Advancing Phantom Fabrication: Exploring 3D-Printed Solutions for Abdominal Imaging Research

Abstract

Background: The development of novel medical imaging technologies and treatment procedures hinges on the availability of accurate and versatile phantoms. This paper presents a cost-effective approach for creating anthropomorphic abdominal phantoms. Methods: This study proposes a cost-effective method using 3D printing and readily available materials (beeswax, plaster, and epoxy resin) to create high-fidelity anthropomorphic abdominal phantoms. The three-dimensionally printed phantoms exhibited X-ray attenuation properties closely matching those of human tissues, with measured Hounsfield unit (HU) values of −115.41 ± 20.29 HU for fat, 65.61 ± 18.06 HU for muscle, and 510 ± 131.2 HU for bone. These values were compared against patient images and a commercially available phantom, and no statistically significant difference was observed in fat tissue simulation (p = 0.428). Differences were observed for muscle and bone tissues, in which the 3D-printed phantom demonstrated higher HU values compared with patient images (p < 0.001). The 3D-printed phantom’s bone simulation was statistically like that of the commercially available phantom (p = 0.063). Conclusion: This method offers a cost-effective, accessible, and customizable alternative for abdominal phantoms. This innovation has the potential to accelerate advancements in abdominal imaging research, leading to improved diagnostic tools and treatment options for patients. These phantoms could be used to develop and test new imaging techniques with high accuracy.

1. Introduction

The development of novel medical imaging technologies and treatment procedures hinges on the availability of accurate and versatile phantoms. Anthropomorphic phantoms, replicating human anatomy, offer a safe and controlled environment for evaluating these advancements [1]. Ideal phantoms possess two key characteristics: geometric fidelity that precisely mirrors human anatomy and tissue-equivalent materials that mimic the X-ray attenuation properties of real tissues across the relevant energy spectrum [2].

Traditionally, abdominal phantoms have been fabricated using materials like silicone or tissue-equivalent plastics. Hydrogels, including those combined with beeswax, offer improved tissue mimicry with adjustable mechanical properties, making them suitable for various medical applications [3,4]. Silicone phantoms can be prohibitively expensive and time-consuming to manufacture due to complex molding processes. Tissue-equivalent plastics, while more readily available, often require specialized equipment and expertise for handling and manipulation. Additionally, both materials can exhibit limitations in their ability to replicate the full spectrum of tissue properties within the abdomen [5].

A novel and cost-effective approach for creating anthropomorphic abdominal phantoms could be to use a combination of 3D printing, beeswax, and epoxy resin for appendix analysis. Three-dimensional printing technology offers unparalleled design flexibility, enabling the creation of intricate anatomical structures with high fidelity. Beeswax, a natural product derived from honeycombs, has garnered attention for its potential applications in various fields, including biomedical engineering. Its unique composition and properties make it a promising candidate for developing tissue-mimicking materials. Chemically, beeswax is primarily composed of esters, hydrocarbons, and free acids [6]. This complex molecular structure imparts beeswax with distinctive characteristics, such as a high melting point and malleability, facilitating its processing and shaping into desired forms [7]. Epoxy resin, a versatile casting material, provides a robust matrix for embedding the 3D-printed components. Additionally, the properties of the epoxy resin can be further tailored by incorporating specific fillers to mimic the X-ray attenuation characteristics of different tissues within the phantom [8].

Furthermore, Fused Deposition Modeling (FDM) has been extensively reviewed in the context of anthropomorphic phantom creation. FDM is a popular 3D printing technique that uses thermoplastic materials to build objects layer by layer. While it offers advantages such as cost-effectiveness and ease of use, its application in creating high-fidelity phantoms is limited by the resolution and material properties achievable with current FDM printers [9].

Recent developments in medical imaging have made more use of 3D printing technology to create highly accurate anthropomorphic phantoms. These phantoms are essential tools for radiation dosage optimization, especially in complex imaging modalities like CT scans [10]. More efficient dose reduction tactics are made possible by the capacity to closely imitate the qualities of human tissue, improving patient safety without sacrificing diagnostic precision. Even with all the advancements in this industry, the choice of materials and printing methods used to create phantoms continues to have a major impact on how effective these models are. Furthermore, it is still unclear how well 3D-printed phantoms compare to solutions that are readily available on the market and patient-derived data [11].

