1 Introduction

Anthropomorphic mannequins, also known as phantoms, are essential tools in the development, calibration, and evaluation of medical imaging techniques, dosimetry, and radiotherapy treatment planning. These models allow for the controlled reproduction of the geometry, density, and radiological properties of human tissues, providing a safe and reproducible experimental environment for the validation of diagnostic and therapeutic methods [1, 2]. Their use is particularly relevant in modalities such as computed tomography (CT), magnetic resonance imaging (MRI), nuclear medicine, and radiotherapy, where accurate tissue characterization is fundamental to ensuring the quality and safety of clinical procedures [2, 3].

The use of mannequins in medical, educational, and research contexts has become a key strategy in the training of healthcare professionals [4, 5]. These devices allow for the controlled simulation of complex clinical scenarios, facilitating the development of technical skills, clinical decision-making, and the evaluation of protocols without compromising patient safety [6, 7]. In this regard, medical simulation has evolved from low-fidelity static anatomical models to advanced, high-fidelity simulators capable of reproducing physiological functions, dynamic responses, and pathological conditions in real time, even integrating computational systems and sensors for immediate feedback [8, 9].

Traditionally, mannequins used in medical applications are manufactured using specialized industrial processes that employ composite materials, epoxy resins, and polymers designed to mimic the physical and radiological properties of human tissues. However, these models often involve high costs, long manufacturing times, and limited customization capabilities, restricting their use in research or teaching environments with specific requirements [8, 9]. Furthermore, the reproducibility of complex internal structures, such as trabecular bone, presents a significant challenge using conventional manufacturing techniques.

In recent years, additive manufacturing has emerged as a disruptive technology in the biomedical field, enabling the fabrication of customized anatomical models with high geometric accuracy and significantly reduced costs. Fused deposition modeling (FDM) has become one of the most accessible and widely used techniques due to its low cost, ease of operation, and compatibility with a wide range of thermoplastic materials such as ABS, PLA, and TPU [10,11,12]. This process is based on the layer-by-layer deposition of molten material from a three-dimensional (3D) digital model, enabling the fabrication of complex structures, including those with controlled internal geometries [13].

The application of 3D printing in medicine has experienced exponential growth in the last decade, ranging from the generation of patient-specific anatomical models for surgical planning to the development of customized prostheses, surgical guides, and clinical training simulators [14,15,16]. In the field of radiological mannequins, several studies have demonstrated the feasibility of using additive manufacturing to reproduce the attenuation, density, and texture properties of biological tissues through the appropriate selection of materials and 3D printing parameters [17, 18]. Furthermore, the incorporation of biomimetic internal structures, such as trabecular patterns, has been explored using mathematical algorithms and generative design techniques, which improves the anatomical and functional fidelity of the models [19].

With the introduction of 3D printing, the manufacture of medical mannequins has undergone a significant transformation, enabling not only a reduction in costs and production times, but also the customization and replicability of complex anatomical models. This technology has complemented and, in some cases, replaced traditional manufacturing methods, facilitating the production of mannequins geared towards both specific clinical applications and quality assurance in medical imaging and radiotherapy systems [20, 21].

Despite these advances, one of the main challenges in mannequin manufacturing using additive manufacturing lies in evaluating the reproducibility of the process, particularly regarding the physical, mechanical, and radiological properties of the materials used. Factors such as variability in printing parameters, the anisotropy inherent in the FDM process, and structural heterogeneity can influence the consistency of the manufactured models, which is critical in applications requiring high precision and repeatability [22,23,24, 27].

In this context, this work aims to evaluate the feasibility and reproducibility of a heterogeneous anthropomorphic mannequin using FDM technology. The proposed model employs various silicone/TPU/ABS materials that simulate both soft and hard tissues, through a layered configuration that reproduces anatomical layers such as skin, muscle, and bone, respectively. Additionally, an internal trabecular structure is incorporated, generated from a mathematical model whose numerical solution emulates the trabecular phase of a long bone, to increase the degree of biomimicry and approximation to real bone microarchitecture. Finally, this research also proposes, as a proof of concept, a methodology for the design and manufacture of more realistic, reproducible, and accessible clinical mannequins that may be used as support tools in medical training, experimental research, and future validation studies in CT imaging.

