Projection-based stereolithography for direct 3D printing of heterogeneous ultrasound phantoms

Abstract

Modern ultrasound (US) imaging is increasing its clinical impact, particularly with the introduction of US-based quantitative imaging biomarkers. Continued development and validation of such novel imaging approaches requires imaging phantoms that recapitulate the underlying anatomy and pathology of interest. However, current US phantom designs are generally too simplistic to emulate the structure and variability of the human body. Therefore, there is a need to create a platform that is capable of generating well-characterized phantoms that can mimic the basic anatomical, functional, and mechanical properties of native tissues and pathologies. Using a 3D-printing technique based on stereolithography, we fabricated US phantoms using soft materials in a single fabrication session, without the need for material casting or back-filling. With this technique, we induced variable levels of stable US backscatter in our printed materials in anatomically relevant 3D patterns. Additionally, we controlled phantom stiffness from 7 to >120 kPa at the voxel level to generate isotropic and anisotropic phantoms for elasticity imaging. Lastly, we demonstrated the fabrication of channels with diameters as small as 60 micrometers and with complex geometry (e.g., tortuosity) capable of supporting blood-mimicking fluid flow. Collectively, these results show that projection-based stereolithography allows for customizable fabrication of complex US phantoms.

Fig 1. 3D printing of phantoms.

A CAD model was translated into (A) the desired phantom. (B) PEGDA material was extruded in layers of set thicknesses and (C) photocured with ultraviolet light (purple) from the bottom. (D) Inclusions were incorporated within the PEGDA material by spatially varying light exposure time. (E) Subsequent layers were then extruded and photocured using the same technique until (F) the physical phantom was fabricated.

Fig 2. Imaging phantom schematics.

(A) Layer thickness phantom (Phantoms 1–4). (B) Contrast phantoms with printed inclusions of different sizes and photocuring times (Phantoms 5–7). (C) Speed of sound phantom (Phantom 9). (D) Temporal stability phantom (Phantoms 10,11). (E) Elasticity phantom with three inclusions of increased photocuring time (Phantoms 12,13). (F) Anisotropic elasticity phantom with columns printed in a checkerboard pattern in the X-Z plane (Phantom 14). (G) Serpentine flow phantom (Phantom 16). (H) Tumor flow phantom with a vascular-mimicking branched channel and a tumor-mimicking inclusion in the center (Phantom 17). Note that ∅ denotes object diameter.

Fig 3. Investigation of acoustic backscatter generation using optical and US imaging.

(A) Fluorescence imaging slice of Phantom 3 printed with 200-μm layers showing photobleached areas of high light exposure (blue arrow) and regions of low light exposure (red arrow). (B) Schematic of assumed phantom density, where dark regions indicate higher local density (blue arrow) and light regions indicate lower density (red arrow). The light source is below the phantom in the Z-direction (purple arrow). (C) Zoomed-in subsection (purple box in [D]) from Phantom 1 with 200-μm layers and imaged using 50-MHz B-mode US. Density mismatches appear as hyperechoic, while continuous-density regions appear anechoic. (D) 50-MHz (left) and 12-MHz (right) B-mode US images of Phantom 1 with 200-μm layers. (E) B-mode US images of Phantom 5 with two rectangular inclusions (orange arrows) printed with 50-μm layers at 50 MHz (top) and 12 MHz (bottom).

Fig 4. Backscatter patterning and photocuring dependence.

C-scan US images of patterned phantoms (A) without silica particles (Phantom 4) and (B) with silica particles (Phantom 7; 0.1 mg/mL). (C) US backscatter signal and (D) CNR for each additional cure time over 6 weeks imaged at 30 MHz in Phantom 7. The data points for additional cure time in (C) and (D) each correspond to a unique printed target in the phantoms shown in (A) and (B).

Fig 5. Generation of phantoms with flow-supporting channels.

(A) B-mode US images (left) of Phantom 15 showing fluorescent microbeads injected into channels and zoomed-in regions (right; cyan dashed boxes) show channels as small as 60 μm remain open and support flow. (B) Zoomed-in views of two regions within the red box in (C). A physical imperfection (i.e., the hyperechoic point denoted by yellow arrow) within the fabricated channel caused disturbances in the symmetry of the parabolic flow profiles within this region (top); downstream from the imperfection, the flow returned to normal (bottom). Dashed black box indicates the location of the image cut-out shown in S4 Fig. (C) Doppler-derived flow velocity vectors through the serpentine channel (Phantom 16). (D) Zoomed-in view of the region in the green box in (C) showing a symmetric, parabolic velocity profile.

Fig 6. 3D reconstruction of flow around a hypoechoic tumor region.

(A) B-mode C-scan of Phantom 17 showing a hypoechoic tumor inclusion flanked by two vessels. (B) 3D B-mode US rendering of Phantom 17 and (C) this rendering fused with the magnitude of Doppler-estimated flow velocity through the channel network; to the right, cross-sectional flow profiles are provided for three representative locations, indicated by the dashed red lines.