StepBrain: A 3-Dimensionally Printed Multicompartmental Anthropomorphic Brain Phantom to Simulate PET Activity Distributions

Design

StepBrain was obtained from a digital phantom (voxels, 0.9375 × 0.9375 × 1 mm) derived from multiparametric segmentation (9,10) of the MR images of a 38-y-old male volunteer (http://lab.ibb.cnr.it/Phantomag_Desc.htm) (11). Cerebrospinal fluid and extracranial tissues were preliminarily merged with the background, leaving only the brain parenchyma voxels. The deep GM nuclei (pallidum, dentate nucleus, and thalamus) were associated with the GM compartment, whereas the substantia nigra and red nucleus were associated with the WM. The caudate nucleus and the putamen were included in a separate compartment defining the dorsal striatum. Left and right caudates and putamen were transformed into a single compartment adding cylindric connections (lumen, 5 mm).

Voxel islands of GM or WM, 3-dimensionally disconnected from the main corresponding tissue compartment, were assigned to the surrounding prevalent tissue.

For each compartment, 2 bilateral cylindric pipes for filling (lumen, 6 mm) were added to the model, reaching from the top of the apical surface of each compartment.

Subsequently, nearest-neighbor interpolation was performed to obtain voxels of 0.75 × 0.75 × 0.75 mm, to model submillimeter-thick walls, still compatible with the minimum wall thickness achievable by fused deposition modeling 3D printing after optimization of the printing parameters (12).

Single voxelized surfaces were extracted at the compartments’ interface by selecting voxels (from the outside toward the inside) sharing vertices, edges, and faces (13), thus ensuring the extraction of a nonperforated Standard Tesselation Language model for 3D printing (12). The Standard Tesselation Language was derived using the Marching Cubes high-resolution 3D surface construction algorithm (14). The Standard Tesselation Language shells had no problems in the triangular mesh but needed further refinements to complete the phantom access via the tubes and eliminate any noise shells (Fig. 1). Finally, threaded joints were added to each tube to allow hermetic closure of the compartments after filling.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1.

From left to right: axial, sagittal, and coronal views of 3D model, and photograph of StepBrain phantom.

Materialization

The phantom was materialized using the Stratasys F370 professional fused deposition modeling 3D printer. As printing material, acrylonitrile butadiene styrene was used, along with the specific material (QSR Support; Stratasys) for soluble supports (13). Advanced fused deposition modeling 3D printing settings were optimized to ensure that the printed walls had continuity, minimizing the presence of microholes and air gaps in the same walls (12). Printing the brain phantom took 7 d 3 h. After printing, support structures were removed by 3 d in an agitated bath (4.5% NaOH, at 70°).

To verify complete cleaning of internal supports, a high-resolution CT scan of the empty phantom was obtained, which was also used for measurement of compartment volumes (“Phantom Loading” section) and for definition of compartment maps for partial-volume effect (PVE) correction (“PET/CT acquisitions and processing” section).

To guarantee the impermeability of the outermost compartments, so that, once filled, the liquid and radiotracer would not cross from one compartment to another, or outward, the phantom was treated with a polyvinyl acetate solution that could penetrate the print weft by physically closing the micropores in the printed surfaces.

Phantom Loading

The typical GM/WM contrast of ¹⁸F-FDG PET studies was simulated. Accordingly, the reference ratios between GM, WM, and striatum were preliminarily calculated on the basis of data reported by Greve et al. (15). To this end, for each compartment the average SUV was age-corrected to the age (38 y) of the volunteer represented in the phantom, resulting in a GM/WM ratio of 2.80 and a striatum/WM ratio of 2.63.

Phantom compartments were preliminarily filled by immersion in water containing 0.05% surfactant and continuously manually rotated for approximately 20 min to eliminate air bubbles. Because the compartment volumes, calculated by manual segmentation of the high-resolution CT images, were 710.7, 408.3, and 14.6 mL for the GM, WM, and striatum, respectively, 51, 10.7, and 1.1 MBq were then injected into the 3 compartments, respectively, and allowed to freely diffuse for 30 min, during which the phantom was placed on a dedicated rotating system to favor a homogeneous diffusion of the radiotracer within each phantom compartment.

PET/CT Acquisitions and Processing

The PET/CT acquisition of StepBrain was performed at the Nuclear Medicine Unit of the University of Naples “Federico II” on an Ingenuity TF-128 PET/CT scanner (Philips) using the acquisition and reconstruction protocol routinely applied for clinical ¹⁸F-FDG studies in adults. The spatial resolution of the scanner is 4.8 mm in full width at half maximum at the center and 5.1 mm at 10 cm, resulting in 6.25 and 6.48 mm, respectively, for images reconstructed with the reconstruction filter routinely used for clinical studies (4 mm in full width at half maximum) (16).

The PET/CT images were processed using PVELab software (17) (release 2.2, https://github.com/swederik/pvelab), assuming a uniform resolution of 6.36 mm in full width at half maximum (mean value in 10-cm field of view). To this end, the automated anatomic labeling atlas (18) was applied to the high-resolution CT images previously manually segmented to provide the reference GM, WM, and striatal volumes, which in turn were coregistered to the CT scan of the PET/CT study, thus providing an automatically parcellated GM map registered with the PET images. All the different region-of-interest–based and voxel-based PVE correction methods implemented in PVELab were then applied to obtain PVE-corrected values for GM, WM, and the GM regions of interest.

For visual comparison, a PET/CT acquisition of the widely used Hoffman brain phantom (Data Spectrum Corporation) (1) was also obtained.