Clinical relevance of accuracy in in-house 3D printing in craniomaxillofacial surgery: A comparative study of FFF, SLA, and MJ Technologies

Clinical relevance of accuracy in in-house 3D printing in craniomaxillofacial surgery: A comparative study of FFF, SLA, and MJ Technologies | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Clinical relevance of accuracy in in-house 3D printing in craniomaxillofacial surgery: A comparative study of FFF, SLA, and MJ Technologies Lievens Mauranne, Van Paepegem Wim, Tom Goffin, Villeirs Geert, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7633361/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 4 You are reading this latest preprint version Abstract Background Considering the Medical Device Regulation (MDR 2017/745), this study aims to evaluate the accuracy of three commonly used 3D printing technologies—Fused Filament Fabrication (FFF), Stereolithography (SLA), and Material Jetting (MJ)—within a clinically realistic in-house 3D printing workflow. Methods In this study, the accuracy of 3D printers was assessed using highly accurate anatomical models to mimic an in vitro clinical workflow, unlike previous validation studies that relied on geometric calibration models or simplified shapes. Accuracy was defined through precision (intra- and inter-build variability) and trueness (deviation from the digital reference model), quantified using root mean square (RMS) error. To analyse the data structure—comprising repeated prints, multiple models, and timepoints—we used a Linear Mixed Model (LMM), which enabled evaluation of fixed effects (printer type, anatomical model, and comparison type) while correcting for internal clustering of data. Results All printers achieved clinically acceptable accuracy. MJ was significantly more accurate (RMS 67 µm) compared to SLA (109 µm) and FFF (130 µm). The Linear Mixed Model showed that accuracy was influenced by printer type and anatomical complexity. Conclusion This is the first study in 3D printing accuracy research to apply a Linear Mixed Model, providing a statistically robust analysis framework. Among the evaluated technologies, MJ achieved the highest accuracy in clinically realistic conditions, supporting its use in applications that require high anatomical fidelity within the MDR framework. 3D printing anatomical models medical device regulation (MDR 2017/745) accuracy validation in-house manufacturing linear mixed model point-of-care manufacturing Figures Figure 1 Figure 2 Figure 3 Background 3D printing, also known as additive manufacturing or rapid prototyping, is a process that involves constructing a three-dimensional solid object of virtually any shape by adding material one layer at a time based on a digital model. Common 3D printing technologies include fused filament fabrication (FFF), stereolithography (SLA), selective laser sintering (SLS), material jetting (MJ) and binder jetting (BJ) ( 1 , 2 ). In the last decade, the increasing availability of low-cost 3D printers, combined with advancements in CAD/CAM technology and medical imaging, has led to an increase in 3D printing applications within the medical field. Initially, 3D printing was used primarily for treatment planning and education. More recently, it has expanded to patient-specific applications such as surgical guides, customized implants, and models for complex anatomy. These applications have demonstrated significant clinical benefits, including improved comprehension of complex anatomy, reduction of operating time, and more precise preoperative planning ( 1 – 8 ). Among the various 3D printing technologies, FFF and SLA are the most accessible and widely used in hospital settings due to their affordability, ease of use, and the availability of compatible materials. FFF, which extrudes thermoplastic filaments layer by layer, is popular for producing anatomical models for surgical planning, patient education, and medical device prototyping. It is cost-effective, but it has limitations in resolution and surface finish​. SLA, a vat photopolymerization method using liquid resin cured by ultraviolet (UV) light, offers higher resolution and smoother surface finishes, making it ideal for printing surgical guides and dental models​. However, SLA-printed parts require post-processing to remove uncured resin and enhance mechanical properties. In contrast, MJ is primarily used for industrial applications such as automotive engineering, aerospace, and high-precision product manufacturing. This method enables high-resolution multi-material and full-colour printing, which is beneficial in complex prototyping and high-detail modelling​. Despite its advantages, MJ is less common in hospital settings due to its high equipment and material costs, as well as the complexity of operating and maintaining the system​ ( 2 , 9 ). To address the cost and logistical challenges of outsourcing 3D printing services, many hospitals are now developing in-house 3D printing laboratories, referred to as point-of-care (POC) manufacturing. This approach has opened new possibilities, such as bioprinting and the fabrication of custom implants ( 3 , 7 , 8 ). However, with these advancements come additional regulatory considerations. Under the European Union's Medical Device Regulation (MDR) 2017/745, which governs in-house production of medical devices, including 3D printed models, hospitals must ensure compliance with strict safety, quality, and performance standards for patient-specific devices. MDR places stringent requirements on POC manufacturing, particularly when creating patient-specific medical devices, to ensure the reliability and safety of such models in clinical settings ( 10 ). Accuracy (including trueness and precision) in the 3D printing process is critical for safe intraoperative use of these models. However, errors can occur at various stages of the manufacturing process, such as data acquisition, image processing, and model fabrication. These steps are influenced by the equipment used, the software, and the expertise of the operator. Additionally, the segmentation process from Digital Imaging and Communications in Medicine (DICOM) to Standard Tessellation Language (STL), particularly when using Cone Beam Computed Tomography (CBCT) or Multi-slice Computed Tomography (MSCT), is highly user-dependent and may introduce errors such as inaccurate volumes or incorrect geometries. To better assess the precision and trueness of 3D printers, this study eliminates the segmentation step by using a high-precision optical 3D scanner ( 6 , 9 – 11 ). Although manufacturers provide specifications on the accuracy of their 3D printers, these tests often involve simple symmetrical objects that do not reflect the complexity of anatomical models, leading to potential inaccuracies such as incorrect volumes or deformations in clinical practice ( 5 , 7 , 9 , 11 , 12 ). The aim of this study is to evaluate the accuracy of 3D printing in a clinically relevant setting using highly accurate anatomical models. This investigation focuses on three 3D printing technologies: FFF, SLA, and MJ. By doing so, we aim to provide a deeper understanding of the accuracy of 3D printed models for medical applications, particularly in the context of stringent quality control requirements under MDR regulations. Methods Study Design This study was designed to evaluate the accuracy of three 3D printing technologies in reproducing clinically relevant anatomical models. Accuracy was assessed in terms of precision and trueness, following the metrological definitions described in ISO 5725-1( 13 ). Precision refers to how consistent repeated measurements are under the same conditions, reflecting the variability or spread (i.e., low variance). Trueness describes how closely a measurement matches the reference value, reflecting systematic error (i.e., low bias). In this study, trueness indicates how closely a printed model matches its original STL file, while precision reflects how similar the printed models are to each other. To evaluate precision and trueness, two experimental setups were used, as illustrated in Fig. 1 . Precision was assessed by examining both intra-build variability (repeatability) and inter-build variability (reproducibility). Intra-build variability was measured by printing multiple copies of the same anatomical model simultaneously in a single build under identical conditions. Inter-build variability was assessed by printing the same models on different days using the same printer and settings. All models used for precision assessment were subsequently also compared to their corresponding original STL files to evaluate trueness. Figure 1 . Methodological study design: Visual overview of the experimental workflow used to assess the accuracy of three 3D printing technologies (FFF, SLA, MJ) using clinically relevant anatomical models. Trueness was evaluated by comparing each printed model to its original STL reference (green arrow). Precision was assessed through intra-build variability and inter-build variability (red arrow), by comparing models printed simultaneously or across different builds. Anatomical models and 3D printing workflow Four different anatomical models were selected from a high-quality anatomically accurate skull model (SOMSO-Plast® Bauchene QS9/1, Coburg, Germany ( 14 )): the sphenoid bone, frontal bone, zygomatic bone, and mandible. These models were chosen for their variability in size, geometry, and complexity. Each model was scanned using a MSCT with a high-resolution object scanning protocol. The scan settings were: slice thickness 0.6 mm, increment 0.4 mm, pitch 0.8, rotation time 1.0 s, effective mAs 420, with CARE Dose and CARE kV turned off. DICOM (Digital Imaging and Communications in Medicine) data from the MSCT was processed using Materialise Innovation Suite® (MIS, Leuven, Belgium). Segmentation was performed using a fixed threshold of -200 Hounsfield units (HU) across all models. The resulting voxel-based segmentations were converted into STL files for 3D printing. Three 3D printing technologies were used. The Felix Pro 3® printer was used for Fused Filament Fabrication (FFF), with a build volume of 237 × 244 × 235 mm. A 0.4 mm nozzle and a layer height of 200 microns were used, printing with PLA filament ( 15 , 16 ). The Formlabs Form 3B+® printer was used for Stereolithography (SLA), with a build volume of 145 × 145 × 165 mm. It printed with Formlabs Model Resin at a 100 micron layer thickness ( 17 , 18 ). The The Stratasys J5 MediJet® printer was used for Material Jetting (MJ), offering a build volume of 140 × 200 × 190 mm and printing with MED620 material at a 27 micron layer thickness ( 19 , 20 ). All printers were properly calibrated prior to use. All prints were produced under identical conditions (temperature, relative humidity, conditioned room) using raw materials