Table 1. Summary of key studies focused on the development and application of 3D-printed anthropomorphic phantoms in medical imaging.

Source/Comparison	Key Findings	Description
Filippou et al. [12]	Three-dimensional printing offers precision and customization.	Three-dimensional printing enables the production of phantoms that closely mimic human tissue properties, which is crucial for accurate radiation dose measurements.
Wang et al. [13]	Potential for dose reduction through personalized phantom design.	This study highlights the comparison of different materials and printing techniques used to achieve tissue-equivalent properties, emphasizing dose reduction.
Coles-Black et al. [14]	Clinical applications in surgery and radiation therapy.	This article discusses the importance of phantoms in training and preoperative planning, in which accurate tissue simulation is critical for clinical applications.
Higgins et al. [15]	Comparison of commercial and 3D-printed phantoms.	This research compares the pros and cons of commercial and 3D-printed phantoms in terms of cost, accuracy, and clinical utility, providing insights into their relative effectiveness.

summarizes the key findings from this literature review, highlighting the advantages, limitations, and clinical implications of various approaches.

We performed this study to investigate the development and characterization of a 3D-printed composite beeswax–epoxy abdominal phantom for a better analysis of the appendix and scanning parameter optimization.

A crucial aspect of this study involves the comprehensive evaluation of the phantom’s performance. This evaluation encompasses three key areas: anatomical accuracy, mechanical properties, and X-ray attenuation characteristics. The obtained data is compared to reference values for human abdominal tissues, allowing for a comprehensive assessment of the phantom’s ability to mimic real tissue behavior under X-ray irradiation.

2. Materials and Methods

2.1. Characterization of Appendix Simulation in an Anthropomorphic Phantom

The quality of diagnostic images in radiology is generally assessed by three key parameters: noise, contrast, and resolution. In computed tomography (CT), noise, often referred to as image graininess, represents unwanted variations in pixel values within a uniform image region. It is commonly quantified using the standard deviation (SD) of CT numbers within a selected homogenous region of interest (ROI). Noise arises from various sources, including quantum noise, electronic noise, structural noise, and artifacts. In addition to the standard deviation, the signal-to-noise ratio (SNR) is another metric used to quantify noise. Contrast (C) is most often defined as the difference between the mean value of the CT number of the object (N) and the background (N₀). Also, the contrast-to-noise ratio (CNR) is used as a parameter to describe the contrast. Resolution in CT, also called high-contrast spatial resolution, is the system’s ability to distinguish structures from structures that are in proximity. The parameter that describes resolution in CT is the most visible frequency, that is, the maximum number of line pairs that the system can resolve per unit length.

To evaluate the visibility of the printed object that simulated the appendix, the values of CT numbers were evaluated along the x- and y-axes. Figure 1 gives a schematic representation of the evaluation of the circular object corresponding to the appendix. In the idealized case (upper part of the figure), the values of the CT numbers (N) change discretely, so the boundary between the background and the object is infinitely sharp. This transition describes a rectangular Π function. However, the border between the background and the object is not infinitely sharp, but there is a transitional gradient. The function that describes this transition can be expressed as the convolution of Π and the Gaussian function (G), that is:

$$f(x)=\Pi (x)\times G(x;\sigma )$$

(1)

where, for our analysis, the significant parameter σ is a measure of the sharpness of the object display, that is, the resolution of the system.

The result of the convolution of the functions Π and G can be expressed through the error function erf in the following form:

$$f(x)=N_0- {\displaystyle \frac{C}{2}}\left\{\left[\left.1-\mathrm{erf}\left({\displaystyle \frac{x-\frac{D}{2}-x_0}{\sigma }}\right)\right]-\left[1-\mathrm{erf}\left({\displaystyle \frac{x+\frac{D}{2}-x_0}{\sigma }}\right)\right.\right]\right\}$$

(2)

The parameters x₀ and N₀ represent the offset for the x- and y-axis, and D is the diameter of the object.

