Anatomically realistic ultrasound phantoms using gel wax with 3D printed moulds

Phantoms are essential clinical training tools for ultrasound imaging, and they are becoming more prominent as the use of this modality for guiding percutaneous procedures increases. Conventional, off-the-shelf commercial phantoms represent a very limited range of tissue structures and they can have anatomical imperfections (Hunt et al 2013). Moreover, their high costs preclude routine use as training tools. Several studies have showed promising results from anatomically realistic Doppler flow phantoms (Poepping et al 2004, King et al 2011). However, obtaining complex patient-specific geometries has been challenging. With the advent of 3D printing, there is a wide range of new methods for creating inexpensive, custom ultrasound phantoms. For instance, these printers can create polymeric structures that represent highly hyperechoic hard tissue structures such as the spine, which can be embedded in a tissue-mimicking material (TMM) (West et al 2014). Additionally, they can create specialized moulds for TMMs with which to create anatomically realistic soft tissue structures such as wall-less blood vessels (Nikitichev et al 2016), the kidney (Hunt et al 2013), and the heart (Holmes et al 2008). One of the limiting factors for creating ultrasound imaging phantoms using 3D printing has been a dearth of choices for suitable TMMs.

Many TMMs have been proposed for ultrasound phantoms (Madsen et al 2005, Zell et al 2007, Troia et al 2017), but there is a need for new materials, and new methods to use them in phantoms. Currently-used TMMs include silicone (McLeod et al 2016), agar/gelatin (Pavan et al 2010), poly(vinyl alcohol) cryogel (Surry et al 2004, Xia et al 2011), polyvinyl chloride-plastisol (McLeod et al 2015), and urethane rubber (Culjat et al 2010). Identifying a TMM that allows for independent tuning of parameters such as the speed of sound, ultrasonic scattering, and attenuation, and which can be developed from non-toxic, mechanically robust, and inexpensive materials, has been particularly challenging. Longevity and low manufacturing costs are also important considerations. Silicone has been used for sophisticated patient-specific phantoms (Holmes et al 2008, McLeod et al 2012), but its attenuation is high (Browne et al 2003). Aqueous phantoms such as agar and gelatin gels can be tuned so that their ultrasonic properties are similar to those of tissue (Madsen et al 2005), but evaporation limits their longevity and these materials are not mechanically robust.

Recently, several studies have highlighted oil-based materials as promising TMMs. Oudry et al (2009) reported a mineral oil-based TMM that included styrene-ethylene/butylene-styrene (SEBS)—type copolymers to influence the phantoms' elasticity. In a recent study by Cabrelli et al (2016), both the acoustic and optical properties of mineral-oil based mixtures were found to vary with the concentrations of SEBS and low-density polyethylene additives. Vieira et al (2013) used paraffin-gel wax TMMs based on mineral oil to develop heterogeneous phantoms for ultrasound-guided breast needle biopsies. Additives can provide acoustic backscattering; they include silica or graphite powder (Oudry et al 2009) and glass spheres (Vieira et al 2013, Cabrelli et al 2016). However, to the authors' knowledge, there have not been any demonstrations of oil-based TMM phantoms with complex patient-specific geometries.

In this study, we demonstrate that the acoustic properties (acoustic attenuation and backscattering) of a mineral-oil based material called 'gel wax', can be tuned with paraffin wax and glass spheres additives. Additionally, we demonstrate, for the first time, 'gel wax' can be shaped with 3D printed moulds to create anatomically realistic phantoms that are derived from patient-specific anatomy. The developed phantoms cover a wide range of clinical fields including regional anesthesia, cardiac electrophysiology and fetal therapy.

The methodology in this study had four components. First, we investigated how two additives to gel wax could be used to tune acoustic attenuation and backscattering. Second, we measured the Young's modulus of native gel wax. Third, we designed three types of moulds for gel wax, and fabricated them with 3D printing. Finally, we fabricated the phantoms, and validated them with a clinical ultrasound imaging scanner.

2.1. Acoustic properties tuning and characterisation

In its native form, the gel wax used here (Product number: FF1 003, Mindsets Online, Waltham Cross, UK) attenuated and scattered ultrasound far less than typical biological tissues (see section 3). Glass spheres (Boud Minerals, Lincolnshire, UK) with a nominal diameter range of 0–63 µm were added to increase acoustic backscattering, and paraffin wax (Alec Tiranti, London, UK) was added to increase acoustic attenuation.