of equal age. Printer settings were based on the respective manufacturers' recommendations and were kept constant across builds to avoid introducing variability through slicing or process parameters. After 3D printing, post-processing of the parts was performed. For FFF the supports were removed manually. The SLA printed parts first underwent a washing process of 10 minutes in ≥ 99% isopropyl alcohol (IPA), to remove any excess resin on the surface of the parts, and 5 minutes of post-curing at 60°C to achieve their optimal mechanical properties. Afterwards, the supports were removed manually. For MJ printed parts, the gel coat surrounding the parts is removed by submerging the parts in water followed using a water jet. Precision assessment For intra-build precision, multiple identical copies of each anatomical model were printed on the same build platform. The number of replicas per build was determined by the available build volume of each printer. For example, the MJ printer consistently produced four copies per model, while FFF and SLA produced two to three copies depending on the model size. For inter-build precision, one copy of each model was printed on three separate occasions over the course of one week, using identical settings. The consistency of each group of prints was evaluated by calculating the deviations between all possible pairwise combinations of the models within the same group (intra-build or inter-build). These deviations quantify how similar the models are to each other and reflect the precision of the printing process. Trueness assessment To evaluate trueness, each printed model was compared to the original segmented STL file. All models printed for intra-build and inter-build analysis were included in the trueness analysis. Thus, trueness was assessed across a range of build conditions and printers, providing a comprehensive picture of how faithfully each technology reproduces the reference geometry. Optical scan of printed models Each 3D printed model was scanned using an optical white light desktop 3D scanner (EinScan-SP, SHINING 3D Tech. Co., Ltd.). The scanner uses white light and coherence scanning interferometry (CSI) to perform non-contact surface digitization, with accuracy better than 50 microns ( 12 , 21 ). Each model was scanned in four orientations (cranial, caudal, left lateral, right lateral), and the scans were merged into a single non-watertight STL model. No post-processing (such as smoothing or hole filling) was applied. The final STL model was used for dimensional comparison with either the original STL file (for trueness) or other scans within the same group (for precision)( 12 ). Part comparison analysis and accuracy metric The comparison analysis was performed in Materialise Mimics 18.0 (Materialise NV, Leuven, Belgium). STL files of both the 3D printed models and the original model were imported. For each analysis, the pair of STL models were aligned using a two-step process. An initial alignment was performed using six manually placed landmarks (N-point registration), followed by fine-tuning via global surface registration. To quantify accuracy, a part comparison analysis was conducted subsequently, which is based on point-cloud comparison. This analysis measures deviations and calculates the Root Mean Square (RMS) deviation, which is the square root of the mean of the squared distances between corresponding surface points of the two aligned STL models. RMS reflects the overall magnitude of deviation, regardless of direction, making it more robust than the mean deviation (MD), which can be affected by the cancellation of positive and negative errors. Unlike MD, RMS accounts for both the distribution and magnitude of deviations and is therefore a more appropriate measure of overall dimensional error. During this part comparison analysis, a heat map was generated to visually highlight areas of dimensional deviation (Fig. 2 ). Figure 2 : Demonstration of the part comparison analysis conducted in Materialise Mimics 18.0 (Materialise NV, Leuven, Belgium) for two 3D printed mandible models. First, the two models were aligned (N-point registration and global registration), then a point cloud part comparison analysis is carried out. A heat map shows the areas of aberrations. RMS values were used as the primary outcome in the statistical analysis to evaluate the effects of printer type, anatomical model, and comparison type (intra-build, inter-build, trueness) on the accuracy of 3D printed models. Statistical analysis To address the complex and partially dependent structure of the data, a Linear Mixed Model (LMM) was employed (Fig. 3 ). LMMs are particularly well suited for datasets with repeated measures, hierarchical nesting, or clustering, as they allow the modelling of both fixed effects (systematic factors like printer type) and random effects (uncontrolled variability across clusters). This approach allows for more accurate estimation of fixed effects while controlling for within-group correlation. Because LMMs assume normally distributed residuals, the RMS values in this study were log-transformed to better meet this assumption and improve model fit. The transformed RMS values were then used as the dependent variable in the LMM. Fixed effects included anatomical model, printer, and comparison type (intra-build, inter-build, and trueness). To correct for internal clustering of measurements under shared experimental conditions, a random intercept was added for each unique combination of anatomical model, printer, and comparison type. Figue 3. Illustration of the linear mixed model used to analyze log-transformed RMS values (LogRMS). The model included fixed effects for anatomical model, printer type, and comparison type (shown in green). A random intercept was added for each unique combination of these three fixed factors (illustration of the clustering shown in red) to account for the hierarchical structure and repeated conditions in the experimental design. This modeling approach allows appropriate estimation of fixed effects while controlling for internal variance within clustered measurements. Results The accuracy of the three printing technologies was assessed using a Linear Mixed Model (LMM), which evaluated the effects of printer type, anatomical model, and comparison type on RMS error, while adding a random intercept to account for data clustering within printer–model–comparison combinations (Table 1 – 2 ). Table 1 Linear Mixed Model Analysis Results (RMS accuracy) Effect Group Mean (LogRMS) 95% CI Back-transformed RMS (µm) F df p-value Fixed Effects Printer 15.646 2 < 0.001 Felix (FFF) –2.041 [–2.222, − 1.861] 130 Forms3 (SLA) –2.221 [–2.402, − 2.040] 109 Stratasys (MJ) –2.695 [–2.874, − 2.517] 67 Anatomical model 11.764 3 < 0.001 Frontal bone –2.045 [–2.250, − 1.839] 129 Mandible –2.127 [–2.333, − 1.920] 119 Sphenoid bone –2.305 [–2.509, − 2.100] 100 Zygomatic bone –2.801 [–3.005, − 2.596] 61 Comparison type 1.206 2 0.305 Inter-build –2.324 [–2.487, − 2.160] 98 Intra-build –2.424 [–2.692, − 2.156] 89 Trueness –2.211 [–2.350, − 2.072] 110 Table 1 : Estimated marginal means from the linear mixed model with fixed effects for printer type, anatomical model, and comparison type. RMS values are log-transformed; back-transformed RMS values are provided in micrometers (µm). Significance was found for printer and anatomical model (p < .001), but not for comparison type (p = .305). The estimated variance of the random intercept was 0.197, compared to a residual variance of 0.045 (Table 2 ). This implies that approximately 81.4% of the total variance in LogRMS was attributable to clustering within printer-model-comparison combinations, with only 18.6% attributable to residual error. Table 2 Random effects (intercept clustering) Effect Variance Estimate Standard Error Intercept (ClusterID) 0.197 0.037 Residual 0.045 0.009 Table 2 : Random effects from the linear mixed model, showing variance estimates and standard errors for intercept clustering and residual error. In the following sections, mean refers to the estimated marginal mean of the log-transformed RMS values, with corresponding 95% confidence intervals (CI) and back-transformed RMS values (in µm) presented for interpretability (Table 1 ). Printer Effect Printer type had a significant effect on LogRMS ( p < 0.001). The Stratasys printer had the lowest LogRMS (mean = − 2.695, 95% CI [–2.874, − 2.517]) corresponding to a back-transformed RMS of 67 µm. This was followed by the Forms3 (mean = − 2.221, 95% CI [–2.402, − 2.040]; 109 µm) and the Felix printer (mean = − 2.041, 95% CI [–2.222, − 1.861]; 130 µm). Post hoc comparisons confirmed that Stratasys differed significantly from both other printers ( p < 0.001). Anatomical Model Effect Anatomical model also had a significant effect on LogRMS ( p < 0 .001). The Zygoma model showed the lowest LogRMS (mean = − 2.801, 95% CI [–3.005, − 2.596]; 61 µm), followed by Sphenoid model (mean = − 2.305, 95% CI [–2.509, − 2.100]; 100 µm), Mandible (mean = − 2.127, 95% CI [–2.333, − 1.920]; 119 µm), and Frontal bone (mean = − 2.045, 95% CI [–2.250, − 1.839]; 129 µm). Post hoc testing showed that Zygoma model differed significantly from all other models ( p < .01), while other pairwise differences were less consistent. Part Comparison Type Effect Comparison type did not have a significant effect on LogRMS ( p = 0.305). The intra-build group had a mean of − 2.424 (95% CI [–2.692, − 2.156]), corresponding to a back-transformed RMS of 89 µm. The inter-build group had mean of − 2.324 (95% CI [–2.487, − 2.160]; 98 µm), and the trueness group had a mean of − 2.211 (95% CI [–2.350, − 2.072]; 110 µm). Post hoc comparisons revealed no significant differences between the three comparison types. Discussion This study aimed to evaluate the dimensional accuracy of 3D printed anatomical models using three 3D printing technologies—Fused Filament Fabrication (FFF), Stereolithography (SLA), and Material Jetting (MJ)—within the framework of in-house clinical use under MDR regulations. Accuracy was approached as a combination of trueness (closeness to a reference) and precision (reproducibility). The experimental design included multiple builds and comparison types (intra-build, inter-build, and trueness comparisons), providing a clinically relevant validation strategy. Although the study design was structured to separately assess trueness and precision through various comparison types (intra-build, inter-build, trueness), all measurements were analysed together in a single, unified Linear Mixed Model (LMM). This approach allowed for the inclusion of all relevant data points while appropriately accounting for their nested structure. By modelling printer type, anatomical model, and comparison type as fixed effects, and incorporating a random intercept for each printer–model–comparison combination, the