Additive manufacturing, also known as 3-D printing and formerly known as rapid prototyping, is being increasingly used to produce bone replicas consisting of polymer that then are used to enhance the management of orthopedic problems in companion animals.1,2 Historically, polymer replicas were produced with high-end, professional machines, and the cost of fabrication was high. Several of the original patents have lapsed,3 and a new generation of low-end (office-based) polymer 3-D printers is currently commercially available.4 However, little is known about the accuracy and repeatability of these low-end 3-D printers.
The purpose of the study reported here was to assess the accuracy and repeatability of commercially available 3-D printers. We hypothesized that 3-D printing would be highly repeatable and that replicas made with low-end printers would be less accurate than replicas made with high-end printers but that the replicas for the low-end printers would have acceptable dimensional errors (< 0.5 mm). To test these hypotheses, polymer replicas of small, medium, and large bones of small animals were fabricated on 2 low-end and 2 high-end printers, replicas were scanned with a laser, and dimensions for the replicas were compared with dimensions of the 3-D renderings of these bones.
Materials and Methods
Sample
The CT images of 3 bones were selected from the veterinary hospital database of North Carolina State University. The right femur of a 19-month-old male cat (small replica), left radius of a 13-month-old male chondrodystrophic mixed-breed dog (medium replica), and right tibia of an 18-month-old female Mastiff (large replica) were used. The cat femur was scanned at 120 kV and 102 to 129 mA by use of a 16.2 × 16.2-cm field of view, the dog radius was scanned at 120 kV and 113 to 124 mA by use of a 23.5 × 23.5-cm field of view, and the dog tibia was scanned at 120 kV and 118 to 131 mA by use of a 33.8 × 33.8-cm field of view. All images were reconstructed into 0.75-mm slices by use of a sharp (B60s) kernel.
CT renderings
Cross-sectional DICOM images were imported into a commercially available modeling software program.a Bone thresholds (226 to 1634 Hounsfield units) were used to prepare 3-D renderings. The 3-D renderings of the bones were imported into a 3-D reverse-engineering software program.b Bones were aligned and imported into an additive manufacturing software program.c
Length, width of the proximal aspect of a bone, and width of the distal aspect of a bone were measured in triplicate on consecutive days (Figure 1); data were recorded on a spreadsheet. Length of the femur was the distance between the greater trochanter and distal aspect of the lateral condyle, width of the proximal aspect of the femur was the distance between the medial aspect of the femoral head and lateral aspect of the greater trochanter, and width of the distal aspect of the femur was the distance between the medial aspect of the medial condyle and lateral aspect of the lateral condyle. Length of the radius was the distance between the proximomedial aspect of the radial head and distal aspect of the medial styloid process, width of the proximal aspect of the radius was the distance between the medial aspect of the radial tuberosity and medial aspect of the radial cortex opposite the tuberosity, and width of the distal aspect of the radius was the distance between the medial and lateral aspects of the physeal scar. Length of the tibia was the distance between the proximal aspect of the intercondylar eminence and distal aspect of the medial malleolus, width of the proximal aspect of the tibia was the distance from the cranial aspect of the tibial tuberosity to the caudal aspect of the medial tibial condyle, and width of the distal aspect of the tibia was the distance between the medial aspect of the medial malleolus and lateral aspect of the tibial shaft proximal to the articular surface.
