Three fresh frozen human cadavers were obtained from the Department of Anatomy. The pelvis, hands and feet were dissected and freed from all soft tissue exposing the bony structures. The ligamentous structures on the bone were left intact. Nine anatomic specimens − 3 pelvis, 3 hands and 3 feet—were used (Fig. 1).
Titanium Kirschner (K-) wires were inserted to mark anatomical landmarks (Fig. 2).
Pelvic landmarks were defined as the left and right: (1) tubercle on the pubic bone, (2) anterior superior iliac spine (ASIS), (3) posterior superior iliac spine (PSIS), (4) sacroiliac (SI) joint and (5) distance between the SI-joint and pubic bone on the right side of the pelvic bone.
Hand landmarks were defined as: (1) the radial styloid process and distal radioulnar articulation, (2) base and head of the second metacarpal bone and (3) base and head of the fifth metacarpal bone. Figure 3 shows a hand of a cadaver with these marker points.
Foot landmarks were defined as: (1) the distal medial malleolus, (2) between base and head of the first metatarsal bone and (3) base and head of the fifth metatarsal bone. Figure 4 shows a foot with the marker points. The distances between all landmarks were subsequently measured by two independent observers using a Vernier caliper. The point of intersection was defined as the intersection between bone and K wire.
Process of creating 3D prints from CT data
In order to create a 3D print, a standard tessellation language (STL) file is needed. This is a specific file format used by 3D software to generate 3D prints. Converting CT scans in digital imaging and communication (DICOM) file format to STL occurs in three stages : image acquisition,  image post-processing  and 3D printing.
The nine specimens were scanned using a Siemens Somatom Definition AS 64-slice CT (Siemens Healthcare, Forchheim, Germany). Slice thickness of 0.6 mm and soft reconstruction filters were used for our protocol in order to generate high resolution images and minimalize soft tissue image noise.
DICOM data of all cadavers were saved in Picture Archiving and Communication System (PACS). The two independent reviewers used the hospital’s integrated Philips Intellispace Portal® software to measure the distance between the markers in two-dimensional views.
The image post processing was divided into three phases:
Phase 1: creating a volume-rendered model of the object
We used Philips Intellispace Portal software to volume render the DICOM data into 3D reconstructions and to ascertain measurements of the 3DCT landmarks by the two independent observers. Figure 5 shows a 3DCT of the pelvis and a hand with respectively the 5 and 3 anatomical landmarks.
Phase 2: cleaning of the model and creating an STL file from the volume-rendered model
The 3D reconstruction was digitally cleaned from all surrounding artifacts and remnants of the soft tissue in the Philips Intellispace Portal and then saved as an STL file. The landmarks in the STL file were measured by the two independent reviewers using Meshlab, an open-source program (Fig. 6).
Phase 3: importing the STL file in 3D print software and generating the print code
Our hospital uses both the Makerbot Replicator Z18 (Makerbot Industries, USA)—a high end consumer extrusion 3D printer with a large build volume and the Ultimaker 3 (Ultimaker B.V., the Netherlands) a desktop 3D printer with a dual extruder. These printers use Polylactic Acid (PLA), a thermoplastic polyester, to extrude the plastic on a build platform where it solidifies.
The print code (G-code) for the Makerbot was generated using Simplify 3D and the print code for the Ultimaker was generated using Cura. The following process settings were standardized: extruder temperature 215 °C, chamber temperature 24 °C, primary layer height 0.2 mm, infill 2% (the outer side of bone exists of cortical bone, therefore the model supports itself and less infill can be used), support infill 20%, maximum overhang without support 60%.
The 3D models of the cadavers were printed in a ratio of 1:1. A 3D-printed model of a hand and the cadaver hand can be seen in Fig. 7. The amount of material used, PLA and support, printing time and filament costs were also noted. The two independent observers measured all distances on all 3D-printed models.
All of the measurements described above were undertaken by two independent observers. In summary, they measured the distances between the anatomical landmarks on the human cadavers (cadaver), 2DCT (Port_2D), 3D reconstructions (Port_3D), Meshlab (Mesh_3D) and 3D-printed models on the Ultimaker and Makerbot (Print_UM, Print_MB).
After 1 month, both observers were asked to measure all distances again to measure the inter-observer and intra-observer agreement. The distances between the K-wires on the fresh human cadavers were only measured once, because the cadavers had to be disposed of after two days.
Descriptive statistics were calculated to provide an overview of the print process settings. Observer data were analyzed and expressed in terms of intra- and inter-observer agreement. We used Pearson’s correlation to calculate the correlation coefficient r and to analyse the relationship between the measurements of both observers.
The measurements between both 3D printers and cadavers were also expressed as a percentage of cadavers. A Paired Samples T test was used to compare the measurements between cadavers, 2DCT, 3DCT, Meshlab and both 3D printers. A p value of 0.05 was determined as significant. IBM SPSS Statistics 24 was used for the database (Statistical Package for the Social Sciences, Chicago, IL, USA).