LMM controlled for shared experimental conditions and internal clustering. This statistical framework offered a robust and scalable way to capture both systematic influences and internal variability, making it well-suited for evaluating accuracy in complex clinical 3D printing workflows. The LMM revealed significant effects of both printer type and anatomical model on accuracy, while the comparison type (intra-build, inter-build, trueness) had no significant impact. Notably, 81.4% of the total variance in LogRMS was attributed to the random intercept, confirming strong clustering and validating the need for mixed modeling. Printer type had a clear and significant influence on accuracy. After adjusting for anatomical model and internal clustering, the MJ printer (Stratasys) showed the highest accuracy with an estimated RMS of 67 µm, significantly outperforming both SLA (Forms3, 109 µm) and FFF (Felix, 130 µm). These results confirm that printer technology remains a dominant factor in achieving accurate prints and suggest that MJ is particularly suitable for applications requiring high dimensional fidelity, such as surgical guides. The anatomical model also significantly influenced the results. The Zygoma model consistently showed the lowest deviations, likely due to its smaller size and geometric simplicity compared to more complex models like the mandible or sphenoid. These findings highlight that anatomical geometry is an important confounder and should be considered in validation studies of 3D printing accuracy. Interestingly, the comparison type—whether evaluating intra-build, inter-build, or trueness—did not significantly influence the outcome when corrected for printer and model. This suggests that all three evaluation types capture similar aspects of dimensional accuracy under controlled printing conditions, and that the dominant variation arises from the printer and model rather than from how the comparison is structured. It should also be noted that the printers were evaluated using their standard manufacturer-recommended settings, which included different layer thicknesses: 200 µm for FFF, 100 µm for SLA, and 27 µm for MJ. These inherent differences in vertical resolution likely contributed to the superior accuracy observed for the MJ printer. However, this was not an intentional bias; the goal was to assess each printer as it would realistically be used in clinical environments. Evaluating devices under their typical settings enhances the validity of the results and ensures that the conclusions are aligned with real-world applications. Clinical Implications All printers performed within the clinically acceptable tolerance, in craniomaxillofacial surgery procedures, of 0.5 mm, but the superior accuracy of MJ makes it the most reliable option for applications demanding high precision. SLA may be sufficient for visual anatomical models, whereas FFF, despite its accessibility and low cost, may require additional quality control protocols before being adopted in high-risk clinical procedures. Limitations This study was limited by the number of anatomical models tested and the use of fixed manufacturer settings. It did not evaluate the effects of build orientation, printer calibration variations, or post-processing steps such as sterilization or surface finishing, all of which could impact final accuracy. Future studies should incorporate these parameters to develop a more comprehensive quality assurance framework aligned with MDR standards. Statistical methodology and future perspective Previous studies on 3D printing accuracy often relied on univariate or non-parametric tests, which do not accommodate repeated measurements or internal clustering. By using a Linear Mixed Model, this study modelled both fixed and random effects simultaneously, revealing that over 80% of the variance was attributable to shared printing conditions—something traditional methods fail to capture. Only one study in the field of dental model printing (Brown et al., 2018) applied a generalized linear mixed model to assess print accuracy. However, to the best of our knowledge, no previous studies have applied a Linear Mixed Model (LMM) in the context of clinically realistic anatomical 3D printing workflows. The present study is therefore the first to implement a LMM framework to evaluate the accuracy of in-house 3D printing, correcting for clustering effects due to repeated prints, anatomical model variation, and comparison type. This approach enabled a more robust analysis of fixed effects (e.g., printer type) while reducing the risk of type I and type II errors common in simpler statistical methods. As such, it offers a statistically rigorous standard that may serve as a reference for future validation studies under MDR compliance. Conclusion This study evaluated the dimensional accuracy of FFF, SLA, and MJ 3D printing technologies using anatomical models in a clinically relevant context. All printers performed within the commonly accepted tolerance of 0.5 mm, supporting their use in clinical workflows under MDR 2017/745. However, MJ technology demonstrated significantly higher accuracy, making it the most suitable option for applications demanding high dimensional fidelity. SLA and FFF remain cost-effective alternatives where such precision is less critical. To maintain safety and compliance in in-house 3D printing, regular accuracy validation—addressing both trueness and precision, including intra- and inter-build variability—should be integrated into ongoing quality control. Our application of LMM provided a clinically realistic and statistically appropriate framework to analyse accuracy across multiple dimensions. Given its ability to account for nested data and repeated measures, this method should be considered a recommended standard for future 3D printing validation studies. Future research should also explore the clinical impact of 3D-printed models on procedural outcomes, assess long-term dimensional stability of materials, and investigate how post-processing steps (e.g., sterilization, surface finishing) affect final model accuracy. Declarations Ethics approval and consent to participate Since this is an in vitro study, no ethics approval is required, but due to its clinical importance, this study was submitted to the Ghent University Hospital Medical Ethics Committee. Official approval was obtained on July 2, 2024. Consent for publication Not applicable. This study does not contain any individual person’s data in any form. Competing interests The authors report there are no competing interests to declare. Funding This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. Author Contribution M.L. was responsible for the study design, data collection, analysis, interpretation of results, and writing of the manuscript. R.C. was the main responsible for the study design. Together with G.V., T.G., and W.V.P., he provided guidance, critical feedback, and supervision throughout the research process. All authors reviewed and approved the final manuscript. Acknowledgements The author would like to thank Prof. Dr. Renaat Coopman and co-authors for their guidance and critical feedback throughout the research process. Their support was instrumental in the development of this study. Data Availability Data is provided within the manuscript or supplementary information files. References Aimar A, Palermo A, Innocenti B. The Role of 3D Printing in Medical Applications: A State of the Art. J Healthc Eng. 2019;2019. Diment LE, Thompson MS, Bergmann JHM. Clinical efficacy and effectiveness of 3D printing: A systematic review. BMJ Open 1 Dec 2017;7(12). Louvrier A, Marty P, Barrabé A, Euvrard E, Chatelain B, Weber E. e.a. How useful is 3D printing in maxillofacial surgery? J Stomatol Oral Maxillofac Surg. 2017;118(4):206 – 12. King BJ, Park EP, Christensen BJ, Danrad R. On-Site 3-Dimensional Printing and Preoperative Adaptation Decrease Operative Time for Mandibular Fracture Repair. J Oral Maxillofac Surg. 2018;76(9):1950.e1-1950.e8. Weijs WLJ, Coppen C, Schreurs R, Vreeken RD, Verhulst AC, Merkx MAW. e.a. Accuracy of virtually 3D planned resection templates in mandibular reconstruction. J Cranio-Maxillofac Surg. 2016;44(11):1828-32. George E, Liacouras P, Rybicki FJ, Mitsouras D. Measuring and establishing the accuracy and reproducibility of 3D printed medical models. Radiographics. 2017;37(5):1424–50. Calvo-Haro JA, Pascau J, Asencio-Pascual JM, Calvo-Manuel F, Cancho-Gil MJ. Del Cañizo López JF, e.a. Point-of-care manufacturing: a single university hospital’s initial experience. 3D Print Med. december 2021;7(1). Willemsen K, Magré J, Mol J, Noordmans HJ, Weinans H, Hekman EEG. e.a. Vital Role of In-House 3D Lab to Create Unprecedented Solutions for Challenges in Spinal Surgery, Practical Guidelines and Clinical Case Series. J Pers Med. 1 maart 2022;12(3). Msallem B, Neha S, Shuaishuai C, Halbeisen FS. Hans-Florian Zeilhofer FMThieringer. Evaluation of the Dimensional Accuracy of 3D-Printed Anatomical Mandibular Models Using. J Clin Med. 2020;9(3):1–18. Huotilainen E, Jaanimets R, Valášek J, Marcián P, Salmi M, Tuomi J. e.a. Inaccuracies in additive manufactured medical skull models caused by the DICOM to STL conversion process. J Cranio-Maxillofac Surg. 2014;42(5). Dorweiler B, Baqué PE, Chaban R, Ghazy A, Salem O. Quality control in 3D printing: Accuracy analysis of 3D-printed models of patient-specific anatomy. Materials. 2021;14(4):1–13. Lievens M, De Kock L, Ureel M, Villeirs G, Van Paepegem W, Coopman R. The Accuracy of an Optical White Light Desktop 3D Scanner and Cone Beam CT Scanner Compared to a Multi-Slice CT Scanner to Digitize Anatomical 3D Models: A Pilot Study. Craniomaxillofacial Trauma Reconstr juni. 2025;18(2):27. ISO 5725-1. 2023(en), Accuracy (trueness and precision) of measurement methods and results — Part 1: General principles and definitions [Internet]. [geciteerd 17 augustus 2025]. Beschikbaar op: https://www.iso.org/obp/ui/#iso:std:iso:5725:-1:ed-2:v1:en SOMSO. QS 9/1, Bauchene Skull. Felix printers. Quick Start Guide FELIX Pro 3. 2019. Felix printers. Filaments - PLA TDS. Formlabs. september. Form 3B+ - Low Force Stereolithography (LFS) 3D Printer. 2023. Formlabs. Dental resin model. 11 september 2021. Stratasys. Stratasys J5 MediJet. Strat J5 MediJet. 2022. Stratasys. Dentajet Dental Materials. Shining 3D, Einscan. SE/SP2 Dekstop 3D Scanner. 