2.1.1. Gel wax material fabrication

To fabricate the sample mixtures, gel wax was first melted to a temperature greater than 30 °C beyond its melting point (ca. 65 °C), in a glass beaker with a heating plate. Glass spheres and/or paraffin wax were gradually added during continuous mechanical stirring until the solution was visually homogeneous. Subsequently, sonication (CL18, Fisher Scientific, Pittsburgh, PA; 120 W) was performed for 30 s (65% of maximum power) to avoid particle aggregation. After sonication, the solution was reheated and degassed for approximately 2 min until air bubbles were absent with visual inspection. It was then allowed to cool down, and then poured into different types of moulds, as described in the following sections.

2.1.2. Acoustic properties characterisation

Two series of samples were fabricated with different concentrations of paraffin wax and glass spheres. In the first series, paraffin wax concentrations varied from 0 to 8 w/w% and no glass spheres were included. In the second series, the glass sphere concentration was varied from 0.05 to 0.9 w/w% and no paraffin wax was added. When melted, these samples were poured into rectangular moulds. These moulds were created with an acrylic spacer with a nominal thickness of either 2 mm or 5 mm, which was positioned between two glass microscope slides. This spacer had an opening that allowed for gel wax to be poured in between the two slides. The gel wax was cooled by placing it in room temperature water, with the spacer opening facing upward. For characterising acoustic properties, the glass slides were removed, so that a rectangular gel wax slab (ca. 35 mm  ×  50 mm) surrounded by the acrylic spacer remained. The speed of sound and acoustic attenuation were measured using an insertion method (Xia et al 2013) in reflection mode, with a single-piezo transducer (V312-SU, 10 MHz central frequency, Olympus, Shinjuku, Japan). The transducer and a metal reflector were mounted in deionized water at room temperature (ca. 20 °C), so that the acoustic axis of the transducer was perpendicular to the surface of the metal reflector. An ultrasonic pulser/receiver (5077PR, Olympus, Shinjuku, Japan) was used to drive the transducer and receive the echo signal from the reflector. The received echo signal was then digitized at 100 MS s⁻¹ by a digital acquisition card (USB-5132, National Instruments, Austin, USA) and transferred to a personal computer for processing.

For each material composition, pulse-echo ultrasound signals (S₁, S₂) with a pair of two gel wax slabs with nominal thicknesses of 2 mm and 5 mm were acquired separately. Using tabulated values for the speed of sound and the frequency dependent acoustic attenuation of water, c_w and ${{\alpha }_{{\rm w}}}\left(\,f \right)$ (Bilaniuk and Wong 1993), the speed of sound (c_s) was estimated by solving for this variable in equation (1):

$$\begin{align} \newcommand{\e}{{\rm e}} \displaystyle \Delta t=\frac{2({{d}_{2}}-{{d}_{1}})}{{{c}_{{\rm s}}}}-\frac{2({{d}_{2}}-{{d}_{1}})}{{{c}_{{\rm w}}}}\nonumber \end{align} \tag{ 1 }$$

where d₁ and d₂ are the thicknesses of the sample pair (measurements performed on the surrounding rigid acrylic frames with digital callipers), and Δt is the difference in the time-of-flights, which was calculated using cross-correlation of S₁ and S₂. Likewise, the acoustic attenuation of the sample, ${{\alpha }_{{\rm s}}}\left(\,f \right)$, was estimated using equation (2):

$$\begin{align} \newcommand{\e}{{\rm e}} \displaystyle {{\alpha }_{{\rm s}}}\left(\,f \right)=\frac{10}{{{d}_{2}}-{{d}_{1}}}{{\log }_{10}}\left[ \frac{{{I}_{1}}\left(\,f \right)}{{{I}_{2}}\left(\,f \right)} \right]+{{\alpha }_{{\rm w}}}\left(\,f \right)\nonumber \end{align} \tag{ 2 }$$