2 Materials and methods

Materials: Two commercially available thermoplastic filaments were used to fabricate the samples using 3D fused deposition modeling (FDM): thermoplastic polyurethane (TPU) and acrylonitrile butadiene styrene (ABS) (see Table 1). The elastomeric filament was Kingroon TPU, with a nominal diameter of 1.75 mm. This material has a typical hardness of 95 A (Shore), high abrasion resistance, and high elastic deformation capacity, making it suitable for applications requiring flexibility and energy absorption. Its recommended extrusion temperature range is 190–230 °C, with a bed temperature between 60 and 80 °C. The reported density is approximately 1.20 g/cm³.

On the other hand, XTZL3D ABS filament, also with a nominal diameter of 1.75 mm, was used. ABS is a rigid thermoplastic polymer widely used in additive manufacturing due to its good mechanical strength, thermal stability, and ease of post-processing. It has an approximate density of 1.04 g/cm³, with an extrusion temperature range of 220–260 °C and a bed temperature between 80 and 110 °C. However, this material is susceptible to thermal shrinkage, which can cause deformation if printing conditions are not properly controlled.

Additionally, liquid silicone (room-temperature curing elastomer, RTV) was used to fabricate a surface layer that simulates skin tissue. This material was selected for its low hardness, high flexibility, and ability to reproduce fine details, allowing for a skin-like texture. The silicone was poured into pre-designed molds and subsequently cured under controlled environmental conditions, forming a continuous layer that was placed over the printed structures.

Table 1 General properties of the materials used
Full size table

Equipment Used: The samples were manufactured using 3D fused deposition modeling (FDM) printing with a Bambu Lab A1 mini printer equipped with a 0.4 mm diameter nozzle. This printer is characterized by its high precision and repeatability, which allowed for consistent manufacturing conditions during mannequin production. The printer features an automatic calibration system and precise temperature control, which is essential for processing materials with different thermal behaviors, such as TPU and ABS.

To achieve future standardization in mannequin development, a well-established production workflow is necessary. The basic steps that can be applied to the workflow for developing mannequins using 3D printing are shown in Fig. 1.

Fig. 1
Fig. 1
Full size image

Workflow diagram for developing mannequins using 3D printing

The conceptualization phase of the heterogeneous anthropomorphic mannequin was developed as a systematic process aimed at establishing the functional, geometric, and structural foundations of the design, prior to its digital modeling and fabrication using additive manufacturing with FDM modeling. Initially, the mannequin’s design was planned. It was determined that the mannequin would be used for both research and educational purposes, which entailed establishing criteria that would allow for its corresponding analysis using Computed Tomography images. This stage guided the design toward a model that met the standards of realism and functionality required for use in academic and scientific environments.

Subsequently, a preliminary sketch of the mannequin’s structure was developed, defining the model’s segmentation into its main anatomical components. In this sense, its application was limited to the anatomical and radiological study of a cross-section of the human arm. The overall dimensions were determined from real biometric data, and to make the representation more accurate, four main components were defined: cortical bone, trabecular bone, muscle, and skin. An average body width of 47.68 mm (cortical bone plus trabecular bone), a muscle body width of 25.5 mm, and a skin layer thickness of 2 mm were established [22]. Based on this segmentation, a detailed analysis of materials compatible with FDM technology was carried out, evaluating their compatibility with the densitometric and visual properties of real tissues. This analysis was crucial for selecting materials that would allow not only an accurate anatomical representation but also an adequate response in tomographic imaging studies.

For the manufacture of the mannequin, the following materials were selected: silicone for skin emulation, which offers a flexible texture and surface resistance comparable to the dermis. Thermoplastic polyurethane (TPU) was used to represent muscle, as it is an elastomeric material with good deformation capacity. Acrylonitrile butadiene styrene (ABS) was selected to emulate hard tissue, both cortical and trabecular, due to its rigidity and ease of printing. This process facilitated the early evaluation of the design’s structural coherence and allowed for anticipating potential manufacturing limitations, such as print volume restrictions and part orientation. The silicone layer that emulates the skin of the heterogeneous anthropomorphic mannequin was obtained through a molding process.