10 oktober 2024. Additional Declarations No competing interests reported. Supplementary Files LinearMixedModelArtikel2SPSS.xlsx Cite Share Download PDF Status: Under Review Version 1 posted Reviewers invited by journal 06 Oct, 2025 Editor assigned by journal 16 Sep, 2025 Submission checks completed at journal 16 Sep, 2025 First submitted to journal 16 Sep, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7633361","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":530669422,"identity":"dc1a34ff-4a7d-4a0d-8439-18470605bfd5","order_by":0,"name":"Lievens Mauranne","email":"","orcid":"","institution":"University Hospital Ghent","correspondingAuthor":false,"prefix":"","firstName":"Lievens","middleName":"","lastName":"Mauranne","suffix":""},{"id":530669423,"identity":"dcebed98-3312-4d9f-807c-3403d0700da8","order_by":1,"name":"Van Paepegem Wim","email":"","orcid":"","institution":"Ghent University","correspondingAuthor":false,"prefix":"","firstName":"Van","middleName":"Paepegem","lastName":"Wim","suffix":""},{"id":530669425,"identity":"67c5cd23-64bf-41a2-bd43-d9726420921d","order_by":2,"name":"Tom Goffin","email":"","orcid":"","institution":"Ghent University","correspondingAuthor":false,"prefix":"","firstName":"Tom","middleName":"","lastName":"Goffin","suffix":""},{"id":530669426,"identity":"64d8dc9c-4c6d-4e09-a1d8-f895bd48e61c","order_by":3,"name":"Villeirs Geert","email":"","orcid":"","institution":"University Hospital Ghent","correspondingAuthor":false,"prefix":"","firstName":"Villeirs","middleName":"","lastName":"Geert","suffix":""},{"id":530669427,"identity":"a99e0c89-088f-481a-a71b-b6483eda1f91","order_by":4,"name":"Coopman Renaat","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA8klEQVRIie3QPQrCMACA0XRyibhGVHqFlkIdLF29RkKhUw7RKS5216lX0K2jUtQl1NXSxQNU0M2CotaqW6yOYj4IyZBHfgCQyX4gq5wUr2ojfq3cEnxDomrSHfjzfX4EaqPmLyZ5uCZB319tldASkjaPneYQA308jN3E5ymZ8ph4CneFBCGqIYgB1jbU3NTZjYyo7iksEhM1M/LTgyRnFpNpkBXk8uYUaLaep6R1NiMBggWZiQmkZq/tovtb0g5zjAmk+ohwR0xq3Egyyyp+bJnsmN0JBlzbH0JbSJ7yNmC51Ior4SpQ9iCq99l2mUwm+5+u5b5YQQFJsa8AAAAASUVORK5CYII=","orcid":"","institution":"University Hospital Ghent","correspondingAuthor":true,"prefix":"","firstName":"Coopman","middleName":"","lastName":"Renaat","suffix":""}],"badges":[],"createdAt":"2025-09-16 19:08:09","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7633361/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7633361/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":93777379,"identity":"deadfdd0-3ede-442a-931b-5e2ce81f6c8d","added_by":"auto","created_at":"2025-10-17 12:41:09","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2036115,"visible":true,"origin":"","legend":"","description":"","filename":"2025.09.16MLClinicalRelevanceofAccuracyinInHouse3DPrintingInCraniomaxillofacialsurgeryAComparativeStudyofFFFSLAandMJTechnologies3Dprintinginmedicine.docx","url":"https://assets-eu.researchsquare.com/files/rs-7633361/v1/4dfdcd69fbfe553f644012e6.docx"},{"id":93777380,"identity":"a11a34c3-6646-485a-9877-c3fbb95010df","added_by":"auto","created_at":"2025-10-17 12:41:09","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6786,"visible":true,"origin":"","legend":"","description":"","filename":"d6b4e2e5d4704269a4b474258b955ff5.json","url":"https://assets-eu.researchsquare.com/files/rs-7633361/v1/4f9175566b95fcf7ad8860a9.json"},{"id":93775752,"identity":"92acd351-7efd-4588-9a60-75966f6641fb","added_by":"auto","created_at":"2025-10-17 12:33:09","extension":"xlsx","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":24720,"visible":true,"origin":"","legend":"","description":"","filename":"LinearMixedModelArtikel2SPSS.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7633361/v1/0c50b85e0c20401c43a4ee74.xlsx"},{"id":93775748,"identity":"c0180266-3415-4509-b717-29322a13c48d","added_by":"auto","created_at":"2025-10-17 12:33:09","extension":"xml","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":70553,"visible":true,"origin":"","legend":"","description":"","filename":"d6b4e2e5d4704269a4b474258b955ff51enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7633361/v1/700a2071c4f558037f383494.xml"},{"id":93775755,"identity":"acccd36d-440e-44d9-8749-7fb18a22909d","added_by":"auto","created_at":"2025-10-17 12:33:09","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":500647,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7633361/v1/8123cc550eec0c2e4b6e4a3d.png"},{"id":93775753,"identity":"769a60fe-c3d4-4233-84b8-fa0fa2f5d942","added_by":"auto","created_at":"2025-10-17 12:33:09","extension":"emf","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":4489524,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.emf","url":"https://assets-eu.researchsquare.com/files/rs-7633361/v1/0a94875256f39b6de1d59d42.emf"},{"id":93775759,"identity":"f9eee13f-06b1-4f5b-8db1-f17eb86f6ed3","added_by":"auto","created_at":"2025-10-17 12:33:10","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":495053,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7633361/v1/b4e235d915ebf3896de811da.png"},{"id":93777383,"identity":"bfcf0f95-72fb-4b4f-a002-7be0c8bb62ab","added_by":"auto","created_at":"2025-10-17 12:41:09","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":52588,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7633361/v1/6712f8499d28e885475ad61d.png"},{"id":93778836,"identity":"ebb62e37-78e8-401c-8522-9a23f489d74d","added_by":"auto","created_at":"2025-10-17 12:49:09","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":566043,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7633361/v1/304ca5642911a423af572396.png"},{"id":93775751,"identity":"1b4d4f36-c2a8-4122-a1ae-573305be983a","added_by":"auto","created_at":"2025-10-17 12:33:09","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":65971,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7633361/v1/92d45c7355c35d13980d806a.png"},{"id":93777382,"identity":"047d7925-33f2-47de-99c7-1d63360c3bbd","added_by":"auto","created_at":"2025-10-17 12:41:09","extension":"xml","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":70342,"visible":true,"origin":"","legend":"","description":"","filename":"d6b4e2e5d4704269a4b474258b955ff51structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7633361/v1/c0b7fbace2d1d788bf29a394.xml"},{"id":93775757,"identity":"36372c65-ebcb-4389-84e7-aefde82aeb78","added_by":"auto","created_at":"2025-10-17 12:33:09","extension":"html","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":78644,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7633361/v1/c75cc9516060048c92eb9138.html"},{"id":93775745,"identity":"0a64219b-0cc3-4c25-b3c2-21e58e2ac5d2","added_by":"auto","created_at":"2025-10-17 12:33:09","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":139555,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eMethodological study design\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMethodological study design: Visual overview of the experimental workflow used to assess the accuracy of three 3D printing technologies (FFF, SLA, MJ) using clinically relevant anatomical models. Trueness was evaluated by comparing each printed model to its original STL reference (green arrow). Precision was assessed through intra-build variability and inter-build variability (red arrow), by comparing models printed simultaneously or across different builds.\u003c/p\u003e","description":"","filename":"1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7633361/v1/50381133e1d05c8d356aae81.jpg"},{"id":93775743,"identity":"8066ece9-f450-4cf6-90cf-4ae1a7f287f2","added_by":"auto","created_at":"2025-10-17 12:33:09","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":122351,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePart Comparison Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDemonstration of the part comparison analysis conducted in Materialise Mimics 18.0 (Materialise NV, Leuven, Belgium) for two 3D printed mandible models. First, the two models were aligned (N-point registration and global registration), then a point cloud part comparison analysis is carried out. A heat map shows the areas of aberrations.\u003c/p\u003e","description":"","filename":"2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7633361/v1/3222683d84cb8bc450eff401.jpg"},{"id":93777381,"identity":"d0708f45-26b1-4fd9-a77e-a0f6284723d0","added_by":"auto","created_at":"2025-10-17 12:41:09","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":175118,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eStatistical model design (Linear Mixed Model)\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIllustration of the linear mixed model used to analyze log-transformed RMS values (LogRMS). The model included fixed effects for anatomical model, printer type, and comparison type (shown in green). A random intercept was added for each unique combination of these three fixed factors (illustration of the clustering shown in red) to account for the hierarchical structure and repeated conditions in the experimental design. This modeling approach allows appropriate estimation of fixed effects while controlling for internal variance within clustered measurements.\u003c/p\u003e","description":"","filename":"3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7633361/v1/5bdc2f11cb708f473f813f8b.jpg"},{"id":93780054,"identity":"3d4998ce-71b7-43f8-a5bf-a2b0e4e83cc7","added_by":"auto","created_at":"2025-10-17 12:57:10","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1267307,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7633361/v1/08529f4a-3bee-4209-845c-9f33a54e2f8e.pdf"},{"id":93775746,"identity":"7badca20-ffe4-4548-a5fe-d60c0958c9a6","added_by":"auto","created_at":"2025-10-17 12:33:09","extension":"xlsx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":24720,"visible":true,"origin":"","legend":"","description":"","filename":"LinearMixedModelArtikel2SPSS.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7633361/v1/d915574289a9d976ffacab2a.xlsx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Clinical relevance of accuracy in in-house 3D printing in craniomaxillofacial surgery: A comparative study of FFF, SLA, and MJ Technologies","fulltext":[{"header":"Background","content":"\u003cp\u003e3D printing, also known as additive manufacturing or rapid prototyping, is a process that involves constructing a three-dimensional solid object of virtually any shape by adding material one layer at a time based on a digital model. Common 3D printing technologies include fused filament fabrication (FFF), stereolithography (SLA), selective laser sintering (SLS), material jetting (MJ) and binder jetting (BJ) (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eIn the last decade, the increasing availability of low-cost 3D printers, combined with advancements in CAD/CAM technology and medical imaging, has led to an increase in 3D printing applications within the medical field. Initially, 3D printing was used primarily for treatment planning and education. More recently, it has expanded to patient-specific applications such as surgical guides, customized implants, and models for complex anatomy. These applications have demonstrated significant clinical benefits, including improved comprehension of complex anatomy, reduction of operating time, and more precise preoperative planning (\u003cspan additionalcitationids=\"CR2 CR3 CR4 CR5 CR6 CR7\" citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eAmong the various 3D printing technologies, FFF and SLA are the most accessible and widely used in hospital settings due to their affordability, ease of use, and the availability of compatible materials. FFF, which extrudes thermoplastic filaments layer by layer, is popular for producing anatomical models for surgical planning, patient education, and medical device prototyping. It is cost-effective, but it has limitations in resolution and surface finish​. SLA, a vat photopolymerization method using liquid resin cured by ultraviolet (UV) light, offers higher resolution and smoother surface finishes, making it ideal for printing surgical guides and dental models​. However, SLA-printed parts require post-processing to remove uncured resin and enhance mechanical properties.