3D printed patient-specific thorax phantom with realistic heterogenous bone radiopacity using filament printer technology

Abstract

Current medical imaging phantoms are usually limited by simplified geometry and radiographic skeletal homogeneity, which confines their usage for image quality assessment. In order to fabricate realistic imaging phantoms, replication of the entire tissue morphology and the associated CT numbers, defined as Hounsfield Unit (HU) is required. 3D printing is a promising technology for the production of medical imaging phantoms with accurate anatomical replication. So far, the majority of the imaging phantoms using 3D printing technologies tried to mimic the average HU of soft tissue human organs. One important aspect of the anthropomorphic imaging phantoms is also the replication of realistic radiodensities for bone tissues. In this study, we used filament printing technology to develop a CT-derived 3D printed thorax phantom with realistic bone-equivalent radiodensity using only one single commercially available filament. The generated thorax phantom geometry closely resembles a patient and includes direct manufacturing of bone structures while creating life-like heterogeneity within bone tissues. A HU analysis as well as a physical dimensional comparison were performed in order to evaluate the density and geometry agreement between the proposed phantom and the corresponding CT data. With the achieved density range (-482 to 968 HU) we could successfully mimic the realistic radiodensity of the bone marrow as well as the cortical bone for the ribs, vertebral body and dorsal vertebral column in the thorax skeleton. In addition, considering the large radiodensity range achieved a full thorax imaging phantom mimicking also soft tissues can become feasible. The physical dimensional comparison using both Extrema Analysis and Collision Detection methods confirmed a mean surface overlap of 90% and a mean volumetric overlap of 84,56% between the patient and phantom model. Furthermore, the reproducibility analyses revealed a good geometry and radiodensity duplicability in 24 printed cylinder replicas. Thus, according to our results, the proposed additively manufactured anthropomorphic thorax phantom has the potential to be efficiently used for validation of imaging- and radiation-based procedures in precision medicine.

Keywords: 3D printing; Filament printer technology; Heterogenous bone radiopacity; Imaging phantom; Patient-specific thorax phantom.

Figure 1

Representation of different heterogeneity of bone density inside one axial slice of the patient CT. The thicknesses of the cortical bone were approximately 1 mm, 2 mm and 2.5 mm for the vertebral body, ribs and dorsal vertebral column, respectively. Average density values were 20, and 107, 184 and -85 HU for bone marrow areas, corresponding to vertebral body, ribs, dorsal vertebral column and surrounding soft tissue respectively.

Figure 2

Representation of the designed phantom in 3D Slicer software, a) Anterior view, b) Caudal view. Vertebral body, dorsal vertebral column, ribs, soft tissue and flat bearing are shown in red, blue, yellow, brown and green respectively.

Figure 3

Illustration of the design of the cortical bone in the PrusaSlicer, a) Cranio-caudal view. Magnification view of b) vertebral body and dorsal vertebral column c) ribs.

Figure 4

Workflow of physical dimensional analysis using the vertebral body as an example. Mesh refinement using Materialise-3-matic software.

Figure 5

The three printed cylinders, S0-S6 (a, Left), S7-S12 (a, Middle), S13-S18 (a, Right), (b) the eight printed replicas (S0-S12) with 1 mm wall thickness, (c) eight replicas (S0-S12) with 2 mm wall thickness, and (d) the eight replicas (S0-S12) with 2.5 mm wall thickness.

Figure 6

The transverse and coronal views from the CT scan from samples S0-S15.

Figure 7

The transverse and coronal views from the CT scan from samples S13-S14 for non-isotropic pixel size with pixel sizes of 0.5 mm, 0.4 mm, 0.3 mm as well as isotropic pixel size equal to 0.6 mm.

Figure 8

The 3D printed thorax phantom, a) Caudal-cranial and b) Cranial-caudal view of thorax phantom in supine position, (a, b) and longitudinal position (c).

Figure 9

Representation of an axial slice of CT from the patient and the 3D printed phantom. Axial and coronal view from the CT scan from the printed cylinder replicas (Axial view only includes S12 (c, left) an S6 (d, right)). In (c) and (d) we have a contour thickness of 0.9 mm, 1.8 mm, 2.25 mm, each repeated for each 4 cylinders, from left, middle and right, respectively. The display window is set to the range [0,101–1154] HU.

Figure 10

HU spectrum related to the cortical bone segmentation of a) the patient CT (230-1170 HU), b) phantom CT (230-910 HU), c) patient CT after removing the values related to the 910-1170 HU from the segmentation. The pink areas in left side show the HU range from which the cortical bone in the right side is segmented.

Figure 11

Results from physical dimensional comparison between the CT-derived patient and phantom STL, a) The result of surface overlap in Extrema analysis (in this case vertebral body, selected axis direction x (left), y (middle), z (right)), b) Pictorial and graphical representation of the results of the volume match in Collision detection is shown represented graphically. Collided triangles marked in pink on the target object.