Material selection was guided by previously reported applications in heterogeneous medical phantoms [23], in which silicone, TPU, and ABS have been used to represent skin, soft tissue, and rigid bone-like components, respectively. In the present proof-of-concept study, these materials were selected to provide layers with differentiated structural and radiological characteristics. No claim of direct mechanical equivalence with native tissues is made, as mechanical characterization of the fabricated components was outside the scope of the present work.

The design phase of the heterogeneous anthropomorphic mannequin was developed based on the guidelines established during the conceptualization stage, with the objective of materializing the model in a three-dimensional digital environment and ensuring its structural and manufacturability using FDM technology. This phase integrated detailed geometric modeling, the incorporation of internal structures developed from mathematical modeling [25, 26] and the preparation of the model for additive manufacturing.

This stage begins with the three-dimensional modeling of the mannequin in Meshmixer (version 3.5, Autodesk Inc., San Rafael, CA, USA, 2020), a computer-aided design (CAD) program. The external geometries of each anatomical segment were precisely defined, respecting the previously established anthropometric proportions. The model was developed in a modular fashion, dividing the system into independent components to facilitate its fabrication and assembly, as shown in Fig. 2. During this stage, appropriate dimensional tolerances were incorporated to ensure the correct fit between parts, considering the inherent limitations of the FDM process, such as material shrinkage and printing resolution.

The mathematical model of the trabecular phase was then solved to simulate internal structures similar to trabecular bone, optimizing the material’s strength-to-density ratio. This approach allowed for the integration of biomimetic design criteria with the capabilities of additive manufacturing, favoring the production of lightweight, strong, and reproducible structures.

Subsequently, the internal structures corresponding to the trabecular phase, generated using the Komarova-Hohenberg mathematical model [25], were integrated, allowing for a heterogeneous distribution of bone mass. This structure was incorporated into the mannequin’s bone design before printing. The model was modified by adjusting according to the biometric measurements. The implementation of these internal geometries was carried out using generative design tools or specialized software, ensuring a homogeneous distribution of material in the required areas and a smooth transition between solid and porous regions.

Fig. 2
Fig. 2
Full size image

Three-dimensional modeling of the mannequin in Meshmixer, (a) side view of each simulated fabric layer, (b) top view of each simulated fabric layer, (c) representation of the body section that the mannequin is emulating

The manufacturing phase of the heterogeneous anthropomorphic mannequins was carried out using 3D fused deposition modeling (FDM) printing with a Bambu Lab A1 Mini printer. The previously designed mannequin layers were exported in STL format and processed using Bambu Studio slicing software (version 1.x, Bambu Lab, Shenzhen, China, 2024).

Two thermoplastic materials were used: acrylonitrile butadiene styrene (ABS) and thermoplastic polyurethane (TPU). The complete FDM processing parameters, including nozzle characteristics, extrusion and bed temperatures, layer height, line width, extrusion flow, wall loops, infill density and pattern, raster orientation, cooling conditions, printing speed, and build orientation, are summarized in Table 2. Automatic bed leveling was performed before each printing process.

Table 2 FDM printing parameters used for the fabrication of the ABS and TPU components of the heterogeneous anthropomorphic mannequins
Full size table

Due to the model’s geometry, support material was generated only at the edges of the parts on the build plate for better adhesion and easier removal. Once the printing process was complete, the parts were cooled to room temperature before being removed from the print bed. Finally, post-processing consisted of manual support material removal and a visual inspection of the surface and dimensional quality of the printed parts.

The validation phase of the anthropomorphic mannequin aimed to verify that the designed and manufactured model met the functional, structural, and geometric requirements established in the previous stages. This phase was carried out through a series of experimental and comparative evaluations, focused on analyzing the correspondence between the digital design, the physical model, and its performance under real-world conditions. Aspects such as stability, ease of assembly, and component compatibility were verified. This stage allowed for the identification of potential design improvements, particularly in the joints and the system’s ergonomics.