Photographs of CT-derived 3-D renderings used as the basis for fabrication of 3-D printed replicas of a cat femur (cranial [left] and lateral [right] views; A), dog radius (cranial [left] and lateral [right] views; B), and dog tibia (cranial [left] and medial [right] views; C). Length (black arrowheads) and width of the proximal (gray arrowheads) and distal (red arrowheads) aspects of the bones were measured by use of the measurement feature of an additive manufacturing software program.c Length of the femur was the distance between the greater trochanter and distal aspect of the lateral condyle, width of the proximal aspect of the femur was the distance between the medial aspect of the femoral head and lateral aspect of the greater trochanter, and width of the distal aspect of the femur was the distance between the medial aspect of the medial condyle and lateral aspect of the lateral condyle. Length of the radius was the distance between the proximomedial aspect of the radial head and distal aspect of the medial styloid process, width of the proximal aspect of the radius was the distance between the medial aspect of the radial tuberosity and medial aspect of the radial cortex opposite the tuberosity, and width of the distal aspect of the radius was the distance between the medial and lateral aspects of the physeal scar. Length of the tibia was the distance between the proximal aspect of the intercondylar eminence and distal aspect of the medial malleolus, width of the proximal aspect of the tibia was the distance from the cranial aspect of the tibial tuberosity to the caudal aspect of the medial tibial condyle, and width of the distal aspect of the tibia was the distance between the medial aspect of the medial malleolus and lateral aspect of the tibial shaft proximal to the articular surface. Magnification of bones differs among panels. Bar = 20 mm.
Citation: American Journal of Veterinary Research 78, 8; 10.2460/ajvr.78.8.900
Photographs of CT-derived 3-D renderings used as the basis for fabrication of 3-D printed replicas of a cat femur (cranial [left] and lateral [right] views; A), dog radius (cranial [left] and lateral [right] views; B), and dog tibia (cranial [left] and medial [right] views; C). Length (black arrowheads) and width of the proximal (gray arrowheads) and distal (red arrowheads) aspects of the bones were measured by use of the measurement feature of an additive manufacturing software program.c Length of the femur was the distance between the greater trochanter and distal aspect of the lateral condyle, width of the proximal aspect of the femur was the distance between the medial aspect of the femoral head and lateral aspect of the greater trochanter, and width of the distal aspect of the femur was the distance between the medial aspect of the medial condyle and lateral aspect of the lateral condyle. Length of the radius was the distance between the proximomedial aspect of the radial head and distal aspect of the medial styloid process, width of the proximal aspect of the radius was the distance between the medial aspect of the radial tuberosity and medial aspect of the radial cortex opposite the tuberosity, and width of the distal aspect of the radius was the distance between the medial and lateral aspects of the physeal scar. Length of the tibia was the distance between the proximal aspect of the intercondylar eminence and distal aspect of the medial malleolus, width of the proximal aspect of the tibia was the distance from the cranial aspect of the tibial tuberosity to the caudal aspect of the medial tibial condyle, and width of the distal aspect of the tibia was the distance between the medial aspect of the medial malleolus and lateral aspect of the tibial shaft proximal to the articular surface. Magnification of bones differs among panels. Bar = 20 mm.
Citation: American Journal of Veterinary Research 78, 8; 10.2460/ajvr.78.8.900
Photographs of CT-derived 3-D renderings used as the basis for fabrication of 3-D printed replicas of a cat femur (cranial [left] and lateral [right] views; A), dog radius (cranial [left] and lateral [right] views; B), and dog tibia (cranial [left] and medial [right] views; C). Length (black arrowheads) and width of the proximal (gray arrowheads) and distal (red arrowheads) aspects of the bones were measured by use of the measurement feature of an additive manufacturing software program.c Length of the femur was the distance between the greater trochanter and distal aspect of the lateral condyle, width of the proximal aspect of the femur was the distance between the medial aspect of the femoral head and lateral aspect of the greater trochanter, and width of the distal aspect of the femur was the distance between the medial aspect of the medial condyle and lateral aspect of the lateral condyle. Length of the radius was the distance between the proximomedial aspect of the radial head and distal aspect of the medial styloid process, width of the proximal aspect of the radius was the distance between the medial aspect of the radial tuberosity and medial aspect of the radial cortex opposite the tuberosity, and width of the distal aspect of the radius was the distance between the medial and lateral aspects of the physeal scar. Length of the tibia was the distance between the proximal aspect of the intercondylar eminence and distal aspect of the medial malleolus, width of the proximal aspect of the tibia was the distance from the cranial aspect of the tibial tuberosity to the caudal aspect of the medial tibial condyle, and width of the distal aspect of the tibia was the distance between the medial aspect of the medial malleolus and lateral aspect of the tibial shaft proximal to the articular surface. Magnification of bones differs among panels. Bar = 20 mm.