Using the non-linear approach method, it is possible to determine the parameters of the function f(x) so that it best describes the actual measurement results (CT number values). For this purpose, this study used Gnuplot software 6.0 (Release 6.0.1), whose algorithm changes the values of the parameters through an iterative process until the sum of the squared deviations of the actual values from the curve is less than the given limit, which, in this work, had a value of 10⁻⁶. The results of the nonlinear iterative fitting of the curve are the experimentally determined values of the parameters σ, C, D, and others, as shown in Figure 2.

2.2. Material Selection and Characterization

Commercially available beeswax pellets were chosen for their tissue-equivalent properties. Prior to use, the beeswax was meticulously characterized to determine its X-ray attenuation properties. This characterization involved creating a series of calibrated samples with varying thicknesses [16]. These samples were then irradiated using an X-ray source with a known energy spectrum. The resulting attenuation data were used to calculate the Hounsfield unit (HU) values of the beeswax material, allowing for comparison with the HU values of human adipose tissue at the relevant X-ray energies [17].

A two-part epoxy resin system (Pan Asel Chemicals Sdn. Bhd Company, Setapak, Federal Territory of Kuala Lumpur, Malaysia) was selected for its casting properties. The specific resin and hardener combinations were chosen on the basis of their compatibility with beeswax, cure time, and mechanical properties once cured. Like beeswax, the epoxy resin was characterized to determine its X-ray attenuation properties [18]. The characterization process involved creating a set of calibrated epoxy castings with varying thicknesses and filler materials. The attenuation data obtained from these samples was used to calculate the HU values of the epoxy resin across the X-ray energy spectrum.

Standard casting plaster was utilized to create a bone mold within the abdominal cavity. The plaster offered a cost-effective and readily available solution for replicating the high-density nature of bone tissue [18]. While plaster itself does not perfectly mimic bone in terms of HU values, it provides a suitable base for further refinement using additional contrast agents within the final casting.

2.3. Printing of the Abdominal Mold

This research employed a custom-designed 3D printer to fabricate the foundation of the abdominal phantom. The printer utilized a fused deposition modeling (FDM) technique, offering a balance between affordability, resolution, and material compatibility. We used the custom 3D printer Signage 3D S1 with a build volume of 1200 × 800 × 250 to build the mold shown in Figure 3.

The design of the 3D model for the abdominal mold was generated using DICOM images from the CT scans of a reference anatomy. Segmentation techniques were employed to isolate individual organs and structures within the abdominal cavity using 3D Slicer, a free open-source software application; 3D Slicer’s versatile image processing capabilities enabled the precise delineation of abdominal organs from CT scan data. The segmented volumes were subsequently exported as STL files for additional optimization, ensuring the accurate representation of the target anatomical structures.

These segmented regions were then converted into 3D models using Meshmixer software Version 3.5, which allowed for precise plane cuts to optimize model orientation for subsequent 3D printing. This software enabled the repair of any imperfections, such as holes and inconsistencies, ensuring optimal printability. Additionally, Meshmixer was used to manipulate model geometry, including scaling and combining components, to create the desired phantom configuration. The software’s slicing capabilities were instrumental in generating support structures, enhancing the overall print quality.

The final 3D model incorporated features that facilitated the subsequent casting process, such as registration points for alignment and channels for introducing the casting materials.

2.4. Phantom Fabrication

The 3D model of the abdominal mold was printed using the custom 3D printer and the chosen filament material. The printing parameters were optimized to ensure proper dimensional accuracy, surface finish, and the structural integrity of the mold.

The molten beeswax was carefully poured into the 3D-printed mold, filling the designated cavities intended to represent adipose tissue. The mold was designed to allow for controlled pouring and prevent air bubbles within the beeswax casting. Once filled, the beeswax was allowed to cool and solidify.

The epoxy resin was mixed with the chosen muscle-mimicking filler material in a predetermined ratio to achieve the desired X-ray attenuation properties. This mixture was then poured into the remaining cavities of the mold, ensuring complete encapsulation of the beeswax casting. The epoxy resin was allowed to cure completely, resulting in a solidified structure mimicking skeletal muscle tissue.

A separate plaster mold was created within the abdominal cavity of the assembled beeswax and epoxy resin structure. This plaster mold served as a base for replicating the high-density nature of bone tissue.