\u003c/p\u003e\u003cp\u003eIn contrast, MJ is primarily used for industrial applications such as automotive engineering, aerospace, and high-precision product manufacturing. This method enables high-resolution multi-material and full-colour printing, which is beneficial in complex prototyping and high-detail modelling​. Despite its advantages, MJ is less common in hospital settings due to its high equipment and material costs, as well as the complexity of operating and maintaining the system​ (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eTo address the cost and logistical challenges of outsourcing 3D printing services, many hospitals are now developing in-house 3D printing laboratories, referred to as point-of-care (POC) manufacturing. This approach has opened new possibilities, such as bioprinting and the fabrication of custom implants (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e). However, with these advancements come additional regulatory considerations. Under the European Union's Medical Device Regulation (MDR) 2017/745, which governs in-house production of medical devices, including 3D printed models, hospitals must ensure compliance with strict safety, quality, and performance standards for patient-specific devices. MDR places stringent requirements on POC manufacturing, particularly when creating patient-specific medical devices, to ensure the reliability and safety of such models in clinical settings (\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eAccuracy (including trueness and precision) in the 3D printing process is critical for safe intraoperative use of these models. However, errors can occur at various stages of the manufacturing process, such as data acquisition, image processing, and model fabrication. These steps are influenced by the equipment used, the software, and the expertise of the operator. Additionally, the segmentation process from Digital Imaging and Communications in Medicine (DICOM) to Standard Tessellation Language (STL), particularly when using Cone Beam Computed Tomography (CBCT) or Multi-slice Computed Tomography (MSCT), is highly user-dependent and may introduce errors such as inaccurate volumes or incorrect geometries. To better assess the precision and trueness of 3D printers, this study eliminates the segmentation step by using a high-precision optical 3D scanner (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan additionalcitationids=\"CR10\" citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eAlthough manufacturers provide specifications on the accuracy of their 3D printers, these tests often involve simple symmetrical objects that do not reflect the complexity of anatomical models, leading to potential inaccuracies such as incorrect volumes or deformations in clinical practice (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThe aim of this study is to evaluate the accuracy of 3D printing in a clinically relevant setting using highly accurate anatomical models. This investigation focuses on three 3D printing technologies: FFF, SLA, and MJ. By doing so, we aim to provide a deeper understanding of the accuracy of 3D printed models for medical applications, particularly in the context of stringent quality control requirements under MDR regulations.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003eStudy Design\u003c/h2\u003e\u003cp\u003eThis study was designed to evaluate the accuracy of three 3D printing technologies in reproducing clinically relevant anatomical models.\u003c/p\u003e\u003cp\u003eAccuracy was assessed in terms of precision and trueness, following the metrological definitions described in ISO 5725-1(\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e). Precision refers to how consistent repeated measurements are under the same conditions, reflecting the variability or spread (i.e., low variance).\u003c/p\u003e\u003cp\u003eTrueness describes how closely a measurement matches the reference value, reflecting systematic error (i.e., low bias). In this study, trueness indicates how closely a printed model matches its original STL file, while precision reflects how similar the printed models are to each other.\u003c/p\u003e\u003cp\u003eTo evaluate precision and trueness, two experimental setups were used, as illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. Precision was assessed by examining both intra-build variability (repeatability) and inter-build variability (reproducibility). Intra-build variability was measured by printing multiple copies of the same anatomical model simultaneously in a single build under identical conditions. Inter-build variability was assessed by printing the same models on different days using the same printer and settings.\u003c/p\u003e\u003cp\u003eAll models used for precision assessment were subsequently also compared to their corresponding original STL files to evaluate trueness.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eFigure \u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. Methodological study design: Visual overview of the experimental workflow used to assess the accuracy of three 3D printing technologies (FFF, SLA, MJ) using clinically relevant anatomical models. Trueness was evaluated by comparing each printed model to its original STL reference (green arrow). Precision was assessed through intra-build variability and inter-build variability (red arrow), by comparing models printed simultaneously or across different builds.\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eAnatomical models and 3D printing workflow\u003c/h3\u003e\n\u003cp\u003eFour different anatomical models were selected from a high-quality anatomically accurate skull model (SOMSO-Plast\u0026reg; Bauchene QS9/1, Coburg, Germany (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e)): the sphenoid bone, frontal bone, zygomatic bone, and mandible. These models were chosen for their variability in size, geometry, and complexity. Each model was scanned using a MSCT with a high-resolution object scanning protocol. The scan settings were: slice thickness 0.6 mm, increment 0.4 mm, pitch 0.8, rotation time 1.0 s, effective mAs 420, with CARE Dose and CARE kV turned off. DICOM (Digital Imaging and Communications in Medicine) data from the MSCT was processed using Materialise Innovation Suite\u0026reg; (MIS, Leuven, Belgium). Segmentation was performed using a fixed threshold of -200 Hounsfield units (HU) across all models. The resulting voxel-based segmentations were converted into STL files for 3D printing.\u003c/p\u003e\u003cp\u003eThree 3D printing technologies were used.\u003c/p\u003e\u003cp\u003eThe Felix Pro 3\u0026reg; printer was used for Fused Filament Fabrication (FFF), with a build volume of 237 \u0026times; 244 \u0026times; 235 mm. A 0.4 mm nozzle and a layer height of 200 microns were used, printing with PLA filament (\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThe Formlabs Form 3B+\u0026reg; printer was used for Stereolithography (SLA), with a build volume of 145 \u0026times; 145 \u0026times; 165 mm. It printed with Formlabs Model Resin at a 100 micron layer thickness (\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e). The The Stratasys J5 MediJet\u0026reg; printer was used for Material Jetting (MJ), offering a build volume of 140 \u0026times; 200 \u0026times; 190 mm and printing with MED620 material at a 27 micron layer thickness (\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eAll printers were properly calibrated prior to use. All prints were produced under identical conditions (temperature, relative humidity, conditioned room) using raw materials of equal age. Printer settings were based on the respective manufacturers' recommendations and were kept constant across builds to avoid introducing variability through slicing or process parameters.\u003c/p\u003e\u003cp\u003eAfter 3D printing, post-processing of the parts was performed. For FFF the supports were removed manually. The SLA printed parts first underwent a washing process of 10 minutes in \u0026ge;\u0026thinsp;99% isopropyl alcohol (IPA), to remove any excess resin on the surface of the parts, and 5 minutes of post-curing at 60\u0026deg;C to achieve their optimal mechanical properties. Afterwards, the supports were removed manually. For MJ printed parts, the gel coat surrounding the parts is removed by submerging the parts in water followed using a water jet.\u003c/p\u003e\n\u003ch3\u003ePrecision assessment\u003c/h3\u003e\n\u003cp\u003eFor intra-build precision, multiple identical copies of each anatomical model were printed on the same build platform. The number of replicas per build was determined by the available build volume of each printer. For example, the MJ printer consistently produced four copies per model, while FFF and SLA produced two to three copies depending on the model size.\u003c/p\u003e\u003cp\u003eFor inter-build precision, one copy of each model was printed on three separate occasions over the course of one week, using identical settings.\u003c/p\u003e\u003cp\u003eThe consistency of each group of prints was evaluated by calculating the deviations between all possible pairwise combinations of the models within the same group (intra-build or inter-build). These deviations quantify how similar the models are to each other and reflect the precision of the printing process.\u003c/p\u003e\n\u003ch3\u003eTrueness assessment\u003c/h3\u003e\n\u003cp\u003eTo evaluate trueness, each printed model was compared to the original segmented STL file. All models printed for intra-build and inter-build analysis were included in the trueness analysis. Thus, trueness was assessed across a range of build conditions and printers, providing a comprehensive picture of how faithfully each technology reproduces the reference geometry.