1. Introduction
Medical Imaging Phantoms (MIPs) are specially designed models used to act as proxies for human tissues, playing a crucial role in calibrating and validating medical imaging devices and in the training of medical professionals [
1]. Traditionally, MIPs were made using industrial techniques such as casting and molding, which limited the options to make patient-specific phantoms as they were often mass-produced. With the advent and improvement of 3D printing technology and increased accessibility, a transition has been made to developing 3D-printed MIPs. Along with this have come many benefits, including the ability to quickly produce patient-specific phantoms using previous imaging studies, change the properties of different parts of the MIP to imitate those of human tissues, and incorporate contrast agents [
2,
3,
4].
Recent studies have demonstrated the utility of 3D-printed phantoms in various medical imaging applications. 3D-printed phantoms have been used in radiation therapy planning, dosimetry, evaluating the performance of computed tomography (CT) scanners, and developing dedicated nuclear medicine protocols and geometries [
5,
6,
7,
8]. The findings of these studies underscore the potential of 3D printing to enhance the customization and functionality of MIPs. Despite numerous studies, a gap in the literature presents itself in the use of contrast with MIPs, with most studies focusing on developing standardized tools to evaluate imaging modalities, compared to the application of modifying contrast concentration to mimic the physiologic processes as seen in our study. One such study by Driscoll et al. highlights the development of a dynamic flow imaging phantom using dynamic contrast-enhanced CT to compare imaging protocols [
9].
Due to the high level of customization that comes with 3D-printed materials, a field of study has emerged to understand the possible applications of these materials in conjunction with contrast agents to improve visualization of sub-structures and mimic pathophysiologic processes [
10,
11]. In this study, the CT attenuation characteristics of contrast-injectable, chambered 3D-printed phantoms were assessed to aid in the construction of custom tissue-mimicking phantoms.
Many additive manufacturing techniques can be used in the fabrication and development of MIPs. Fused deposition modeling (FDM), stereolithography (SLA), and material jetting are three commonly utilized methods of 3D printing for biomedical applications [
12,
13,
14]. FDM printing is carried out by sequentially extruding heated filament into a series of progressively layered two-dimensional patterns to construct a three-dimensional object. FDM is often the least complex and most cost-effective 3D printing method out of the three aforementioned techniques [
15]. However, limitations in print resolution inhibit the accuracy with which FDM models can mimic complex anatomical structures [
15,
16]. Chances for leakage across the tumor phantom chambers also increase with FDM printing due to gaps that commonly form between adjacent extruded layers of the model [
17]. SLA and material jetting, on the other hand, often utilize photopolymer resins that are cured upon exposure to UV light. During SLA printing, UV light is projected repeatedly into a vat of photopolymer resin until each layer of a model is cured on top of the next [
15]. In mixed jetting printing, droplets of the material in use are deposited in the pattern of a model layer and sequentially cured with UV light [
12]. Both techniques offer exceptional printing accuracy and resolution [
15,
18]. The curing process also chemically bonds each layer of photopolymer resin to the next, creating a less porous print that can contain liquid more effectively than an FDM model [
19]. SLA printing is also more cost-effective than mixed jetting [
20]. Thus, SLA printing was utilized for this study due to its high resolution, which is ideal for the intricacies of anatomical structures, and superior fluid-containing capabilities.
Tough 1500 resin was implemented for phantom material due to its strength and pliability, as well as its consistent radiopacity, with an average attenuation value of 101.4 HU, closely mimicking soft tissues, making it suitable for this application [
21,
22]. A preliminary model was printed in BioMed Durable resin. However, more translucent resins, such as BioMed Durable resin, tend to cause complications when printing hollow parts. The curing light from an SLA printer can pass through a clear section of the print and over-cure resin behind it [
19]. This phenomenon can result in the sealing of smaller hollow structures like small blood vessels within the tumor phantom. The opaque nature of Tough 1500 made it a superior alternative to BioMed Durable resin.
A challenge currently faced by this emerging field that has not been fully addressed in the current research is the variability in contrast absorption and dispersion that occurs with different 3D-printed materials and techniques. In this study, chambered tumor phantoms were printed and injected with varying contrast solutions. By improving the understanding of contrast-injectable 3D-printed phantoms, our study aims to demonstrate the versatility of 3D printing in mimicking physiologic tissues and contrast distribution. These prints can be designed to enable contrast injection to specific parts of a phantom. Contrast concentrations in the injected solution can also be varied to achieve predetermined attenuation characteristics, further enhancing the tissue-mimicking and optimization capabilities of 3D-printed phantoms. The first set of goal attenuations for the tumor phantom was chosen by determining the attenuations exhibited by soft tissues (40 HU) and contrast-infused blood (150–200 HU) based on current literature [
23,
24]. The second set of goal attenuations was chosen based on attenuation ranges not represented by the first set of tumor phantom phases (90–150 HU) and common attenuation of low-density bone (400 HU) [
25,
26]. An illustration of the workflow to manufacture a chambered contrast-injectable 3D-printed phantom is shown briefly in
Figure 1.