The reproducibility of the 3D printing phase was analyzed by comparing three 3D prints of the same heterogeneous anthropomorphic mannequin, evaluating the consistency in dimensions and the mannequin’s quality. To conduct the comparison, the model underwent a clinical computed tomography (CT) scan to assess its structural behavior. The tomographic images obtained were processed using digital image processing techniques, employing Weasis software (version 4.x, Weasis Team, France, 2024) to obtain slices from DICOM files, ImageJ (version 1.54, National Institutes of Health, Bethesda, MD, USA, 2023) for quantitative analysis, fractal dimension analysis was employed as an image-based quantitative descriptor of structural complexity. In this study, FD was used to assess the consistency of the geometric complexity reproduced among independently manufactured phantoms and OriginPro 2026 software (OriginLab Corporation, Northampton, MA, USA) was used to perform a linear mixed-effects model (LMM) analysis. This approach was used to evaluate the consistency in structural complexity among independently manufactured phantoms by assessing differences in fractal dimension while accounting for the correlation among tomographic slices obtained from the same phantom.

Additionally, control cylinders fabricated from each material used in the phantom were included in the CT acquisition to quantitatively evaluate their radiological response. Regions of interest (ROIs) were defined within each control cylinder, and the corresponding Hounsfield Unit (HU) values were obtained. These values were subsequently compared with HU ranges reported in the literature [27] for the human tissues that each material was intended to mimic. This analysis was performed to complement the morphological evaluation and provide a quantitative assessment of the radiological equivalence of the phantom materials, see Table 3.

Table 3 Comparison between the measured Hounsfield Unit (HU) values of the phantom materials and the reported HU ranges of the corresponding human tissues [27]
Full size table

3 Results

As a result of the proposed methodology, a heterogeneous anthropomorphic mannequin was obtained composed of differentiated layers that represent bone tissue (cortical and trabecular), muscle and skin. Figure 3, a-d) shows the different layers resulting from 3D printing, the bone component manufactured with ABS material due to its rigidity and similarity to the properties of cortical and trabecular bone. The muscle layer was printed using TPU and the skin was replicated with a silicone cover. Each component was printed independently using the selected materials (ABS, TPU and silicone) and subsequently assembled, achieving adequate structural integration and correspondence with the digital design.

Fig. 3
Fig. 3
Full size image

Individual components of the heterogeneous anthropomorphic mannequin 3D printed and final assembly. (a) Silicone layer representing skin, (b) TPU layer representing muscle tissue, (c) ABS structure corresponding to trabecular tissue, (d) ABS structure representing cortical tissue; (e) final assembly of all layers that make up the complete mannequin

The images obtained by computed tomography (see Fig. 4) allowed us to visualize the differentiation between the layers, evidencing density contrasts consistent with the materials used. Likewise, a homogeneous internal distribution was observed in each region, without obvious structural defects, which confirms the viability of the manufacturing process to reproduce complex geometries. The equipment used for CT image acquisition in this study was the Siemens Healthineers SOMATOM go.Top, a high-intermediate clinical tomograph designed to offer both routine and advanced procedures with high image quality and automated workflows. The study acquisition parameters are shown in Table 4.

Table 4 Acquisition parameters of the study using Computed Tomography
Full size table
Fig. 4
Fig. 4
Full size image

DICOM images obtained by computed tomography (CT) of the heterogeneous anthropomorphic mannequin. (a) Cross-section of the complete assembly, showing the layers representing skin, muscle, cortical tissue and trabecular tissue; (b) visualization of the layer corresponding to muscle tissue; (c) section corresponding to cortical tissue; (d) internal trabecular structure

3.1 Analysis of the contrast of density of CT data

Morphological characterization was performed by calculating the fractal dimension (Fd) in consecutive tomographic slices of three mannequins manufactured under the same conditions. The mannequins underwent a uniform computed tomography (CT) protocol. Twenty-three consecutive tomographic slices were obtained from each mannequin, on which the region of interest (muscle, cortical bone, trabecular bone) corresponding to the simulated structure was delimited. In each slice, the fractal dimension was calculated using the box-counting method, obtaining a set of values that describe the morphological complexity as a function of the depth of the analyzed volume (see Table 5).