where I₁(f) and I₂(f) are the power spectra of S₁ and S₂. With most materials, acoustic attenuation follows a power law of the form ${{\alpha }_{{\rm s}}}\left(\,f \right)={{\alpha }_{0}}{{f}^{n}}$, in which ${{\alpha }_{0}}$ is termed the pre-factor, and n a positive exponent. The estimated values of ${{\alpha }_{{\rm s}}}\left(\,f \right) ~$ were fitted to the power law using a non-linear least squares regression algorithm (Matlab, Mathworks, Natick, NH) to estimate ${{\alpha }_{0}}$ and n. The water temperature (T), which was recorded using a digital thermometer (4378, Fisher Scientific, Pittsburgh, UK), was used to estimate c_w (Bilaniuk and Wong 1993). For each sample pair, 5 measurements of S₁ and S₂ were obtained from distinct regions. The acoustic impedance (Z) of native gel wax was calculated using measured densities of 10 samples. The density of each sample was obtained with a digital weight scale (BP211D; 0.01 mg nominal precision; Sartorius, Göttingen, Germany) and a digital calliper (CD-15APX;  ±20 µm nominal precision; Mitutoyo, Kawasaki, Japan).

To assess the acoustic backscattering, we compared ultrasound images acquired from phantoms with varying concentration of glass spheres and 2 w/w% paraffin wax, and from the human arms of three healthy volunteers. The images were acquired using a clinical ultrasound scanner (SonixMDP, Analogic Ultrasound, Richmond, Canada) with a linear array imaging probe (bandwidth: 14–5 MHz; L14-5/38, Analogic Ultrasound, Richmond, BC, Canada). The ultrasound acquisition settings, which included excitation frequency, gain, depth, dynamic range and power, were consistent across images; the speed of sound used for reconstruction was different (gel wax phantoms: 1440 m s⁻¹; human tissues: 1540 m s⁻¹). For each image, the histogram, mean value and standard deviation of the image intensities were calculated. Prior to this, the acquired images were scaled globally so that the minimum and maximum across all images were 0 and 1, respectively.

2.2. Mechanical testing

2.3. Phantom creation and evaluation

The addition of paraffin wax had a strong effect on acoustic attenuation and a weak effect on the speed of sound and acoustic impedance. The acoustic attenuation increased rapidly with the paraffin wax concentration (figure 1; table 1). The frequency dependence of the attenuation was found to be well modelled by the power law. As the concentration of added paraffin wax was increased to 8% of the native weight, the attenuation increased by 308% at 3 MHz from 0.72 to 2.91 dB cm⁻¹ and by 289% at 10 MHz from 6.84 to 26.63 dB cm⁻¹. These ranges of attenuation contain reported values of most soft tissues (Mast 2000, Culjat et al 2010). Over the same concentration range, the speed of sound was not substantially affected (table 1). At 1448.5 m s⁻¹, the speed of sound of native gel wax was 6.3% lower than average speed of sound for soft tissue (1540 m s⁻¹), which is a typical value used for reconstruction in most clinical ultrasound imaging scanners. The measured acoustic impedance of native gel wax, 1.22  ±  0.01 kg m⁻² s⁻¹ $\times$ 10⁶, was lower than that of typical soft tissue (see 1.62 kg m⁻² s⁻¹ $\times$ 10⁶ for muscle (Mast 2000)).

Table 1. Measured acoustic properties of gel wax, in which the only additive was paraffin wax. Each property (mean  ±  standard deviation) was obtained from measurements from five distinct regions. Acoustic attenuation as a function of frequency (f) is depicted by the power law curve (${{\alpha }_{0}}{{f}^{n}}$), with parameters obtained from non-linear least squares fitting.

Paraffin wax concentration (w/w%)	Acoustic attenuation ${{\alpha }_{0}}{{f}^{n}}$ (dB cm−1)		Speed of sound cs (m s−1)
	${{\alpha }_{0}}$	$~n$
0	0.080  ±  0.007	1.93  ±  0.06	1448.5  ±  0.8
2	0.117  ±  0.004	1.86  ±  0.03	1447.9  ±  0.5
4	0.156  ±  0.003	1.86  ±  0.01	1447.9  ±  1.1
6	0.142  ±  0.022	2.09  ±  0.07	1448.4  ±  1.7
8	0.348  ±  0.063	1.88  ±  0.10	1443.1  ±  1.5