Citation: American Journal of Veterinary Research 78, 8; 10.2460/ajvr.78.8.900
Polymer replicas
Polymer replicas were fabricated by use of 4 systems: a high-end system that relied on jetting of curable liquid photopolymer,d a high-end system that relied on polymer extrusion,e a triple-nozzle polymer extrusion low-end system that printed ABS and PLA,f and a dual-nozzle polymer extrusion low-end system that printed ABS and a high-impact polystyrene support.g Replicas produced by liquid photopolymer jetting consisted of an epoxy-based photopolymerh with 16-μm-thick layers. The support material consisted of acrylici and was removed with a water jet.j Replicas fabricated by use of the high-end polymer extrusion system were built in a heated chamber (70°C) with 254-μm-thick layers of ABS P400 material. Diameter of nozzles was 300 μm. Support material consisted of polyethersulfonek and was removed by dissolution via immersion in a 2.18% sodium hydroxide solution for 3 hours. Replicas produced by use of the triple-nozzle (nozzle diameter, 400 μm) low-end polymer system consisted of 200-μm-thick layers of PLA on ABS support material; the support material was removed manually. The dual-nozzle (nozzle diameter, 500 μm) polymer extrusion low-end system was placed in a custom-made acrylic enclosure heated to 50°C. Building parameters were controlled by use of a software program.l Replicas produced by use of the dual-nozzle polymer extrusion low-end system were printed with 200-μm-thick layers in ABS with high-impact polystyrene support; the support material was removed by dissolution via immersion in limonene for 24 hours.
Geometric assessment
Bone replicas were scanned by use of a reverse-engineering software programm and a portable laser-based coordinate measurement machinen with repeatability of ± 24 μm and accuracy (volumetric maximum deviation) of ± 34 μm. Polystyrene spheres (10 mm in diameter) were affixed to the replicas for automatic registration (alignment). When automatic registration failed, registration was performed manually. For each bone, several overlapping partial scans of replicas and registration spheres were acquired. These partial scans were combined into a single scan, and the registration spheres were deleted from the scan. Scans were saved as stereolithography files and imported into the additive manufacturing software program.c Dimensions were measured by use of the same method used for CT images.
Statistical analysis
The ICC values for repeated measurements of length, width of the proximal aspect, and width of the distal aspect of 3-D renderings were calculated to evaluate measurement repeatability. The ICC values for length, width of the proximal aspect, and width of the distal aspect measurements of triplicate replicas were calculated to evaluate fabrication repeatability.3 Repeatability was considered high at ICC > 0.9.
To evaluate fabrication accuracy, measurements from replicas fabricated by use of the four 3-D printing methods were compared with measurements of CT-derived 3-D renderings by use of an ANOVA. Least squares means were calculated for each printer. Bland-Altman plots were created. Bias was assessed by comparing mean differences for printers with CT-derived measurements by use of contrasts. Bias proportionality was evaluated by correlating dimensional errors during fabrication and length of measurements. Replicas that differed from CT renderings by > 0.5 mm were considered inaccurate. The accuracy threshold of 0.5 mm was determined on the basis of the 95% CI of measurements of replicas fabricated by use of high-end 3-D printers in a previous study2 Analyses were performed with statistical analysis software.o Values of P < 0.05 were considered significant.
Results
Fabrication of 1 large long-bone replica failed once on each of the 2 low-end printers. No additional attempts were made to fabricate these replicas because of time constraints. Therefore, there were only duplicate replicas (rather than triplicate replicas) for that bone. We collected 129 measurements (27 measurements of CT renderings and 102 measurements for the 34 replicas; Table 1).