Following the creation of the plaster mold, a contrast agent with HU values closer to those of bone tissue was incorporated into the final epoxy resin casting. This casting was then poured into the plaster mold, replicating the location and density of bones within the abdominal cavity. The final epoxy resin casting was allowed to cure completely.

After all individual components were cured and solidified, the different sections of the phantom were assembled to form the complete abdominal structure.

3. Results

This study investigated the impact of various scanning and reconstruction parameters on the image quality of phantom abdominal scans.

Table 2. Technical parameters of computed tomography scanning and reconstruction used for the abdominal phantom: tube voltage (U), tube current (I), rotation time (t), pitch (p), slice thickness (T), computed tomography air kerma index (CTDI), and reconstruction kernel for different scanning series.

		U (kV)	I (mA)	t (ms)	p	T (mm)	CTDI (mGy)	Kernel
Series	1	80	300	750	0.816	0.5	3.9	FC18
	2	100	153	750	0.816	0.5	4.9	FC18
	3	120	87	750	0.816	0.5	6.0	FC18
	4	135	80	750	0.816	0.5	7.6	FC18
	5	80	300	750	0.816	0.5	3.9	FC08
	6	80	300	750	0.816	1.0	3.9	FC18
	7	100	153	750	0.816	0.5	4.9	FC08
	8	100	153	750	0.816	1.0	4.9	FC18
	9	120	87	750	0.816	0.5	6.0	FC08
	10	120	87	750	0.816	1.0	6.0	FC18
	11	135	80	750	0.816	0.5	7.6	FC08
	12	135	80	750	0.816	1.0	7.6	FC18

summarizes the utilized settings, including the tube voltage (kV), tube current (mA), rotation time, pitch, slice thickness, computed tomography air kerma index (CTDI_vol), and convolution kernel.

Automatic tube current modulation (ATCM) was employed during scanning. The CT scanner’s ATCM system utilizes standard deviation (SD) as an image quality metric, with a pre-defined threshold of 12.5 in this study. This resulted in a dynamic anode current range of 50 mA to 300 mA, adapting to each layer within the ROI (as shown in Table 2). To assess reproducibility, specific series were scanned multiple times. The FC18 convolution kernel, known for its noise reduction and smooth image generation, was used as the standard kernel. Conversely, the FC08 kernel was employed for scenarios requiring enhanced sharpness and higher resolution.

The evaluation results of the visualization of the appendix simulation obtained by non-linear fitting, described through parameters C, D, and σ, are presented in

Table 3. Visualization parameters of appendix simulation: contrast (C), diameter (D), and parameter σ from function [7] for different series of scans at different values of tube voltage (U), tube current (I), computed tomography air kerma index (CTDI), and slice thickness (T).

		U	I	CTDI	T	σ	D	C
		(kV)	(mA)	(mGy)	(mm)	(HU)	(mm)	(HU)
Series	1	80	300	3.9	0.5	1.00	7.46	810
	2	100	153	4.9	0.5	1.01	7.55	877
	3	120	87	6.0	0.5	0.93	7.35	884
	4	135	80	7.6	0.5	0.80	7.50	865
	5	80	300	3.9	0.5	1.00	7.46	808
	6	80	300	3.9	1.0	1.09	7.65	825
	7	100	153	4.9	0.5	1.01	7.56	874
	8	100	153	4.9	1.0	1.02	7.68	879
	9	120	87	6.0	0.5	0.93	7.22	883
	10	120	87	6.0	1.0	1.00	7.58	876
	11	135	80	7.6	0.5	0.80	7.50	864
	12	135	80	7.6	1.0	0.84	7.75	869

. Figure 4 shows how the contrast, that is, parameter C, changes depending on the anode voltage. Contrast (C), defined as the difference between the mean values of the CT numbers in the selected object (the simulation of the appendix) and the background (the simulation of the soft tissue in the abdomen), is an indicator in assessing the ability of the system to distinguish objects from the background. It can be observed that there is a significant positive correlation between the two variables. The mean value of parameter C at 80 kV is 818 HU, while at other values of the anode voltage, it is significantly higher. Thus, at 120 kV, which is a value often used in common scanning protocols, C has a mean value of 881 HU. This dependence on the anode voltage is expected, considering that the general definition of Hounsfield units depends on the radiation energy and thus on the anode voltage. Also, the possible dependence of the parameter C on the convolution kernel was examined, but no significant difference between the FC18 and FC08 kernels was observed (Mann–Whitney U test, p = 0.594).