Zoom In Zoom Out Reset image size

The addition of glass spheres to provide acoustic backscattering had a prominent effect on the ultrasonic appearance, and it did not substantially affect the speed of sound and acoustic attenuation (table 2). The ultrasound images from gel wax phantoms and human arm tissues had similar image intensity distributions (figures 2(a) and (c)). The mean value of image intensities for gel wax phantoms increased with glass spheres concentration from 0.27  ±  0.12 at 0.025 w/w% to 0.56  ±  0.08 at 0.9 w/w% (figure 2(b)). The mean values of human arm tissues ranged from 0.36  ±  0.19 to 0.53  ±  0.17. At a concentration of 0.025 w/w%, the mean value was lower than the imaged human arm tissues, and at 0.9 w/w% it was higher. The standard deviations of the image intensity values for gel wax phantoms were slightly lower that human arm tissues; this difference can be attributed to the heterogeneity of human arm tissues. Our assessment suggested that 0.5 w/w% provides an acoustic backscattering appearance similar to that of soft tissues.

Table 2. Measured acoustic properties of gel wax, in which the only additive was glass spheres. Each property (mean  ±  standard deviation) was obtained from measurements from five distinct regions. Acoustic attenuation as a function of frequency (f) was well described by the power law (${{\alpha }_{0}}{{f}^{n}}$), with parameters obtained from non-linear least squares fitting.

Glass spheres concentration (w/w%)	Acoustic attenuation ${{\alpha }_{0}}{{f}^{n}}$ (dB cm−1)		Speed of sound cs (m s−1)
	${{\alpha }_{0}}$	$~n$
0	0.111  ±  0.013	1.80  ±  0.07	1445.7  ±  0.7
0.05	0.108  ±  0.012	1.83  ±  0.05	1443.9  ±  0.6
0.1	0.102  ±  0.005	1.87  ±  0.06	1443.5  ±  0.7
0.5	0.106  ±  0.028	1.88  ±  0.11	1443.4  ±  0.3
0.9	0.113  ±  0.009	1.87  ±  0.05	1444.1  ±  0.8

Zoom In Zoom Out Reset image size

Native gel wax was sufficiently robust to maintain its integrity in the presence of compressive deformations up to 60%. Its stress–strain curve was linear (R²  >  0.99) for deformations up to 40%, with a Young's modulus of 17.4  ±  1.4 kPa. For greater deformations, the stress–strain curve was non-linear, with the derivate tripling from 40% to 60%.

In the nerve and blood vessel phantom, all embedded structures were visible on ultrasound (figure 3). The blood vessels presented as hypoechoic; the nerves, as hyperechoic with a slight difference relative to the background that was reminiscent of that seen in some clinical contexts. Slight deviations from circularity were apparent in both vessels and nerves, which may have arisen from heating of these structures during embedding within the surrounding melted background gel wax during fabrication. The diameters of the vessel and nerves, as measured on the ultrasound images, ranged from 4.3 to 6.3 mm. The mean value of 5.6 mm was 12% different from that of the moulds from which they originated. The background had a largely homogeneous appearance but some scatters were present, which may have corresponded to small air bubbles. There was an absence of bubbles around the vessels and nerves. Moreover, no artefacts or prominent acoustic changes at interfaces with the background were apparent.

Zoom In Zoom Out Reset image size

In the cardiac phantom, the surfaces of the gel wax that formed the heart chambers retained the shapes of the corresponding 3D printed mould (figure 4). In particular, the septum was intact, despite having a minimum measured thickness of only 6.6 mm. The interface between the gel wax and the water that filled the heart chambers was slightly more hyperechoic than the tissue-blood interface encountered in clinical practice. The underlying gel wax had a largely homogeneous appearance that resembled the human myocardium. Due to the compliant nature of gel wax, the phantom was readily deformed with palpation, and these deformations were visualised with M-mode imaging.

