Purpose: Deformable lung phantoms have been proposed to investigate four-dimensional (4D) imaging and radiotherapy delivery techniques. However, most phantoms mimic only the lung and tumor without pulmonary airways. The purpose of this study was to develop a reproducible, deformable lung phantom with three-dimensional (3D)-printed airways.
Methods: The phantom consists of: (a) 3D-printed flexible airways, (b) flexible polyurethane foam infused with iodinated contrast agents, and (c) a motion platform. The airways were simulated using publicly available breath-hold computed tomography (CT) image datasets of a human lung through airway segmentation, computer-aided design modeling, and 3D printing with a rubber-like material. The lung was simulated by pouring liquid expanding foam into a mold with the 3D-printed airways attached. Iodinated contrast agents were infused into the lung phantom to emulate the density of the human lung. The lung/airways phantom was integrated into our previously developed motion platform, which allows for compression and decompression of the phantom in the superior-inferior direction. We quantified the reproducibility of the density (lung), motion/deformation (lung and airways), and position (airways) using breath-hold CT scans (with the phantom compressed and decompressed) repeated every two weeks over a 2-month period as well as 4D CT scans (with the phantom continuously compressed and decompressed) repeated twice over four weeks. The density reproducibility was quantified with a difference image (created by subtracting the rigidly registered baseline and the repeated images) in each of the compressed and decompressed states. Reproducibility of the motion/deformation was evaluated by comparing the baseline displacement vector fields (DVFs) derived from deformable image registration (DIR) between the compressed and decompressed phantom CT images with those of repeated scans and calculating the difference in the displacement vectors. Reproducibility of the airway position was quantified based on DIR between the baseline and repeated images.
Results: For the breath-hold CT scans, the mean difference in lung density between baseline and week 8 was -1.3 (standard deviation 33.5) Hounsfield unit (HU) in the compressed state and 0.4 (36.8) HU in the decompressed state, while large local differences were observed around the high-contrast structures (caused by small misalignments). By visual inspection, the DVFs (between the compressed and decompressed states) at baseline and last time point (week 8 for the breath-hold CT scans) demonstrated a similar pattern. The mean lengths of displacement vector differences between baseline and week 8 were 0.5 (0.4) mm for the lung and 0.3 (0.2) mm for the airways. The mean airway displacements between baseline and week 8 were 0.6 (0.5) mm in the compressed state and 0.6 (0.4) mm in the decompressed state. We also observed similar results for the 4D CT scans (week 0 vs week 4) as well as for the breath-hold CT scans at other time points (week 0 vs weeks 2, 4, and 6).
Conclusions: We have developed a deformable lung phantom with 3D-printed airways based on a human lung CT image. Our findings indicate reproducible density, motion/deformation, and position. This phantom is based on widely available materials and technology, which represents advantages over other deformable phantoms.
Keywords: deformable image registration; deformable phantom; flexible airways; three-dimensional printing.
© 2019 American Association of Physicists in Medicine.