocity. The model has significant potential for assessing both the ultrasonic and mechanical dependence upon cancellous bone structure. For example, the initial CAD image may be modified to exhibit both perforation and thinning observed in type I and type II osteoporotic changes. The authors know of no other technique which may
604
Table 2 BUA and VOS mean and precision data for the McCue phantom and stereolithography models over a 108-day period Samples
CubaClinical Phantom SL1 SL2 SL3 SL4 SL5 SL6
BUA
VOS
Mean
CV%
Mean
CV%
66.64
4.55
1718
0.50
67.00 59.00 67.94 53.03 65.74 67.29
4.15 5.95 5.76 5.77 4.94 3.37
1855 1853 1840 1690 1843 1864
2.46 2.41 2.54 2.44 2.54 2.58
enable the structural complexity of a porous sample to be defined. ACKNOWLEDGEMENTS The authors acknowledge the EC BIOMED 1 Concerted Action Assessment of Bone Quality in Osteoporosis for their support and Professor J. D. Currey, University of York, for assistance with the mechanical analysis of the solid stereolithography sample. REFERENCES 1. Gluer, C. C., Vahlensieck, M., Faulkner, K. G., Engelke, K., Black, D. and Genant, H. K., Site-matched calcaneal measurements of broad-band ultrasound attentuation and single x-ray absorptiometry: do they measure different skeleton properties?. J. Bone Miner. Res., 1992, 7, 1071– 1079. 2. Waud, C. E., Lew, R. and Baran, D. T., The relationship between ultrasound and densitometric measurements of bone mass at the calcaneus in women. Calcif. Tissue Int., 1992, 51, 415–418. 3. Langton, C. M., Riggs, C. M. and Evans, G. P., Pathway of ultrasound waves in the equine third metacarpal bone. J. Biomed. Eng., 1990, 13, 113–118. 4. Gluer, C. C., Wu, C. Y. and Genant, H. K., Broadband ultrasound attenuation signals depend on trabecular orientation: an in vitro study. Osteo. Int., 1993, 3, 185–191. 5. McKelvie, M. L., Fordham, J., Clifford, C. and Palmer, S. B., In vitro comparison of quantitative computed tomography and broadband ultrasonic attenuation of trabecular bone. Bone, 1989, 10, 101–104. 6. Langton, C. M., Njeh, F., Hodgskinson, R. and Currey, J. D., Prediction of mechanical properties of the human calcaneus by broadband ultrasonic attenuation. Bone, 1996, 18, 495–503. 7. Clarke, A. J., Evans, J. A., Truscott, J. G., Milner, R. and Smith, M. A., A phantom for quantitative ultrasound of trabecular bone. Phys. Med. Biol., 1994, 39, 1677–1687. 8. Njeh C. F. The dependence of ultrasound, velocity and attenuation on material properties of cancellous bone. Ph.D. Thesis. Sheffield Hallam University, 1995. 9. Langton, C. M., ZSD: a universal parameter for precision in the ultrasonic assessment of osteoporosis. Physiol. Meas., 1997, 18, 67–72.