Mean ± SD dimensions for CT-derived renderings and scans of 34 long-bone replicas fabricated by use of four 3-D printing methods.
| Width | |||||
|---|---|---|---|---|---|
| Bone | Method | Length (mm)* | P value† | Proximal aspect (mm) | Distal aspect (mm) |
| Cat femur | CT scan | 116.71 ± 0.50 | — | 22.36 ± 0.04 | 19.66 ± 0.29 |
| Low-end dual-nozzle | 117.14 ± 0.09 | 0.272 | 22.32 ± 0.05 | 19.46 ± 0.14 | |
| Low-end triple-nozzle | 117.18 ± 0.15 | 0.216 | 21.84 ± 0.26 | 19.90 ± 0.39 | |
| High-end extrusion | 117.03 ± 0.10 | 0.537 | 21.70 ± 0.19 | 19.36 ± 0.18 | |
| High-end ink jetting | 117.32 ± 0.06 | 0.076 | 22.70 ± 0.28 | 20.12 ± 0.34 | |
| Dog radius | CT scan | 121.81 ± 0.24 | — | 17.32 ± 0.09 | 24.63 ± 0.27 |
| Low-end dual-nozzle | 122.78 ± 0.31 | < 0.001 | 17.45 ± 0.16 | 25.42 ± 0.28 | |
| Low-end triple-nozzle | 121.51 ± 0.09 | 0.256 | 16.91 ± 0.08 | 25.50 ± 0.12 | |
| High-end extrusion | 122.23 ± 0.06 | 0.144 | 16.69 ± 0.08 | 24.39 ± 0.03 | |
| High-end ink jetting | 122.14 ± 0.01 | 0.315 | 17.40 ± 0.16 | 24.45 ± 0.02 | |
| Dog tibia | CT scan | 255.08 ± 0.26 | — | 53.05 ± 0.25 | 32.15 ± 0.08 |
| Low-end dual-nozzle‡ | 258.02 ± 0.01 | 0.012 | 54.17 ± 0.01 | 31.05 ± 1.14 | |
| Low-end triple-nozzle‡ | 258.74 ± 0.03 | 0.003 | 54.44 ± 0.04 | 31.42 ± 0.01 | |
| High-end extrusion | 257.80 ± 0.81 | 0.010 | 54.17 ± 0.10 | 31.51 ± 0.58 | |
| High-end ink jetting | 258.91 ± 1.14 | 0.001 | 54.14 ± 0.06 | 31.28 ± 0.25 | |
Measurements were collected for 3 CT-derived renderings of bones and collected for each of the 3 replicas fabricated by use of each of the four 3-D printing methods.
Comparison of the length measurements of the scanned replicas with those of the CT-derived renderings; values were considered significant at P < 0.05.
Fabrication failed for one of the replicas; thus, this represents results for duplicate replicas rather than triplicate replicas.
— = Not applicable.
Measurements of CT renderings were highly repeatable (ICC, > 0.999; variance, 0.05 mm2; and 95% CI, ± 0.09 mm). Replica fabrication was highly repeatable. The ICC of each 3-D printing method was > 0.999. Variance for measurements of triplicate replicas was 0.06 mm2 (95% CI, 0.03 to 0.10 mm2) for the low-end dual-nozzle printer, 0.02 mm2 (95% CI, 0 to 0.04 mm2) for the low-end triple-nozzle printer, 0.08 mm2 (95% CI, 0.04 to 0.12 mm2) for the high-end extrusion printer, and 0.12 mm2 (95% CI, 0.07 to 0.17 mm2) for the high-end photopolymer jetting printer. Overall variance was 0.07 mm2 (95% CI, 0.02 to 0.12 mm2).