\u003c/p\u003e\n\u003ch3\u003eOptical scan of printed models\u003c/h3\u003e\n\u003cp\u003eEach 3D printed model was scanned using an optical white light desktop 3D scanner (EinScan-SP, SHINING 3D Tech. Co., Ltd.). The scanner uses white light and coherence scanning interferometry (CSI) to perform non-contact surface digitization, with accuracy better than 50 microns (\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eEach model was scanned in four orientations (cranial, caudal, left lateral, right lateral), and the scans were merged into a single non-watertight STL model. No post-processing (such as smoothing or hole filling) was applied. The final STL model was used for dimensional comparison with either the original STL file (for trueness) or other scans within the same group (for precision)(\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e).\u003c/p\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003ePart comparison analysis and accuracy metric\u003c/h2\u003e\u003cp\u003eThe comparison analysis was performed in Materialise Mimics 18.0 (Materialise NV, Leuven, Belgium). STL files of both the 3D printed models and the original model were imported. For each analysis, the pair of STL models were aligned using a two-step process. An initial alignment was performed using six manually placed landmarks (N-point registration), followed by fine-tuning via global surface registration.\u003c/p\u003e\u003cp\u003eTo quantify accuracy, a part comparison analysis was conducted subsequently, which is based on point-cloud comparison. This analysis measures deviations and calculates the Root Mean Square (RMS) deviation, which is the square root of the mean of the squared distances between corresponding surface points of the two aligned STL models. RMS reflects the overall magnitude of deviation, regardless of direction, making it more robust than the mean deviation (MD), which can be affected by the cancellation of positive and negative errors. Unlike MD, RMS accounts for both the distribution and magnitude of deviations and is therefore a more appropriate measure of overall dimensional error.\u003c/p\u003e\u003cp\u003eDuring this part comparison analysis, a heat map was generated to visually highlight areas of dimensional deviation (Fig. \u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eFigure \u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e: Demonstration of the part comparison analysis conducted in Materialise Mimics 18.0 (Materialise NV, Leuven, Belgium) for two 3D printed mandible models. First, the two models were aligned (N-point registration and global registration), then a point cloud part comparison analysis is carried out. A heat map shows the areas of aberrations.\u003c/p\u003e\u003cp\u003eRMS values were used as the primary outcome in the statistical analysis to evaluate the effects of printer type, anatomical model, and comparison type (intra-build, inter-build, trueness) on the accuracy of 3D printed models.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003eStatistical analysis\u003c/h2\u003e\u003cp\u003eTo address the complex and partially dependent structure of the data, a Linear Mixed Model (LMM) was employed (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). LMMs are particularly well suited for datasets with repeated measures, hierarchical nesting, or clustering, as they allow the modelling of both fixed effects (systematic factors like printer type) and random effects (uncontrolled variability across clusters). This approach allows for more accurate estimation of fixed effects while controlling for within-group correlation. Because LMMs assume normally distributed residuals, the RMS values in this study were log-transformed to better meet this assumption and improve model fit.\u003c/p\u003e\u003cp\u003eThe transformed RMS values were then used as the dependent variable in the LMM. Fixed effects included anatomical model, printer, and comparison type (intra-build, inter-build, and trueness). To correct for internal clustering of measurements under shared experimental conditions, a random intercept was added for each unique combination of anatomical model, printer, and comparison type.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eFigue 3. Illustration of the linear mixed model used to analyze log-transformed RMS values (LogRMS). The model included fixed effects for anatomical model, printer type, and comparison type (shown in green). A random intercept was added for each unique combination of these three fixed factors (illustration of the clustering shown in red) to account for the hierarchical structure and repeated conditions in the experimental design. This modeling approach allows appropriate estimation of fixed effects while controlling for internal variance within clustered measurements.\u003c/p\u003e\u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003eThe accuracy of the three printing technologies was assessed using a Linear Mixed Model (LMM), which evaluated the effects of printer type, anatomical model, and comparison type on RMS error, while adding a random intercept to account for data clustering within printer\u0026ndash;model\u0026ndash;comparison combinations (Table \u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eLinear Mixed Model Analysis Results (RMS accuracy)\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"8\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eEffect\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eGroup\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eMean (LogRMS)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003e95% CI\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eBack-transformed RMS (\u0026micro;m)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003eF\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c7\"\u003e\u003cp\u003edf\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c8\"\u003e\u003cp\u003ep-value\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFixed Effects\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u003cb\u003ePrinter\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e\u003cb\u003e15.646\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.001\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eFelix (FFF)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e\u0026ndash;2.041\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c4\"\u003e\u003cp\u003e[\u0026ndash;2.222, \u0026minus;\u0026thinsp;1.861]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e130\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eForms3 (SLA)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e\u0026ndash;2.221\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c4\"\u003e\u003cp\u003e[\u0026ndash;2.402, \u0026minus;\u0026thinsp;2.040]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e109\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eStratasys (MJ)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e\u0026ndash;2.695\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c4\"\u003e\u003cp\u003e[\u0026ndash;2.874, \u0026minus;\u0026thinsp;2.517]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e67\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u003cb\u003eAnatomical model\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e\u003cb\u003e11.764\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.001\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eFrontal bone\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e\u0026ndash;2.045\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c4\"\u003e\u003cp\u003e[\u0026ndash;2.250, \u0026minus;\u0026thinsp;1.839]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e129\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMandible\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e\u0026ndash;2.127\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c4\"\u003e\u003cp\u003e[\u0026ndash;2.333, \u0026minus;\u0026thinsp;1.920]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e119\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSphenoid bone\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e\u0026ndash;2.305\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c4\"\u003e\u003cp\u003e[\u0026ndash;2.509, \u0026minus;\u0026thinsp;2.100]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e100\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eZygomatic bone\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e\u0026ndash;2.801\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c4\"\u003e\u003cp\u003e[\u0026ndash;3.005, \u0026minus;\u0026thinsp;2.596]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e61\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u003cb\u003eComparison type\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e\u003cb\u003e1.206\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e\u003cb\u003e0.305\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eInter-build\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e\u0026ndash;2.324\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c4\"\u003e\u003cp\u003e[\u0026ndash;2.487, \u0026minus;\u0026thinsp;2.160]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e98\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eIntra-build\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e\u0026ndash;2.424\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c4\"\u003e\u003cp\u003e[\u0026ndash;2.692, \u0026minus;\u0026thinsp;2.156]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e89\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrueness\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e\u0026ndash;2.211\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c4\"\u003e\u003cp\u003e[\u0026ndash;2.350, \u0026minus;\u0026thinsp;2.072]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e110\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e: Estimated marginal means from the linear mixed model with fixed effects for printer type, anatomical model, and comparison type. RMS values are log-transformed; back-transformed RMS values are provided in micrometers (\u0026micro;m). Significance was found for printer and anatomical model (p\u0026thinsp;\u0026lt;\u0026thinsp;.001), but not for comparison type (p\u0026thinsp;=\u0026thinsp;.305).\u003c/p\u003e\u003cp\u003eThe estimated variance of the random intercept was 0.197, compared to a residual variance of 0.045 (Table \u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). This implies that approximately 81.4% of the total variance in LogRMS was attributable to clustering within printer-model-comparison combinations, with only 18.6% attributable to residual error.