Table 5 Fractal dimension values obtained for each section of the three mannequins
Full size table

Fractal dimension values were calculated for each group of samples, with variations observed between the three types of tissue. The results showed that trabecular tissue presents consistently higher Fd values (≈ 1.84–1.89) compared to cortical (≈ 1.68–1.73) and muscle tissue (≈ 1.68–1.72), which is indicative of greater structural complexity.

The comparison of dimension fractal (FD) values among phantoms was used to identify variations in the geometric complexity of the reproduced structures. Lower dispersion among FD measurements indicates greater consistency in this morphological descriptor, whereas statistically significant differences indicate detectable variations in structural complexity among independently manufactured phantoms.

3.2 Statistical analysis: Linear Mixed-effects Model (LMM)

The FD was analyzed using a LMM to account for the hierarchical structure of the data. For each structural type—trabecular, cortical, and muscle—three phantoms were independently fabricated under the same conditions. A total of 23 consecutive CT slices were analyzed from each phantom, yielding 207 observations.

Because multiple slices were obtained from the same physical phantom, these measurements were considered clustered within each replica rather than statistically independent observations. Accordingly, structure type was included in the model as a fixed effect, whereas phantom identity was incorporated as a random intercept to account for between-replica variability and the correlation among observations obtained from the same phantom.

The model was expressed as [28]:

$$\:{FD}_{ijk}=\mu\:+{\alpha\:}_{i}+{b}_{ij}+{\epsilon\:}_{ijk},$$

where \(\:{FD}_{ijk}\) represents the fractal dimension of slice \(\:k\) from phantom \(\:j\) belonging to structure \(\:i\); \(\:\mu\:\) denotes the overall mean; \(\:{\alpha\:}_{i}\) represents the fixed effect of structure type; \(\:{b}_{ij}\) corresponds to the random effect associated with each physical phantom; and \(\:{\epsilon\:}_{ijk}\) represents the residual variability among slices within each phantom.

Variance components corresponding to between-phantom variability and residual within-phantom variability were estimated, together with the intraclass correlation coefficient (ICC). Model adequacy was assessed through graphical residual diagnostics, including residual versus fitted values, a histogram of the residuals, and a normal probability plot. Statistical significance was established at an adjusted significance level of \(\:p<0.05\).

Regarding between-replica variability, the estimated variance between phantoms was \(\:2.689\times\:{10}^{-5}\), corresponding to a standard deviation of 0.00519. The residual variance was \(\:8.427\times\:{10}^{-5}\), with a standard deviation of 0.00918. Therefore, the variability attributable to differences among the physical phantoms was lower than the variability observed among CT slices within the same phantom.

The intraclass correlation coefficient (ICC) was 0.2419, while the root mean square error (Root-MSE) was 0.00908.

The trabecular structure exhibited significantly higher FDijk values than the cortical structure, with an estimated difference of 0.15448 (IC 95%: 0.14558–0.16338; \(\:p<0.001\)), and the muscle structure, with an estimated difference of 0.14994 (IC 95%: 0.14104–0.15884; \(\:p<0.001\)). In contrast, the estimated difference between the muscle and cortical structures was 0.00454 (IC 95%: −0.00436–0.01344; \(\:p=0.3155\)), indicating no statistically significant difference between these two structures.

As shown in Fig. 5, the distribution of fractal dimension values exhibited a distinct pattern among the structures analyzed. The three replicas of the trabecular structure showed higher FDijk values than those observed for the cortical and muscle structures, whereas the latter two displayed considerably more similar distributions. Within each structure type, the three physical replicas showed partially overlapping distributions and relatively limited dispersion compared with the differences observed between the trabecular structure and the other two structures. This descriptive pattern was consistent with the results obtained from the linear mixed-effects model. Overall, Fig. 5 illustrates the separation among structure types and the relative consistency of the replicas fabricated under the same conditions.

Fig. 5
Fig. 5
Full size image

Distribution of fractal dimension (FD) according to structure type and phantom replica. For each structure—cortical, muscle, and trabecular—three independently fabricated phantoms (Phantom1–Phantom3) were analyzed, with 23 consecutive CT slices per replica. The boxes represent the interquartile range (IQR), the central line indicates the median, and the points represent the individual DF values obtained from each slice

The inspection of the diagnostic plots showed that the residuals were predominantly distributed around zero. The central region of the normal probability plot showed an approximately linear pattern, although deviations were observed in the tails, together with some heterogeneity in the residual dispersion across the fitted values (Fig. 6).