Zoom In Zoom Out Reset image size

In the placenta phantom, the visual and ultrasonic appearances of the chorionic vessels and underlying chorionic tissue strongly resembled those of the human placenta from which they were derived, according to A.L.D., consultant specialist in fetal medicine. With ultrasound imaging, the superficial chorionic fetal vessels were clearly apparent with hypoechoic centres and hyperechoic boundaries that were recessed from the chorionic surface of the placenta (figure 5). The hyperechoic boundaries of these surface vessels and the chorionic surface of the placenta were larger than those encountered in clinical practice. The background was speckled and largely homogeneous. Visually, the veins were much darker than the arteries; the background was translucent. The gel wax proved to be sufficiently compliant to allow for the arteries and veins to cross over each other, a situation that frequently occurs in the human placenta.

Zoom In Zoom Out Reset image size

Here, we demonstrated for the first time that heterogeneous phantoms with anatomical realism can be generated from gel wax. We found that gel wax is well suited to moulding with 3D printed plastics due to its favourable mechanical properties. These moulds can readily be derived directly from tissue to create patient-specific phantoms. Acoustic attenuation, speed of sound and acoustic backscattering were obtained within the range of values quoted for soft tissues by varying concentrations of two additives: paraffin wax and glass spheres.

The speed of sound and the acoustic impedance of gel wax are lower than that of most soft tissues and water (Mast 2000, Culjat et al 2010). One consequence is that immersion of gel wax phantoms in water can lead to hyperechoic boundaries on ultrasound imaging. These challenges could be mitigated by the use of different surrounding fluids, such as mineral oil. Furthermore, the low speed of sound could affect the scaling of the images during reconstruction. Copolymers could be added to modify the speed of sound of the material (Oudry et al 2009). Here, frequency dependent acoustic attenuation involved a power law exponent (n  ≈  1.9) that falls within the range of values reported for different soft tissue types (1.0–7.4) (Duck 1990, Azhari 2010 (appendix A)). As a result, tuning the acoustic attenuation of gel wax to match specific soft tissue types within a broad frequency range may be challenging.

The native gel wax used in this study is readily deformable. Indeed, its Young's modulus is similar to the average values of 18–22 kPa reported for normal breast fat tissue (Krouskop et al 1998). Its stiffness could be increased significantly by incorporating additives such as paraffin wax and copolymers (Oudry et al 2009). These additives could reduce the range of deformations over which the stress–strain relationship is substantially linear, which could be of interest to introduce non-linearities encountered in soft-tissues during ultrasound elastography studies (Pavan et al 2010).

There are several ways in which the anatomically realistic ultrasound gel wax phantoms in this study could be improved. First, spatial variations in glass sphere concentrations were visible within regions that were designed to be homogenous. These variations likely arose from differences in the densities of glass spheres and the surrounding gel wax. Increasing the mixing speed and increasing the rate of cooling of the gel wax after it is poured in the mould will likely reduce heterogeneities but it might also introduce air bubbles. To mitigate the risk of air bubble formation, it may be advantageous to mix and pour gel wax within low pressure chambers. Second, the acoustic properties (acoustic attenuation and back scattering) in this study were assigned to provide contrast between the different tissue structures, such as blood vessels and underlying background. In future work, acoustic properties values could be assigned to represent specific tissue types.

In this study, anatomically realistic ultrasound phantoms with feature sizes as small as 2 mm were created using gel wax with 3D printed moulds. Direct printing of gel wax could be useful to create structures with greater 3D complexity. One recent step in this direction was provided by Dong et al (2015), who developed a fuse deposition modelling system to fabricate 3D heterogeneous phantoms for optical imaging. Gel wax was mixed with additives and directly printed using a custom print head with multiple channels and an embedded mixing device. We conclude that gel wax is a material which could lead to a new generation of ultrasound phantoms with a step change in anatomical realism.

This work was supported by an Innovative Engineering for Health award by the Wellcome Trust (No. WT101957)and the Engineering and Physical Sciences Research Council (EPSRC) (No. NS/A000027/1), by a Starting Grant from the European Research Council (Grant No. ERC-2012-StG, Proposal 310970 MOPHIM), by an EPSRC First Grant (No. EP/J010952/1), by a Wellcome/EPSRC Centre award [203145Z/16/Z & NS/A000050/1], by a Royal Society Research Grant [RG160569], by the UCL EPSRC Centre for Doctoral Training in Medical Imaging (EP/L016478/1) and by an EPSRC Award (EP/N02124X/1). A.L.D. is supported by the UCL/UCLH NIHR Comprehensive Biomedical Research Centre.