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eRandom effects (intercept clustering)\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"4\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eEffect\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eVariance Estimate\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eStandard Error\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eIntercept (ClusterID)\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.197\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.037\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eResidual\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.045\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.009\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e: Random effects from the linear mixed model, showing variance estimates and standard errors for intercept clustering and residual error.\u003c/p\u003e\u003cp\u003eIn the following sections, mean refers to the estimated marginal mean of the log-transformed RMS values, with corresponding 95% confidence intervals (CI) and back-transformed RMS values (in \u0026micro;m) presented for interpretability (Table \u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003ePrinter Effect\u003c/h2\u003e\u003cp\u003ePrinter type had a significant effect on LogRMS (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001).\u003c/p\u003e\u003cp\u003eThe Stratasys printer had the lowest LogRMS (mean = \u0026minus;\u0026thinsp;2.695, 95% CI [\u0026ndash;2.874, \u0026minus;\u0026thinsp;2.517]) corresponding to a back-transformed RMS of 67 \u0026micro;m. This was followed by the Forms3 (mean = \u0026minus;\u0026thinsp;2.221, 95% CI [\u0026ndash;2.402, \u0026minus;\u0026thinsp;2.040]; 109 \u0026micro;m) and the Felix printer (mean = \u0026minus;\u0026thinsp;2.041, 95% CI [\u0026ndash;2.222, \u0026minus;\u0026thinsp;1.861]; 130 \u0026micro;m).\u003c/p\u003e\u003cp\u003ePost hoc comparisons confirmed that Stratasys differed significantly from both other printers (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003eAnatomical Model Effect\u003c/h2\u003e\u003cp\u003eAnatomical model also had a significant effect on LogRMS (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0 .001).\u003c/p\u003e\u003cp\u003eThe Zygoma model showed the lowest LogRMS (mean = \u0026minus;\u0026thinsp;2.801, 95% CI [\u0026ndash;3.005, \u0026minus;\u0026thinsp;2.596]; 61 \u0026micro;m), followed by Sphenoid model (mean = \u0026minus;\u0026thinsp;2.305, 95% CI [\u0026ndash;2.509, \u0026minus;\u0026thinsp;2.100]; 100 \u0026micro;m), Mandible (mean = \u0026minus;\u0026thinsp;2.127, 95% CI [\u0026ndash;2.333, \u0026minus;\u0026thinsp;1.920]; 119 \u0026micro;m), and Frontal bone (mean = \u0026minus;\u0026thinsp;2.045, 95% CI [\u0026ndash;2.250, \u0026minus;\u0026thinsp;1.839]; 129 \u0026micro;m).\u003c/p\u003e\u003cp\u003ePost hoc testing showed that Zygoma model differed significantly from all other models (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.01), while other pairwise differences were less consistent.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003ePart Comparison Type Effect\u003c/h2\u003e\u003cp\u003eComparison type did not have a significant effect on LogRMS (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.305).\u003c/p\u003e\u003cp\u003eThe intra-build group had a mean of \u0026minus;\u0026thinsp;2.424 (95% CI [\u0026ndash;2.692, \u0026minus;\u0026thinsp;2.156]), corresponding to a back-transformed RMS of 89 \u0026micro;m. The inter-build group had mean of \u0026minus;\u0026thinsp;2.324 (95% CI [\u0026ndash;2.487, \u0026minus;\u0026thinsp;2.160]; 98 \u0026micro;m), and the trueness group had a mean of \u0026minus;\u0026thinsp;2.211 (95% CI [\u0026ndash;2.350, \u0026minus;\u0026thinsp;2.072]; 110 \u0026micro;m).\u003c/p\u003e\u003cp\u003ePost hoc comparisons revealed no significant differences between the three comparison types.\u003c/p\u003e\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study aimed to evaluate the dimensional accuracy of 3D printed anatomical models using three 3D printing technologies\u0026mdash;Fused Filament Fabrication (FFF), Stereolithography (SLA), and Material Jetting (MJ)\u0026mdash;within the framework of in-house clinical use under MDR regulations. Accuracy was approached as a combination of trueness (closeness to a reference) and precision (reproducibility). The experimental design included multiple builds and comparison types (intra-build, inter-build, and trueness comparisons), providing a clinically relevant validation strategy.\u003c/p\u003e\u003cp\u003eAlthough the study design was structured to separately assess trueness and precision through various comparison types (intra-build, inter-build, trueness), all measurements were analysed together in a single, unified Linear Mixed Model (LMM). This approach allowed for the inclusion of all relevant data points while appropriately accounting for their nested structure. By modelling printer type, anatomical model, and comparison type as fixed effects, and incorporating a random intercept for each printer\u0026ndash;model\u0026ndash;comparison combination, the LMM controlled for shared experimental conditions and internal clustering. This statistical framework offered a robust and scalable way to capture both systematic influences and internal variability, making it well-suited for evaluating accuracy in complex clinical 3D printing workflows.\u003c/p\u003e\u003cp\u003eThe LMM revealed significant effects of both printer type and anatomical model on accuracy, while the comparison type (intra-build, inter-build, trueness) had no significant impact. Notably, 81.4% of the total variance in LogRMS was attributed to the random intercept, confirming strong clustering and validating the need for mixed modeling.\u003c/p\u003e\u003cp\u003ePrinter type had a clear and significant influence on accuracy. After adjusting for anatomical model and internal clustering, the MJ printer (Stratasys) showed the highest accuracy with an estimated RMS of 67 \u0026micro;m, significantly outperforming both SLA (Forms3, 109 \u0026micro;m) and FFF (Felix, 130 \u0026micro;m). These results confirm that printer technology remains a dominant factor in achieving accurate prints and suggest that MJ is particularly suitable for applications requiring high dimensional fidelity, such as surgical guides.\u003c/p\u003e\u003cp\u003eThe anatomical model also significantly influenced the results. The Zygoma model consistently showed the lowest deviations, likely due to its smaller size and geometric simplicity compared to more complex models like the mandible or sphenoid. These findings highlight that anatomical geometry is an important confounder and should be considered in validation studies of 3D printing accuracy.\u003c/p\u003e\u003cp\u003eInterestingly, the comparison type\u0026mdash;whether evaluating intra-build, inter-build, or trueness\u0026mdash;did not significantly influence the outcome when corrected for printer and model. This suggests that all three evaluation types capture similar aspects of dimensional accuracy under controlled printing conditions, and that the dominant variation arises from the printer and model rather than from how the comparison is structured.\u003c/p\u003e\u003cp\u003eIt should also be noted that the printers were evaluated using their standard manufacturer-recommended settings, which included different layer thicknesses: 200 \u0026micro;m for FFF, 100 \u0026micro;m for SLA, and 27 \u0026micro;m for MJ. These inherent differences in vertical resolution likely contributed to the superior accuracy observed for the MJ printer. However, this was not an intentional bias; the goal was to assess each printer as it would realistically be used in clinical environments. Evaluating devices under their typical settings enhances the validity of the results and ensures that the conclusions are aligned with real-world applications.\u003c/p\u003e\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\u003ch2\u003eClinical Implications\u003c/h2\u003e\u003cp\u003eAll printers performed within the clinically acceptable tolerance, in craniomaxillofacial surgery procedures, of 0.5 mm, but the superior accuracy of MJ makes it the most reliable option for applications demanding high precision. SLA may be sufficient for visual anatomical models, whereas FFF, despite its accessibility and low cost, may require additional quality control protocols before being adopted in high-risk clinical procedures.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\u003ch2\u003eLimitations\u003c/h2\u003e\u003cp\u003eThis study was limited by the number of anatomical models tested and the use of fixed manufacturer settings. It did not evaluate the effects of build orientation, printer calibration variations, or post-processing steps such as sterilization or surface finishing, all of which could impact final accuracy. Future studies should incorporate these parameters to develop a more comprehensive quality assurance framework aligned with MDR standards.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\u003ch2\u003eStatistical methodology and future perspective\u003c/h2\u003e\u003cp\u003ePrevious studies on 3D printing accuracy often relied on univariate or non-parametric tests, which do not accommodate repeated measurements or internal clustering. By using a Linear Mixed Model, this study modelled both fixed and random effects simultaneously, revealing that over 80% of the variance was attributable to shared printing conditions\u0026mdash;something traditional methods fail to capture.\u003c/p\u003e\u003cp\u003eOnly one study in the field of dental model printing (Brown et al., 2018) applied a generalized linear mixed model to assess print accuracy. However, to the best of our knowledge, no previous studies have applied a Linear Mixed Model (LMM) in the context of clinically realistic anatomical 3D printing workflows.\u003c/p\u003e\u003cp\u003eThe present study is therefore the first to implement a LMM framework to evaluate the accuracy of in-house 3D printing, correcting for clustering effects due to repeated prints, anatomical model variation, and comparison type. This approach enabled a more robust analysis of fixed effects (e.g., printer type) while reducing the risk of type I and type II errors common in simpler statistical methods. As such, it offers a statistically rigorous standard that may serve as a reference for future validation studies under MDR compliance.\u003c/p\u003e\u003c/div\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThis study evaluated the dimensional accuracy of FFF, SLA, and MJ 3D printing technologies using anatomical models in a clinically relevant context. All printers performed within the commonly accepted tolerance of 0.5 mm, supporting their use in clinical workflows under MDR 2017/745. However, MJ technology demonstrated significantly higher accuracy, making it the most suitable option for applications demanding high dimensional fidelity. SLA and FFF remain cost-effective alternatives where such precision is less critical.\u003c/p\u003e\u003cp\u003eTo maintain safety and compliance in in-house 3D printing, regular accuracy validation\u0026mdash;addressing both trueness and precision, including intra- and inter-build variability\u0026mdash;should be integrated into ongoing quality control.