Fig. 6
Fig. 6
Full size image

Residual diagnostic plots of the linear mixed-effects model used for the analysis of fractal dimension. (a) Residuals versus observation order; (b) frequency distribution of the residuals; (c) residuals versus model-fitted values; and (d) normal probability plot of the residuals

4 Discussion

The results obtained demonstrate that additive manufacturing using FDM is a viable strategy for fabricating heterogeneous anthropomorphic mannequins with adequate structural differentiation between simulated tissues. The clear layer differentiation observed in the computed tomography (CT) images shows that the material selection (ABS, TPU, and silicone) was appropriate for simulating different tissue types, allowing the reproduction of CT-derived density contrasts that are essential in medical imaging applications.

Fractal dimension provides a quantitative description of structural complexity and therefore offers an image-based approach for comparing the morphology reproduced among independently manufactured phantoms. Similar FD values indicate that the fabrication process produced structures with comparable levels of geometric complexity; however, this should not be interpreted as evidence of complete manufacturing reproducibility. FD does not directly quantify dimensional accuracy, mechanical repeatability, or material homogeneity. Consequently, the present results support the reproducibility of specific morphological characteristics observed by CT, whereas a comprehensive evaluation of manufacturing reproducibility would require additional dimensional and mechanical characterization.

In trabecular structures fabricated from ABS, the presence of significant differences among phantoms suggests that the resulting geometric complexity may be sensitive to variability inherent to the FDM manufacturing process. Although parameters such as infill pattern, density, and deposition orientation may affect the architecture of printed structures, these variables were not systematically varied in the present study. Therefore, a direct relationship between individual printing parameters and fractal dimension cannot be established.

The behavior observed in the muscle system (TPU), characterized by a higher level of statistical significance, can be attributed to the viscoelastic nature of the material. During the deposition process, TPU exhibits higher levels of deformation and relaxation, which can amplify the geometric variations between configurations. This greater structural sensitivity is reflected in the fractal dimension, revealing a more pronounced heterogeneity. From an applied perspective, this result is particularly relevant in the development of biomimetic models of soft tissues.

The quantitative HU analysis showed that the ABS control samples exhibited values ranging from 150 to 280 HU. This range is comparable to the value reported for trabecular bone (approximately 262 HU), indicating reasonable radiological agreement for the trabecular component. However, these values were substantially lower than those reported for cortical bone (approximately 1000–1500 HU). This discrepancy can be attributed to the intrinsic radiological properties of ABS, which is a non-mineralized polymer with a lower effective density and X-ray attenuation than highly mineralized cortical bone. Furthermore, although a nominal infill density of 100% was used, the layer-by-layer FDM process may introduce small inter-road voids and interfaces that can reduce the effective radiodensity of the printed component.

Therefore, while ABS provided an adequate approximation of the HU values associated with trabecular bone, it did not fully reproduce the radiological attenuation of cortical bone. This result indicates that further optimization of the material would be necessary to achieve attenuation equivalent to that of cortical bone, possibly by incorporating mineral-based fillers into the impression material or even by changing the impression material itself.

In contrast, the TPU samples exhibited HU values of 40–50 HU, showing good agreement with the reported range for muscle tissue (40–44 HU). Silicone exhibited values between 50 and 75 HU, partially overlapping the reported range for skin tissue (70–77 HU), indicating a reasonable, although not complete, approximation of its radiological response.

Overall, the HU analysis demonstrates different degrees of radiological agreement between the selected phantom materials and their target tissues. TPU showed close agreement with muscle tissue, silicone partially reproduced the reported attenuation of skin, and ABS showed better agreement with trabecular bone than with cortical bone.

The linear mixed-effects model enabled the simultaneous assessment of morphological differences among the fabricated structures and the variability associated with the physical replicas. The trabecular structure exhibited significantly higher fractal dimension (DF) values than the cortical and muscle structures, reflecting the greater morphological complexity of its geometry. This finding is consistent with the heterogeneous and interconnected organization of the trabecular architecture compared with the relatively more homogeneous geometries of the other structures.

