Platform development of easily customizable brain phantoms using 3D-printing workflow

Introduction: A next-generation ultra-high performance human brain PET scanner, the NeuroEXPLORER (NX), is targeted to have <2.0mm resolution. Phantoms are key tools for testing and validation of new systems. In brain imaging, simple structured phantoms (such as point/line sources) may be insufficient, and realistic brain phantoms would better reflect scanner capabilities (Gravel, et al. MIC2021). Phantoms need fine features that simulate brain structures such as gyri and sulci. However, commercially available phantoms are expensive and cannot be easily or quickly customized. Also developing new full-scale brain phantom models can be costly with conventional methods. Therefore, we present a new low-cost, scalable, and easy-to-operate workflow for developing custom 3D printed phantoms designed to be filled by a pump system.

Methods: For testing purposes, three prototypes were designed; 1)Micro prototype v0.0, 2)Mini prototype v0.1, 3)Brain prototype v1.0. The Micro phantom is cylindrical (20mm diameter) with two channels per quadrant of different thicknesses (0.5, 1, 2, and 4mm). The Mini phantom is a larger version of the Micro with the same configuration of inlets, outlets, and channels (110mm diameter, Fig.1A). The Brain phantom was designed using the cortical gray matter surface data of BigBrain atlas (Fig.1B) (Wagstyl, et al. PLOS Bio.2020). For this prototype (v1.0), a cropped area of the brain was chosen that includes the finer structures of the occipital lobe (~2.5mm in the average of visual cortex). A continuous flow system allows for quick filling, thorough mixing of the desired solution, increased ease in eliminating bubbles, cleaning, and drying, which is useful for complex structures. To quickly connect and disconnect the phantom, commercially available plastic quick-disconnect tube couplings with plastic pipe fittings are secured into the 3D printed phantom with glue (Fig.1C). The phantoms were printed by sterolithography printers (Form 3L, clear resin). The Micro phantom was printed first to ensure correct part fit and sealing. Next, the Mini phantom was printed, and the inner structure was scanned using the CT of the Siemens mCT. The PET phantom study was conducted on the Mini phantom filled with ¹⁸F-FDG solution (98kBq/mL) in the Focus 220 (acquisition:10min, reconstruction:3D MAP-OSEM, voxel:0.95×0.95×0.8mm).

Results: The Micro phantom was estimated to take between 4-12 hours to print with approximate material cost of $4. The Mini phantom requires between 23-40 hours per print with approximate material cost of $75 (print time estimates vary depending on the orientation and chosen layer height). Both the Micro and Mini phantoms were printed successfully and filled within minutes using the continuous flow system (Fig.1C). During the preparation, there were no major issues such as leaking or contamination (Fig.1D). The CT scan from the Siemens mCT resolved all internal structures with channel widths down to 0.5 mm (Fig.1E). The PET images (average of 10 slices, Fig.1F) from the Mini phantom delivered the expected results. Only a few microbubbles were observed by the CT image on the filled Mini phantom.

Conclusions: Phantoms have historically been expensive and limited in their ability to test the full capabilities of imaging technologies. Laser-based 3D printing technologies such as SLA are capable of high-resolution models of complex geometries, with lower costs and time input. The Micro and Mini phantoms were designed for quick in-house engineering validation of 3D printing and scanning workflows. They are structurally similar to older generation phantoms for comparison purposes. We successfully showed imaging consistency between the CAD model, CT scan, and PET scan. Moreover, the entire process from print, to assembly, to scan can be completed within a week for the Mini. This is an extremely flexible, quick, and low-cost workflow for developing and scanning simple phantoms, and more complex geometry of the brain will be developed for our next iteration.

• Download figure
• Open in new tab
• Download powerpoint