3. Results
3.1. Contrast Injection and Phantom Attenuation
All six phases of the 3D-printed tumor phantom were visible on CT, as depicted in
Figure 4. Despite observable leakage between chambers on CT imaging, the injected chambers in each phase were filled to capacity. Tumor phase I exhibited a mean HU of 37.49 ± 3.95, venous phase I exhibited a HU of 200.50 ± 4.03, and arterial phase I exhibited a HU of 227.92 ± 5.21. Tumor phase II exhibited a mean HU of 326.20 ± 2.70, venous phase II exhibited a HU of 91.32 ± 1.90, and arterial phase II exhibited a HU of 132.08 ± 5.43 (
Table 3).
As expected, the attenuation values reflected the relative concentrations of contrast solutions utilized for each phase. Tumor phase II was injected with the most concentrated contrast solution with a v/v percentage of 3.61% and exhibited the highest attenuation value (326.20 ± 2.70) out of the three phases. Tumor phase I was injected with the least concentrated solution at 0.50%, resulting in the lowest mean attenuation value at 37.49 ± 3.95. The other tumor phantom phases exhibited increasing observed mean attenuation as their respective injected contrast solution concentrations increased. Tumor phantom phases arranged in ascending order of their exhibited mean attenuation are outlined as follows: tumor I, venous II, arterial II, venous I, arterial I, and tumor II.
The percent differences between goal attenuation and observed attenuation were 6.48% for tumor phase I, 13.05% for arterial phase I, 28.82% for venous phase I, 20.32% for tumor phase II, 12.71% for arterial phase II, and 1.45% for venous phase II. Venous phase II had the lowest percentage difference, whereas venous phase I exhibited the highest. The mean percent difference decreased from 16.1% for the phase I scans to 11.5% for the phase II scans. Although percent differences were considerably high, the observed attenuation values were similar to goal attenuation values with regard to relative HU differences between phases. The difference between observed arterial I and venous phase I attenuation values was still considerably smaller (28 HU) than the difference between observed tumor I and arterial phase I values (191 HU) or tumor and venous phase I values (163 HU). Similarly, observed arterial II and venous phase II attenuations had a relatively low difference of 41 HU, while the attenuation differences between tumor phase II and arterial phase II (194 HU) or tumor phase II and venous phase II (235 HU) were relatively larger. For both sets of scans, the relative differences in observed mean attenuations reflect the attenuation differences between goal attenuation values.
3.2. Attenuation versus Concentration
Phantom mean attenuations for each concentration were consistently higher compared to syringe attenuations except for the tumor phase I solution (
Figure 7). The difference in mean attenuations between the phantom and syringe datasets proved to be statistically significant upon analysis with a paired
t-test at a significance threshold of 0.05 (
p-value = 0.03137).
The observed mean attenuation of each phase increased for both the tumor phantom and syringes as contrast solution concentration increased (
Table 5,
Figure 7). R-values were 0.990 for tumor phantom attenuation and 0.995 for syringe attenuation, indicating a strong positive linear correlation between attenuation and concentration for both datasets. Phantom attenuation and syringe attenuation plotted against concentration yielded R-squared values of 0.981 and 0.991, which convey that changes in concentration are highly predictive of changes in attenuation. Trendlines for phantom attenuation versus concentration and syringe attenuation versus concentration exhibited slopes of 90.82 and 78.99, respectively.
The variation in phantom and syringe attenuation data acquired by both readers was visualized by plotting the data collected by each investigator against each other for all six phases (
Figure 8). The ICC values for tumor phantom and syringe attenuations were 0.9998415 and 0.9987372, respectively, which are within the ‘excellent’ reliability range (ICC = 0.90–1.00) (
Table 6).
4. Discussion
The objective of this study was to understand the attenuation characteristics and customization capabilities of contrast-injectable, additively manufactured imaging phantoms. A major benefit of additive manufacturing imaging phantoms is the wide selection of available 3D printing materials with diverse radiological properties. This study aimed to expand upon this advantage by enabling the static, intrinsic attenuation characteristics of a 3D-printed material to be customizable via contrast injection.