No statistically significant difference was observed between the cortical and muscle structures. This suggests that, although DF clearly distinguishes the trabecular structure, this parameter alone may not be sufficient to differentiate all tissue-equivalent structures or materials. Accordingly, DF should be interpreted primarily as a descriptor of morphological complexity rather than as an independent parameter for tissue classification.

The variance components further showed that variability among independently fabricated phantoms was lower than the residual variability observed among CT slices within the same phantom. The standard deviation associated with between-phantom variability was 0.00519, whereas the residual standard deviation was 0.00918. These results suggest that the variability introduced by independently fabricating the same structure was relatively limited compared with the spatial variation observed across CT slices, supporting the consistency of the fabrication procedure under the conditions evaluated.

The mixed-effects approach also accounted for the hierarchical structure of the experimental data. Nevertheless, only three physical replicas were analyzed for each structure type; therefore, estimates of between-phantom variability should be interpreted with caution. Increasing the number of independently fabricated replicas would provide a more robust characterization of the repeatability of the manufacturing process.

It is important to consider some inherent limitations of the FDM process, such as material anisotropy, print resolution, and potential changes in material-CT radiation sensitivity that can affect the final microstructure. Furthermore, although fractal dimension is a robust descriptor of geometric complexity, it does not allow for the direct evaluation of mechanical or radiological properties, thus requiring further studies.

The contribution of the proposed approach does not rely on the development of a new FDM process, since the ABS and TPU components were fabricated separately using conventional single-material printing and subsequently assembled with a molded silicone layer. Instead, the methodological contribution lies in the integration of materials with different structural and radiological characteristics within a single anthropomorphic phantom. The incorporation of a mathematically generated trabecular architecture, anatomically based geometry, and CT-based characterization, including HU measurements, provides a reproducible framework for the fabrication of customizable heterogeneous phantoms using accessible manufacturing techniques.

Taken together, these results are consistent with recent literature on the use of 3D printing in the fabrication of medical phantoms and represent an advance by incorporating anthropomorphic and trabecular structures generated using mathematical models, which are quantitatively evaluated in terms of reproducibility and morphological similarity.

5 Conclusions

This work demonstrates the feasibility of designing and manufacturing a heterogeneous anthropomorphic mannequin using FDM technology, integrating multiple materials to simulate different types of biological tissue. The developed methodology yielded models with adequate structural differentiation, both macroscopically and internally, validated through computed tomography studies.

The use of a mathematical model to generate the trabecular structure proved effective in reproducing the morphological complexity of bone. The linear mixed-effects model allowed the characterization of both morphological differences among the fabricated structures and the variability among their physical replicas. Likewise, the variability among independently fabricated phantoms was lower than the variability observed among CT slices within the same phantom. Overall, these results support the consistency and repeatability of the fabrication process under the conditions evaluated.

The results showed consistency among the replicas fabricated under the evaluated conditions; however, because the study included only three independent phantoms per structure, conclusions regarding reproducibility should be interpreted within this experimental scope. Future studies, including a larger number of replicas and independent fabrication runs will be needed to confirm and generalize these findings.

Although ABS was selected for both cortical and trabecular components because of its rigidity and previous use in bone-mimicking phantoms, the differentiation between these tissues in the present model was mainly achieved through geometric design. The trabecular component incorporated a porous mathematically generated architecture, whereas the cortical component was represented as a denser structure. Nevertheless, the HU analysis showed that the radiological response of ABS was closer to that reported for trabecular bone and did not reach the higher attenuation values characteristic of cortical bone. Therefore, while ABS was suitable for reproducing the structural architecture of both regions in this proof-of-concept model, further material modification is required to improve the radiological fidelity of the cortical component.

Finally, this work establishes a methodological basis for the additive manufacturing of low-cost, customized heterogeneous anthropomorphic mannequins, with the potential to be adapted to different anatomical regions and specific applications. As future work, it is suggested to complement the characterization with more detailed mechanical and radiological analyses, as well as to explore new materials and manufacturing techniques that allow for improved resolution and fidelity of the model.