Mean dimensional error (root mean square error) for measurements of bone length was 1.02 mm for replicas fabricated by use of the low-end dual-nozzle printer, 1.26 mm for replicas fabricated by use of the low-end triple-nozzle printer, 1.10 mm for replicas fabricated by use of the high-end extrusion printer, and 1.43 mm for replicas fabricated by use of the high-end photopolymer jetting printer. Overall root mean square dimensional error was 1.22 mm. Overall coefficient of variation was 0.35. Dimensional errors increased as replica size increased (r = 0.79; P < 0.001). Length of small replicas did not differ significantly from the length of CT-derived renderings for all 3-D printing machines. For medium replicas, measurements of replicas fabricated by use of the dual-nozzle polymer extrusion 3-D printer were larger than measurements of CT-derived renderings. Measurements of medium replicas made by use of other 3-D printers did not differ from measurements of CT-derived renderings. For large replicas, measurements of replicas fabricated by use of all 3-D printers were larger than measurements of CT-derived renderings (Figure 2). Overall, 3-D fabricated replicas were larger than 3-D renderings, with a mean ± SD dimensional error of 0.40 ± 1.15 mm (95% CI, 0.18 to 0.63 mm) and limits of agreement of −1.86 mm (95% CI, −1.66 to −2.06 mm) and 2.66 mm (95% CI, 2.47 to 2.86 mm).
Bland-Altman plot of the difference between CT-derived 3-D renderings and 3-D printed replicas for a dual-nozzle polymer extrusion low-end system (white circles [n = 24 measurements]), triple-nozzle polymer extrusion low-end system (white triangles [24 measurements]), high-end system that relied on polymer extrusion (gray diamonds [27 measurements]), and high-end system that relied on photopolymer ink jetting (black circles [27 measurements]). The mean difference (black line) is 0.4 mm. The regression line for all 102 measurements (dotted line), limits of agreement (gray-shaded area), and prediction limits (dashed lines) are indicated.
Citation: American Journal of Veterinary Research 78, 8; 10.2460/ajvr.78.8.900
Bland-Altman plot of the difference between CT-derived 3-D renderings and 3-D printed replicas for a dual-nozzle polymer extrusion low-end system (white circles [n = 24 measurements]), triple-nozzle polymer extrusion low-end system (white triangles [24 measurements]), high-end system that relied on polymer extrusion (gray diamonds [27 measurements]), and high-end system that relied on photopolymer ink jetting (black circles [27 measurements]). The mean difference (black line) is 0.4 mm. The regression line for all 102 measurements (dotted line), limits of agreement (gray-shaded area), and prediction limits (dashed lines) are indicated.
Citation: American Journal of Veterinary Research 78, 8; 10.2460/ajvr.78.8.900
Bland-Altman plot of the difference between CT-derived 3-D renderings and 3-D printed replicas for a dual-nozzle polymer extrusion low-end system (white circles [n = 24 measurements]), triple-nozzle polymer extrusion low-end system (white triangles [24 measurements]), high-end system that relied on polymer extrusion (gray diamonds [27 measurements]), and high-end system that relied on photopolymer ink jetting (black circles [27 measurements]). The mean difference (black line) is 0.4 mm. The regression line for all 102 measurements (dotted line), limits of agreement (gray-shaded area), and prediction limits (dashed lines) are indicated.
Citation: American Journal of Veterinary Research 78, 8; 10.2460/ajvr.78.8.900
Discussion
For the study reported here, we accepted the hypothesis that 3-D printing by use of the low-end and high-end 3-D printers was highly repeatable. Differences among replicas fabricated from an identical 3-D rendering were within 50 μm. That dimensional tolerance represented < 1 fabrication layer for polymer extrusion printers, for which layer thickness ranged from 200 to 254 μm, and represented approximately 3 fabrication layers for the photopolymer jetting printer. The high-end extrusion polymer printer had the thinnest layers. The triple-nozzle low-end printer could print with a layer thickness of 70, 200, or 300 μm. Subjectively, replicas printed with a layer thickness of 70 μm had a poor surface finish, compared with that for replicas fabricated with a layer thickness of 200 μm, potentially because of a slow extrusion rate. Because printing settings (eg, feed rate [nozzle motion in the horizontal plane {x- and y-axes}] and extrusion rate) for the triple-nozzle printer were not adjustable, a decision was made to print with a layer thickness of 200 μm. For the dual-nozzle printer, layer thickness could be selected, and a layer thickness of 200 μm was chosen because, on the basis of the authors' experiences and results of preliminary tests, that was the thinnest layer that resulted in an optimal surface finish. For layer thicknesses < 200 μm, there was a low extrusion rate and nozzle pressure that caused a lack of back pressure in the nozzle, which led to inconsistent extrusion of material. The nozzle diameter differed among the 3 extrusion printers, and those nozzle diameters were fixed. Some extrusion printers have interchangeable nozzles with different diameters. Nozzle diameter influences layer thickness and likely would influence printer accuracy.