The final observed attenuations of each phase (tumor I: 37.49 HU, tumor II: 326.20 HU, venous I: 200.50 HU, venous II: 91.32 HU, arterial I: 227.92 HU, arterial II: 132.08 HU) were different from their respective goal attenuations (40 HU, 400 HU, 150 HU, 90 HU, 200 HU, 150 HU) with only venous phase II exhibiting a percent difference less than 5%. The relative differences in observed attenuation between phases proved to be similar to differences in goal attenuation. CT imaging often relies on the differences in attenuation values between tissues to enhance understanding of underlying pathology. For example, liver tissue is largely homogeneous in non-contrast CT. The triple-phase liver protocol is often used to differentiate between healthy liver parenchyma and a cancerous lesion. Some studies have shown a mean difference of around 26 HU between a liver lesion and the surrounding healthy parenchyma in arterial phase scans [
23]. A 3D-printed phantom with sealed chambers can be injected with different contrast solutions into adjacent chambers to mimic pathologic lesions adjacent to normal physiologic tissue. The difference between the observed arterial and venous phase attenuations in this study was around 28 HU, proving contrast-injectable 3D-printed phantoms capable of reproducing attenuation differences adjacent physiologic tissues can exhibit on CT imaging.
The absolute attenuation values of the tumor phantom proved to be within a useful physiologic attenuation range. The tumor phase I scan exhibited an attenuation of 37.49 HU, which is close to the attenuation values for blood (38.89 HU), renal parenchyma (renal cortex and medulla, 33.2 ± 4.4 HU and 34.2 ± 4.8 HU, respectively), spleen (37.6 HU), unenhanced liver parenchyma (right lobe of liver, 47.5 HU), and psoas muscles (44.0–44.4 HU) [
29,
30,
31]. The attenuation values observed in the arterial I/II and venous phase I/II scans (201.42 HU, 91.32 HU, 229.3 HU, 132.08 HU) correlated closely with the arterial and venous phase aorta CT attenuation ranges noted in pancreatic cancer patients [
32,
33]. Hepatocellular carcinoma and contrast-enhanced liver parenchyma attenuation values during hepatic arterial and venous phases (80 HU to 120 HU, 60 HU to 130 HU) were similar to arterial and venous phase II scan attenuations [
34]. The mean attenuation exhibited by tumor phase II of 326.20 HU also falls within the range of HU values for trabecular bone [
35]. Thus, a single 3D-printed tumor phantom with contrast compatibility can represent a wide array of physiologic and pathologic tissues.
The tumor phantom utilized in this study was constructed out of FormLabs Tough 1500 resin. This photopolymer has an average attenuation value of 101.4 HU [
22]. Through the injection of solutions with varying iodinated contrast concentrations into the chambered phantom, the attenuation of the phantom was modified to have a range of 37.49–326.20 HU. A solid tumor phantom made of the same Tough 1500 material, on the other hand, could only represent tissues with an attenuation of 101.4 HU.
Phantom attenuation exhibited a strong positive correlation with contrast solution concentration, as depicted in
Figure 7. An R-squared value of 0.986 also indicates that contrast concentration is a reliable predictor of attenuation even when contrast is injected into a chambered 3D-printed phantom. Although attenuation was consistently higher in the tumor phantom compared to the syringes, this statistically significant difference is likely attributable to inadequate agitation of solutions before scanning and CT artifact noted within syringes. The similarity in the rate of change of both phantom and syringe attenuation trendlines—90.82 and 78.99, respectively—confirms that contrast affected attenuation in a largely similar, predictable manner within both containers. The tumor phantom in this study exhibited a 90.82 HU increase in attenuation for every 1% increase in concentration. Consequently, the trends gathered from the attenuation values in this study can be utilized to make contrast solutions that can achieve a predetermined HU value within the 3D-printed phantom. Once the change in a specific 3D-printed phantom’s attenuation relative to contrast concentration is determined, its tissue-mimicking capabilities can be enhanced significantly by simply varying the composition of the solution injected.
This study is not without limitations. Leakage between chambers was observed, which could have influenced the accuracy of the attenuation measurements. Additionally, the range of contrast concentrations tested was limited, and only one type of 3D printing resin was utilized. Attempts to conduct a dimensional analysis were also limited by the lack of clear borders when segmenting the contrast-filled tumor phantom cavities. Future studies should address these limitations by using larger sample sizes, improving the sealing of chambers, and exploring a broader range of contrast concentrations and materials. More robust methods for a dimensional comparison between the tumor phantom and the segmented patient anatomy should also be explored.