\u003c/p\u003e\u003cp\u003eOur application of LMM provided a clinically realistic and statistically appropriate framework to analyse accuracy across multiple dimensions. Given its ability to account for nested data and repeated measures, this method should be considered a recommended standard for future 3D printing validation studies.\u003c/p\u003e\u003cp\u003eFuture research should also explore the clinical impact of 3D-printed models on procedural outcomes, assess long-term dimensional stability of materials, and investigate how post-processing steps (e.g., sterilization, surface finishing) affect final model accuracy.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003cp\u003eSince this is an in vitro study, no ethics approval is required, but due to its clinical importance, this study was submitted to the Ghent University Hospital Medical Ethics Committee. Official approval was obtained on July 2, 2024.\u003c/p\u003e\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003cp\u003eNot applicable. This study does not contain any individual person\u0026rsquo;s data in any form.\u003c/p\u003e\u003c/p\u003e\u003cp\u003e\u003ch2\u003eCompeting interests\u003c/h2\u003e\u003cp\u003eThe authors report there are no competing interests to declare.\u003c/p\u003e\u003c/p\u003e\u003ch2\u003eFunding\u003c/h2\u003e\u003cp\u003eThis research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eM.L. was responsible for the study design, data collection, analysis, interpretation of results, and writing of the manuscript. R.C. was the main responsible for the study design. Together with G.V., T.G., and W.V.P., he provided guidance, critical feedback, and supervision throughout the research process. All authors reviewed and approved the final manuscript.\u003c/p\u003e\u003ch2\u003eAcknowledgements\u003c/h2\u003e\u003cp\u003eThe author would like to thank Prof. Dr. Renaat Coopman and co-authors for their guidance and critical feedback throughout the research process. Their support was instrumental in the development of this study.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eData is provided within the manuscript or supplementary information files.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAimar A, Palermo A, Innocenti B. The Role of 3D Printing in Medical Applications: A State of the Art. J Healthc Eng. 2019;2019.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDiment LE, Thompson MS, Bergmann JHM. Clinical efficacy and effectiveness of 3D printing: A systematic review. BMJ Open 1 Dec 2017;7(12).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLouvrier A, Marty P, Barrab\u0026eacute; A, Euvrard E, Chatelain B, Weber E. e.a. How useful is 3D printing in maxillofacial surgery? J Stomatol Oral Maxillofac Surg. 2017;118(4):206\u0026thinsp;\u0026ndash;\u0026thinsp;12.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKing BJ, Park EP, Christensen BJ, Danrad R. On-Site 3-Dimensional Printing and Preoperative Adaptation Decrease Operative Time for Mandibular Fracture Repair. J Oral Maxillofac Surg. 2018;76(9):1950.e1-1950.e8.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWeijs WLJ, Coppen C, Schreurs R, Vreeken RD, Verhulst AC, Merkx MAW. e.a. Accuracy of virtually 3D planned resection templates in mandibular reconstruction. J Cranio-Maxillofac Surg. 2016;44(11):1828-32.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGeorge E, Liacouras P, Rybicki FJ, Mitsouras D. Measuring and establishing the accuracy and reproducibility of 3D printed medical models. Radiographics. 2017;37(5):1424\u0026ndash;50.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCalvo-Haro JA, Pascau J, Asencio-Pascual JM, Calvo-Manuel F, Cancho-Gil MJ. Del Ca\u0026ntilde;izo L\u0026oacute;pez JF, e.a. Point-of-care manufacturing: a single university hospital\u0026rsquo;s initial experience. 3D Print Med. december 2021;7(1).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWillemsen K, Magr\u0026eacute; J, Mol J, Noordmans HJ, Weinans H, Hekman EEG. e.a. Vital Role of In-House 3D Lab to Create Unprecedented Solutions for Challenges in Spinal Surgery, Practical Guidelines and Clinical Case Series. J Pers Med. 1 maart 2022;12(3).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMsallem B, Neha S, Shuaishuai C, Halbeisen FS. Hans-Florian Zeilhofer FMThieringer. Evaluation of the Dimensional Accuracy of 3D-Printed Anatomical Mandibular Models Using. J Clin Med. 2020;9(3):1\u0026ndash;18.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHuotilainen E, Jaanimets R, Val\u0026aacute;šek J, Marci\u0026aacute;n P, Salmi M, Tuomi J. e.a. Inaccuracies in additive manufactured medical skull models caused by the DICOM to STL conversion process. J Cranio-Maxillofac Surg. 2014;42(5).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDorweiler B, Baqu\u0026eacute; PE, Chaban R, Ghazy A, Salem O. Quality control in 3D printing: Accuracy analysis of 3D-printed models of patient-specific anatomy. Materials. 2021;14(4):1\u0026ndash;13.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLievens M, De Kock L, Ureel M, Villeirs G, Van Paepegem W, Coopman R. The Accuracy of an Optical White Light Desktop 3D Scanner and Cone Beam CT Scanner Compared to a Multi-Slice CT Scanner to Digitize Anatomical 3D Models: A Pilot Study. Craniomaxillofacial Trauma Reconstr juni. 2025;18(2):27.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eISO 5725-1. 2023(en), Accuracy (trueness and precision) of measurement methods and results \u0026mdash; Part 1: General principles and definitions [Internet]. [geciteerd 17 augustus 2025]. Beschikbaar op: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.iso.org/obp/ui/#iso:std:iso:5725:-1:ed-2:v1:en\u003c/span\u003e\u003cspan address=\"https://www.iso.org/obp/ui/#iso:std:iso:5725:-1:ed-2:v1:en\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSOMSO. QS 9/1, Bauchene Skull.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFelix printers. Quick Start Guide FELIX Pro 3. 2019.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFelix printers. Filaments - PLA TDS.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFormlabs. september. Form 3B+ - Low Force Stereolithography (LFS) 3D Printer. 2023.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFormlabs. Dental resin model. 11 september 2021.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eStratasys. Stratasys J5 MediJet. Strat J5 MediJet. 2022.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eStratasys. Dentajet Dental Materials.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eShining 3D, Einscan. SE/SP2 Dekstop 3D Scanner. 10 oktober 2024.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"3d-printing-in-medicine","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"tdpm","sideBox":"Learn more about [3D Printing in Medicine](https://threedmedprint.biomedcentral.com/)","snPcode":"41205","submissionUrl":"https://submission.nature.com/new-submission/41205/3","title":"3D Printing in Medicine","twitterHandle":"@BioMedCentral","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"3D printing, anatomical models, medical device regulation (MDR 2017/745), accuracy validation, in-house manufacturing, linear mixed model, point-of-care manufacturing","lastPublishedDoi":"10.21203/rs.3.rs-7633361/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7633361/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cb\u003eBackground\u003c/b\u003e Considering the Medical Device Regulation (MDR 2017/745), this study aims to evaluate the accuracy of three commonly used 3D printing technologies\u0026mdash;Fused Filament Fabrication (FFF), Stereolithography (SLA), and Material Jetting (MJ)\u0026mdash;within a clinically realistic in-house 3D printing workflow.\u003c/p\u003e\u003cp\u003e\u003cb\u003eMethods\u003c/b\u003e In this study, the accuracy of 3D printers was assessed using highly accurate anatomical models to mimic an in vitro clinical workflow, unlike previous validation studies that relied on geometric calibration models or simplified shapes.\u003c/p\u003e\u003cp\u003eAccuracy was defined through precision (intra- and inter-build variability) and trueness (deviation from the digital reference model), quantified using root mean square (RMS) error.\u003c/p\u003e\u003cp\u003eTo analyse the data structure\u0026mdash;comprising repeated prints, multiple models, and timepoints\u0026mdash;we used a Linear Mixed Model (LMM), which enabled evaluation of fixed effects (printer type, anatomical model, and comparison type) while correcting for internal clustering of data.\u003c/p\u003e\u003cp\u003e\u003cb\u003eResults\u003c/b\u003e All printers achieved clinically acceptable accuracy. MJ was significantly more accurate (RMS 67 \u0026micro;m) compared to SLA (109 \u0026micro;m) and FFF (130 \u0026micro;m). The Linear Mixed Model showed that accuracy was influenced by printer type and anatomical complexity.\u003c/p\u003e\u003cp\u003e\u003cb\u003eConclusion\u003c/b\u003e This is the first study in 3D printing accuracy research to apply a Linear Mixed Model, providing a statistically robust analysis framework. Among the evaluated technologies, MJ achieved the highest accuracy in clinically realistic conditions, supporting its use in applications that require high anatomical fidelity within the MDR framework.\u003c/p\u003e","manuscriptTitle":"Clinical relevance of accuracy in in-house 3D printing in craniomaxillofacial surgery: A comparative study of FFF, SLA, and MJ Technologies","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-17 12:33:04","doi":"10.21203/rs.3.rs-7633361/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewersInvited","content":"","date":"2025-10-06T13:11:17+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-09-17T02:11:54+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-09-17T02:08:13+00:00","index":"","fulltext":""},{"type":"submitted","content":"3D Printing in Medicine","date":"2025-09-16T18:52:30+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"3d-printing-in-medicine","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"tdpm","sideBox":"Learn more about [3D Printing in Medicine](https://threedmedprint.biomedcentral.com/)","snPcode":"41205","submissionUrl":"https://submission.nature.com/new-submission/41205/3","title":"3D Printing in Medicine","twitterHandle":"@BioMedCentral","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"b115f9cf-5227-4142-b08f-783a0855eea9","owner":[],"postedDate":"October 17th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2025-10-17T12:33:05+00:00","versionOfRecord":[],"versionCreatedAt":"2025-10-17 12:33:04","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7633361","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7633361","identity":"rs-7633361","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}