The hypothesis that 3-D printed replicas were accurate was rejected because the length of all replicas of the dog tibia exceeded the length of the 3-D rendering of that bone by > 0.5 mm. Mean error was approximately 1% of bone length. That lack of accuracy was not detected in small replicas, possibly because measurement repeatability (± 90 μm) in those replicas represented more than one-fifth of the mean dimensional error (approx 450 μm). It was also possible that 3-D printing was less accurate when fabricating large parts, compared with the accuracy for fabrication of small parts. Small parts are fabricated in the center of the build platform, but large parts are fabricated by use of the platform edges because they require full use of the platform. Printers could potentially have been less accurate at the platform edges because, for many printers, nozzle motion in the horizontal plane (x- and y-axes) is belt driven. By comparison, nozzle motion in a vertical direction (z-axis) is driven by a worm gear, which is a more robust mechanism.4 Warping of a replica could have been responsible for the lack of accuracy. Warping is more likely to occur when fabricating large parts because of increased thermal stresses. However, warping would likely lead to a decrease in part length, and the large replicas fabricated in the study reported here were excessively long. Slight warping could lead to contact of a part with the nozzle, and the resulting interference could flatten and deform the polymer being extruded, which would lead to dimensional inaccuracy.
General accuracy of light polymerization, powder bed fusion, and extrusion deposition 3-D printing has been reviewed.6 The authors of that review6 reported an accuracy of 200 μm for the photopolymer jetting 3-D printer and 0.15% for the high-end polymer extrusion 3-D printer that were used in the present study, and they indicated that accuracy of polymer extrusion 3-D printing may decrease when part size increases. Accuracy of 3-D printed replicas has been evaluated in several studies2,7–11 involving bones. A dimensional error of approximately 2% was reported for a study7 conducted to evaluate human mandible replicas printed by use of a paper-based 3-D printer. A study2 conducted to evaluate variability of replicas of a dog femur fabricated by use of three 3-D printing methods did not detect a discrepancy between replicas and the length of the cadaver bone that was replicated. However, the length of that femur was only 106 mm, which decreased the likelihood of detection of dimensional errors. In a study8 conducted to evaluate the influence of CT reconstruction algorithms and CT image processing (segmentation) on accuracy of fabrication of 5-cm cubes fabricated by use of a low-end polymer extrusion 3-D printer, the printed parts overestimated cube size by approximately 1%. In 1 study9 in which investigators assessed the accuracy of fabrication of a portion of a mandible of approximately 50 mm by use of a high-end polymer extrusion 3-D printer, mean dimensional error was 0.35 mm. Assessment of the accuracy and reproducibility of segmentation and fabrication by use of a high-end polymer extrusion 3-D printer for human hip and shoulder joints revealed a root mean square error for the comparison of cadaver bone and replica of 0.3 ± 0.5 mm for the hip joint and 0.3 ± 0.4 mm for the shoulder joint.11 Overall, accuracy of image segmentation (mean absolute deviation ± root mean square error of 0.3 ± 0.4 mm) was less than the accuracy of 3-D printing (0.1 ± 0.1 mm). The lack of accuracy of image segmentation is more pronounced in the periphery of articular surfaces. An evaluation of replicas of 3 human phalanges revealed similar conclusions: replicas overestimated actual size, accuracy was most impacted by resolution of the 3-D rendering (triangulation resolution) and by segmentation algorithms (thresholding), and errors were most pronounced in areas with high curvature.12 In another study13 conducted to evaluate segmentation of CT images of cancellous bone, the threshold also had a major impact on the volume of bone renderings because a change in threshold of 0.5% led to a change in bone volume of 0.5% for low-density cancellous bone.