The sharpness of the appendix display, that is, the value of the parameter σ, did not depend on the tube voltage. A significant correlation cannot be observed in Figure 5. Also, there is no significant difference between the value of σ obtained using different convolutional kernels in image reconstruction (Mann–Whitney U test, p = 1.000).

The third parameter of interest, which is related to the visualization of the appendix, is the diameter of the visible object (D). No significant difference was expected between the different series of scans, which was confirmed by the correlation test between U and D (p = 0.357) as well as by the comparison between different convolutional kernels (Mann–Whitney U test, p = 1.000), which is also observed in Figure 6.

The 3D-printed phantom exhibited excellent morphological similarity to both a human abdomen and a commercially available phantom, as confirmed by the HU values of CT imaging.

Table 4. Comparison of the CT numbers, expressed in terms of Hounsfield unit mean values ($\overline{x}$) and standard deviation (σ) in 100-pixel regions of interest, for the images of the patient: commercially available phantom and 3D-printed phantom.

Evaluated Object	Fat Tissue $(\overline{\mathit{x}}$ ± σ)	Muscle Tissue $(\overline{\mathit{x}}$ ± σ)	Bone $(\overline{\mathit{x}}$ ± σ)
Patient images	−113.6 ± 10.4	49.72 ± 14.7	376 ± 120.6
3D-printed phantom	−115.41 ± 20.29	65.61 ± 18.06	510 ± 131.2
Commercially available phantom	−74.78 ± 12.83	56.34 ± 12.6	541 ± 101.8
p-Values of Student’s t-test
- Patient vs. 3D-printed phantom	0.428	<0.001	<0.001
- Patient vs. commercially available phantom	<0.001	<0.001	<0.001
- 3D-printed vs. commercially available phantom	<0.001	<0.001	0.063

presents a comparison of CT numbers (measured in Hounsfield units, HUs) across images from a patient, a commercially available phantom, and a 3D-printed phantom. The table aims to evaluate the accuracy of the phantoms in replicating the CT numbers of different tissues—fat, muscle, and bone—found in a patient.

The 3D-printed phantom, which used beeswax to simulate fat tissue, showed a close match to the patient’s fat tissue, with no statistically significant difference (p = 0.428). This indicates that beeswax is an effective material for simulating fat tissue in CT imaging. For muscle tissue, the 3D-printed phantom’s resin material exhibits a slightly higher HU value compared with the patient’s muscle tissue, but this difference remains within one standard deviation, suggesting reasonable accuracy in the simulation. Similarly, for bone tissue, the 3D-printed phantom showed a higher HU value than the patient images, but this value was still within an acceptable range.

By contrast, the commercially available phantom demonstrated a notable discrepancy in representing fat tissue, with a difference of more than two standard deviations from the patient’s values, indicating a less accurate simulation. However, the commercially available phantom more accurately represented muscle tissue, with HU values closer to the patients. For bone tissue, the difference between the 3D-printed phantom and the commercially available phantom was not statistically significant (p = 0.063), suggesting that both phantoms provide a similar level of accuracy in bone tissue simulation.

The measurement region used to assess the phantom’s conformity to real patients is shown in Figure 7. The ROI is a 10-pixel-tall rectangle that covers areas in the vicinity of the abdomen, including the surrounding air, muscle, soft tissue, and fat tissue. The same regions were evaluated on images belonging to a real patient (used to make a realistic design), a commercially available adult phantom, and the 3D-manufactured abdomen phantom. Measurements made in the ROI provide valuable data regarding the quality of the manufactured phantom.

The 3D-printed anthropomorphic phantom demonstrated high accuracy. Figure 8 shows the measurement results of the ROI. All curves start from −1000 HU (air). All curves have a 15 mm-wide plateau at approximately −100 HU, which represents the abdominal fat tissue. It should be noted that the abdomen shape can be different from one patient to another.

