Abstract
In the area of 3D printing, more and more maxillofacial surgery departments are equipped with 3D printers to build their own anatomical models or surgical guides. Prior to be printable, the patients' DICOM imaging data has to be converted to a 3D virtual model, a 3D mesh. The file format most commonly used is the STL (Standard Tesselation Language) file format. Many programs exist that are able to convert DICOM data to STL files. Commercial software, such as Surgicase CMF© are FDA- and CE-approved whereas free programs, available online do not have the approval. However, the latter are often used anyway because of financial reasons. In this article, we investigate whether 6 of these software solutions are equivalent or not.
Thin slice CT imaging data of a patient's mandible (in DICOM file format) was converted to STL meshes with 6 different software solutions. One commercial program, Surgicase CMF©, was used to build the reference model. Then 5 free programs were used to create 5 models of the same mandible, specifying the same thresholding parameters: InVesalius 3.0, 3DimViewer 2.2.4, 3D Slicer, itk-Snap and Seg3D. All of these models were loaded in Netfabb Basic 6.4 to retrieve dimensional data, geometric information and the number of holes in each mesh. Finally, the models were then compared to the reference model using CloudCompare 2.6.2.
All models created with free software differed from the reference model in the 3 dimensions. Mean length difference was −0.74 mm [−2.06; −0.32] (SD: 0.74), mean width difference −0.45 mm [−0.76; −0.25] (SD: 0.19) and mean height difference was 0.41 mm [0.14; 0.62] (SD: 0.18). Although the height was increased in all models, both the length and width were systematically decreased, resulting in an average decrease of volume of −7.1 cm3 [−7.45; −6.77] (SD: 0.32). The number of triangles used to create each mesh ranged from 20944 to 368244, resulting in a variation of the file size from 1023 Ko to 80462 Ko (0.16 to 12.70 times the file size of the reference model). Two of the free programs created meshes with errors, such as the presence of holes (non-watertight meshes) that could be repaired with Netfabb.
Free programs able to convert volume imaging data to a printable virtual mesh do not provide equivalent results. Variations were noted in the three plane of space with a systematic difference between free programs and the commercial FDA-approved one. While the length and width were less than a millimeter different to the reference, the dimension that most varied was the length with a difference reaching −2.06 mm with itk-Snap. Geometric data also varied significantly, the number of triangles composing the meshes being much different than the reference, resulting in variable file sizes. This traduces the fact that algorithms used by the programs are not the same. In the era of 3D printing made directly accessible in surgical departments, great attention should be paid to the accuracy of the models created with free software.