The hypothesis that 3-D printed replicas fabricated by use of high-end printers were more accurate than 3-D printed replicas fabricated by use of low-end printers was rejected. All printers were accurate when fabricating small replicas and inaccurate when fabricating large replicas. The cost of purchase and maintenance for high-end printers may be 100 times as much as that for low-end printers, and low-end printers were acceptable for the purpose of fabricating CT-derived bone replicas. However, low-end 3-D printers have limitations. The volume of the build platform is limited, compared with that of high-end printers, which makes fabrication of large bones more technically challenging or impossible. Although it is possible to modify the orientation of a part within the build platform to allow fabrication of parts longer than the width or depth of the build volume, more support material is required for an oblique orientation, and the duration of fabrication is increased, which increases the likelihood of failure during fabrication. Fabrication of large replicas by use of the low-end printers failed twice. The replicas warped as a result of thermal stresses. For the dual-nozzle printer, strategies used to prevent warping included gluing the support material to the build platform and heating the build chamber to 50°C. These strategies limited, but did not fully prevent, warping. For the triple-nozzle printer, the use of PLA required disabling the air heating unit within the build chamber.
Low-end printers offer limited options for materials used for fabrication. Material options for the low-end printers used in the study reported here were limited to ABS and PLA. Replicas produced by use of the triple-nozzle polymer extrusion low-end system were initially fabricated with ABS on PLA support so that the same material as for the dual-nozzle and triple-nozzle printers could be used and to facilitate removal of the support material because PLA is weaker than ABS. However, the processing temperature for the ABS used in the triple-nozzle 3-D printer (250°C) was higher than the processing temperature for the PLA (210°C). The deposition of ABS on PLA support remelted the PLA surface, which altered the shape and texture of the replica. Therefore, replicas produced by use of the triple-nozzle low-end printer were fabricated with PLA on ABS support. In the dual-nozzle low-end printer, the flowing temperature of ABS for fabrication of replicas (230°C) was identical to that of the high-impact polystyrene support material, and deformation of the support material was not seen. Similarly, for the high-end extrusion printer, flowing temperature of ABS (280°C) was identical to that of the polyethersulfone support material, and deformation of the support material was not seen. The ABS used in the triple-nozzle and low-end printers was from different sources because the triple-nozzle printer required the use of the manufacturer's material with an ABS filament diameter of 1.75 mm, but the dual-nozzle printer could only print with a filament diameter of 3.0 mm. The ABS melting temperature differed for the triple-nozzle and dual-nozzle printers because the triple-nozzle printer had a fixed ABS extrusion temperature of 250°C, whereas the dual-nozzle printer had an adjustable ABS extrusion temperature that was set at 230°C because ABS extruded more evenly in the dual-nozzle printer at 230°C than at higher or lower temperatures. The ABS extrusion temperatures differ among printers on the basis of nozzle diameter, use of a heated enclosure, and specific sources of ABS. The range of materials available for low-end 3-D printing is expanding and currently includes nylon, carbon fiber–reinforced nylon, polyetheretherketone, and other materials.
Consequences of the 3-D printing dimensional errors for the present study differ on the basis of the intended use of the replicas. Small dimensional errors are unlikely to be detrimental for replicas used for educational or diagnostic purposes. For surgical planning, including preparation of an external fixation frame or plate contouring, a 3-mm overestimation of length is unlikely to have clinical consequences. When preparing replicas intended for research purposes, including testing of bone surface strains or evaluating the fit of articulating surfaces, these dimensional errors would have a larger impact.
For the study reported here, we concluded that low-end 3-D printers could fabricate CT-derived bone replicas with high repeatability. The 3-D printed replicas were longer (approx 1%) than the corresponding CT-derived 3-D renderings.