4. Discussion

This research presented the development and characterization of a novel anthropomorphic abdominal phantom fabricated using a combination of 3D printing, beeswax, and epoxy resin. This approach offers a promising alternative to traditional phantom fabrication techniques, potentially accelerating advancements in medical imaging research and development. Here, we delve into the key findings, the limitations of the current design, and exciting possibilities for future exploration [19,20,21]. The success of the phantom depends on its ability to accurately mimic the human abdominal anatomy and its X-ray attenuation properties [22].

The 3D-printed mold design should be meticulously compared to the reference anatomy used for its creation. High-resolution imaging techniques, such as micro-CT, can provide a detailed comparison of the phantom’s geometry to real human abdominal anatomy. This analysis should assess the level of detail captured in the mold, particularly focusing on the accurate representation of organs, blood vessels, and other critical structures [23,24,25].

It is important to acknowledge potential limitations in replicating specific anatomical features. For instance, intricate structures like vascular networks or the intricate folds of the small intestine might require further refinement of the 3D printing process or the use of additional casting materials. Future studies could explore techniques like microfluidic channels within the epoxy resin to create more realistic vascular phantoms [26,27].

The mechanical properties of the phantom, particularly its ability to mimic the elasticity and pliability of real abdominal tissues, are crucial for simulating realistic interaction with X-ray beams. Indentation testing with probes of varying stiffnesses can be employed to evaluate the phantom’s response to pressure, comparing the obtained data to established values for human abdominal tissues like fat and muscle. Tensile testing of isolated sections of the phantom can further assess its ability to stretch and deform under stress [28].

Future research could explore incorporating materials with variable mechanical properties within the epoxy resin. For instance, hydrogels with tunable stiffness could be used to mimic the elasticity of different soft tissues.

The most crucial aspect of the phantom’s performance lies in its ability to mimic the X-ray attenuation properties of real human tissues. The X-ray attenuation characterization process should be thoroughly discussed, including the specific energy spectrum used and the methods employed for calculating Hounsfield units (HUs) [29,30,31].

A comprehensive comparison of the HU values obtained for the beeswax, the epoxy resin with and without fillers, and the final phantom to the reference values for human adipose tissue, skeletal muscle, and bone tissue across the relevant X-ray energy range is vital. This comparison should highlight the success of the chosen materials and characterization techniques in achieving tissue-equivalent properties [32].

Potential deviations from the reference HU values should be acknowledged and discussed. Discrepancies for specific tissues or energy ranges could be attributed to the limitations of the beeswax or the chosen filler materials. Future studies could explore alternative fillers within the epoxy resin or investigate surface treatments for the beeswax to achieve a closer match to specific tissue types [33,34,35]. Beeswax has a density typically around 0.95 to 0.97 g/cm³, which is slightly lower than that of soft tissues, making it a suitable material for simulating certain tissue types. Its organic composition, primarily consisting of esters, fatty acids, and hydrocarbons, provides a tissue-equivalent attenuation profile in imaging modalities like CT. The attenuation coefficient is influenced by the wax’s composition and can be adjusted by incorporating additives or blending with other materials to match the specific attenuation profiles required for the phantom [36,37].

The cost benefits associated with using readily available materials like beeswax and readily accessible fabrication techniques like 3D printing compared to the cost benefits of using traditional materials (e.g., silicone and tissue-equivalent plastics) should be further investigated. A cost analysis comparing the materials and fabrication processes can be presented, showcasing the potential for significant cost savings.

5. Conclusions

This research successfully demonstrated the feasibility and effectiveness of a novel approach for creating anthropomorphic abdominal phantoms using a combination of 3D printing, beeswax, and epoxy resin. The X-ray CT characterization revealed excellent morphological similarity between the 3D-printed phantom and a human abdomen. The analysis of plot profiles further suggests that the smooth transitions between materials in the phantom may contribute to a more accurate representation of tissue attenuation compared with traditional phantoms.

Further research is warranted to explore the full potential of this approach. This includes the optimization of material properties, the incorporation of biocompatible filaments, and validation with advanced imaging modalities. By addressing these future directions, this research can contribute to the development of even more versatile and clinically relevant phantoms for advancing medical imaging research and ultimately improving patient outcomes.