The processing pipeline used to produce a print-ready mesh is outlined in Fig. 1 (For a guide for creating 3D visualizations of MRI data, see Madan 2015)

Figure 1.

Illustration of the processing pipeline involved in converting a structural MRI volume to a print-ready surface mesh.

FreeSurfer (http://freesurfer.net/fswiki) (Dale et al. 1999, Fischl 2012, Fischl et al. 2002) is first used to construct a cortical surface mesh from a standard T1-weighted structural volume (e.g., MPRAGE or SPGR). The standard recon-all procedure was used, though this additionally provides many other metrics that are not needed here (e.g., cortical thickness, cortical parcellations). This process takes many hours to complete.

After the FreeSurfer pipeline has been run, the remaining steps can be executed within a few minutes, provided the necessary programs are already installed. Nearly all of these remaining steps can be executed via a command-line interface, with the commands listed in Table 1. These steps were developed and tested for Mac OS X, but should readily be adapted for use on Linux operating systems.

Table 1. Download as CSV 

Code

1 2 3 4 5 6 7 8 9 10 11 12 13

mg=/Applications/Research/imaging/meshgeometry/meshgeometry_mac ml=/Applications/meshlab.app/Contents/MacOS/meshlabserver td=/Applications/Research/imaging/vcglib/apps/tridecimator/tridecimator   f=rh.pial $mg -i $f -centre -o $f.ply $td $f.ply $f.dec.ply 100000 $ml -i $f.dec.ply -o $f.dec.stl   f=lh.pial $mg -i $f -centre -o $f.ply $td $f.ply $f.dec.ply 100000 $ml -i $f.dec.ply -o $f.dec.stl

First, the cortical surface files (lh.pial and rh.pial) are converted to a generic mesh format (.ply) using meshgeometry (https://github.com/r03ert0/meshgeometry). As an example, resulting ply files are provided as supplemental here (Suppl. material 1), based on the 'bert' subject from FreeSurfer (https://surfer.nmr.mgh.harvard.edu/fswiki/TestingFreeSurfer). Second, tridecimator – a program distributed as part of VCGlib (http://vcg.isti.cnr.it/vcglib/install.html; see *1 for details on compiling tridecimator) – is used to decrease the complexity of the mesh, i.e., polygon reduction via decimation. Prior to this step, the mesh for each hemisphere will be made of 250,000-300,000 polygons; this decimation step will reduce the number of polygons to 100,000 (but the code can be midified to another number). This reduction in polygons is necessary to make the model easier to print, with minimal effect on the shape of the mesh. Third, MeshLab (http://meshlab.sourceforge.net) is used to convert the mesh again (.ply to .stl).

For 3D printing, meshes usually need to be 'repaired' to ensure that there are no holes or other issues with the mesh structure (cf. https://library.ualberta.ca/services/3dprinting/preparing-3d-model). This repair process can be readily accomplished using the online version of Netfabb (https://netfabb.azurewebsites.net), though a local version can instead be installed (https://www.netfabb.com/products/netfabb-basic).

(Also see http://www.shapeways.com/tutorials/how_to_use_meshlab_and_netfabb.)

The resulting files (.stl) are now print-ready. Fig. 2 shows an example of the resulting printed models. The hemisphere surfaces were split by the printers (University of Alberta Libraries) using Netfabb. Each 'quarter' of a brain is approximately the size of a golf ball and took 4 hours to print; each piece weighs approximately 30 grams, printed using a plastic filament (PLA/PHA blend) from ColorFabb (http://colorfabb.com/pla-pha).

Figure 2.

Photos of a resulting 3D-printed surface. (A) View of the full models. (B) Close-up of the coronal cross-section. (C) Close-up of the lateral surface. (D) Photo of the Machina Mk2 X20 printers used to print the models, located at the University of Alberta Libraries.