ABBREVIATIONS
| ABS | Acrylonitrile butadiene styrene |
| CI | Confidence interval |
| ICC | Intraclass correlation coefficient |
| PLA | Polylactic acid |
Footnotes
Mimics, version 16.0 (x64), Materialise, Plymouth, Mich.
Geomagic Studio 2013, 3D Systems, Rock Hill, SC.
3-maticSTL, version 8.0 (x64), Materialise, Plymouth, Mich.
Objet 350 Connex1, Stratasys, Minneapolis, Minn.
FDM Dimension SST, Stratasys, Minneapolis, Minn.
CubePro Trio, 3D Systems, Rock Hill, SC.
Bits from bytes BFB-3000, 3D Systems, Rock Hill, SC.
VeroWhitePlus RGD835, Stratasys, Minneapolis, Minn.
Support SUP705, Stratasys, Minneapolis, Minn.
POWERBLAST 20–120 bar, Objet Geometries, Stratasys, Minneapolis, Minn.
PPSF support material, Stratasys, Minneapolis, Minn.
KISSlicer, version 1.4.5.6. Available at: kisslicer.com/. Accessed Nov 21, 2016.
Geomagic Studio 64bit, version 2013.0.2, 3D Systems, Rock Hill, SC.
FARO Edge, ScanArm, Lake Mary, Fla.
SAS, version 9.4, SAS Institute Inc, Cary, NC.
References
1. Hespel AM, Wilhite R, Hudson J. Invited review. Applications for 3D printers in veterinary medicine. Vet Radiol Ultrasound 2014; 55:347–358.
2. Fitzwater KL, Marcellin-Little DJ, Harrysson OL, et al. Evaluation of the effect of computed tomography scan protocols and freeform fabrication methods on bone biomodel accuracy. Am J Vet Res 2011; 72:1178–1185.
3. Hornick J, Roland D. Many 3D printing patents are expiring soon: here's a round up & overview of them. Available at: 3dprintingindustry.com/news/many-3d-printing-patents-expiring-soon-heres-round-overview-21708/. Accessed Nov 21, 2016.
4. Schwartz A, Money K, Spangehl M, et al. Office-based rapid prototyping in orthopedic surgery: a novel planning technique and review of the literature. Am J Orthop 2015; 44:19–25.
5. Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychol Bull 1979; 86:420–428.
6. Lee PH, Chung H, Lee SW, et al. Review: dimensional accuracy in additive manufacturing processes, in Proceedings. Am Soc Mech Eng Int Manuf Sci Eng Conf 2014;V001T04A045-1–V001T04A045-7.
7. Olszewski R, Szymor P, Kozakiewicz M. Accuracy of three-dimensional, paper-based models generated using a low-cost, three-dimensional printer. J Craniomaxillofac Surg 2014; 42:1847–1852.
8. Ogden KM, Aslan C, Ordway N, et al. Factors affecting dimensional accuracy of 3-D printed anatomical structures derived from CT data. J Digit Imaging 2015; 28:654–663.
9. Ciocca L, De Crescenzio F, Fantini M, et al. CAD/CAM and rapid prototyped scaffold construction for bone regenerative medicine and surgical transfer of virtual planning: a pilot study. Comput Med Imaging Graph 2009; 33:58–62.
10. Hazeveld A, Huddleston Slater JJ, Ren Y. Accuracy and reproducibility of dental replica models reconstructed by different rapid prototyping techniques. Am J Orthod Dentofacial Orthop 2014; 145:108–115.
11. Smith EJ, Anstey JA, Venne G, et al. Using additive manufacturing in accuracy evaluation of reconstructions from computed tomography. Proc Inst Mech Eng H 2013; 227:551–559.
12. Pinto JM, Arrieta C, Andia ME, et al. Sensitivity analysis of geometric errors in additive manufacturing medical models. Med Eng Phys 2015; 37:328–334.
13. Hara T, Tanck E, Homminga J, et al. The influence of micro-computed tomography threshold variations on the assessment of structural and mechanical trabecular bone properties. Bone 2002; 31:107–109.
