Adapting super-resolution reconstruction for skeletal analysis of clinical computed tomography data

Adapting super-resolution reconstruction for skeletal analysis of clinical computed tomography data | 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 Article Adapting super-resolution reconstruction for skeletal analysis of clinical computed tomography data An-Di Yim, Louise K. Corron, Kyra E. Stull This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8195787/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 7 You are reading this latest preprint version Abstract Clinical computed tomography (CT) datasets are increasingly common in skeletal research, yet archived retrospective datasets with thick-slice, anisotropic reconstructions are more commonly available to researchers than the original high-resolution scans. Models reconstructed from these suboptimal scans can produce distorted measurements and irregular surfaces. This study evaluates a super-resolution reconstruction (SRR) framework for generating high-resolution skeletal models from multiple orthogonal, thick-slice CT stacks. Archived CT scans of long bones from 33 individuals (0–16 years) were collected from National Taiwan University Hospital. For each individual, 3D models were generated from both the original thick-slice stacks and SRR-processed volumes. Linear measurements were compared with those taken directly from high-resolution picture archiving and communication system (PACS) renderings, and geometric similarity was quantified by signed Hausdorff distance and Dice similarity coefficient. SRR-reconstructed models showed lower measurement error and greater agreement with the PACS rendering than the thick-slice models. Surface geometry was generally consistent across model types, with localized deviations concentrated at metaphyseal and epiphyseal regions. SRR processing also produced smoother, more anatomically plausible surfaces with a clearer separation of fusing elements. These results demonstrate that SRR can improve virtual skeletal model quality from suboptimal clinical imaging, particularly for applications where anatomically precise surfaces are beneficial. Computed tomography (CT) morphometrics super-resolution reconstruction skeletal biology Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Introduction In recent years, advances in medical imaging and processing techniques have opened new avenues for biological anthropology and anatomy research. In particular, the use of computed tomography (CT) scans as data sources has become increasingly common in skeletal analysis literature (Simmons-Ehrhardt, 2021 ; Uldin, 2017 ). CT imaging offers several benefits, including strong contrast between dense objects (such as bones) and surrounding soft tissues (Weber, 2015 ), and the preservation of internal structures such as trabecular bone architecture, bony cavities, and their internal surfaces (Christensen et al., 2018 ; Scherf, 2013 ; Scherf & Tilgner, 2009 ). These capabilities extend the application of CT data beyond traditional osteometric analysis. CT scan-based datasets have been used to build virtual skeletal collections (e.g., Copes et al., 2016 ; Edgar et al., 2020 ; Fischer, 2023 ; Kistler et al., 2013 ; Stull & Corron, 2022 ), reconstruct incomplete or fragmented specimens (e.g., Gunz et al., 2009 ; Lautenschlager, 2016 ), perform advanced morphometric shape analysis (e.g., Huseynov et al., 2016 ; Zollikofer & Ponce de León, 2002), as well as develop new methods and document case findings in forensic sciences (Christensen et al., 2018 ; Garvin & Stock, 2016 ). Anatomical sciences typically rely on two primary sources of CT scan-based data. First, CT scans directly obtained from fossils, archaeological artifacts, cadavers, or dry skeletal elements can be conducted in research settings under controlled conditions. These scans are characterized by higher radiation doses (Goldman, 2007 ), minimal soft tissue, and stationary conditions to reduce noise and improve imaging quality (Gascho et al., 2018 ). However, this approach is limited by access to equipment and trained personnel. Second, clinical CT scans, generated in the hospital setting by radiologists or trained technicians as part of routine diagnostic practice, can be anonymized and repurposed to provide accessible and valuable retrospective data for clinical and non-clinical research. Previous studies (Colman et al., 2017 ; Colman et al., 2019 ) have demonstrated that 3D skeletal models reconstructed from clinical CT scans are accurate and precise representations of physical skeletal elements in situ or ex situ (e.g., Brough et al., 2013 ; Corron et al., 2017 ; Franklin et al., 2013 ). Measurements taken from these rendered 3D models showed no significant difference from those collected directly from dry bones or dry bone surface scans, with an average intra-observer error of less than 1 mm and an average inter-observer error of less than 3 mm, considerably less than some errors produced via traditional osteometric analysis on dry bones (Langley et al., 2018 ). Surface deviations between 3D surfaces derived from clinical CT scans and dry bone surface scans were also minimal (Colman et al., 2019 ), supporting their use in subsequent morphometric and shape analyses. In most clinical settings, CT scans are acquired as isotropic or near-isotropic volumetric datasets (equal voxel dimensions in all three spatial direction) and subsequently reformatted and resliced at three anatomical planes (axial, sagittal, and coronal) through a process known as multiplanar reconstruction (MPR) (Geijer & El-Khoury, 2006 ). Radiologists and physicians routinely use MPR to assess anatomical structures and pathologies from multiple perspectives (Ebert et al., 2021 ). While sufficient for diagnostic purposes, the MRP images present challenges for researchers. These reconstructions typically consist of thicker slices (3 to 5 mm slice thickness) with anisotropic voxel dimensions (unequal voxel dimensions) and are therefore of lower resolution. For researchers, an MPR dataset may be high-resolution in one dimension but distorted in others, leading to inaccurate measurement and uneven surfaces in 3D reconstructions. Unfortunately, instead of the original high-resolution volumetric datasets, it is often these MPR images that are distributed and stored in the hospital picture archiving and communication system (PACS) (e.g., Lee et al., 2005 ; Yoshinobu et al., 2011 ). As a result, researchers in human skeletal biology conducting retrospective data collection from clinical CT scans may be limited to these suboptimal, thick-slice reconstructions, which can result in poorly visible morphological features on CT slices and varying qualities of virtual reconstructions of skeletal elements after image segmentation, depending on the pre- and post-processing protocols and segmentation parameters set by the user (Stock et al., 2020 ). To reduce voxelation and stepping artifacts, smoothing algorithms can be applied to fill the gaps between voxels via interpolation (Bade et al., 2006 ; Zhang & Xia, 2021 ), improving the visual appearance of bone surfaces (Lasek & Piórkowski, 2021 ; Sawdayee et al., 2023 ; Stock et al., 2020 ). However, when applied to thick-slice CT data, such smoothing rarely eliminates voxelation completely, and researchers risk over-smoothing, altering surface morphology and erasing anatomical features (Carew et al., 2019 ; Johnson et al., 2016 ; Ledig et al., 2017 ), which can lead to inaccurate or unreliable osteological data obtained from these reconstructions (Stock et al., 2020 ). This inherent limitation highlights the need for reconstruction approaches that enhance resolution without relying entirely on smoothing algorithms. Super-resolution reconstruction (SRR) is a group of techniques that produces high-resolution images from low-resolution inputs (Park et al., 2003 ). In medical imaging processing, SRR has been applied to modalities such as CT or magnetic resonance imaging (MRI) to enhance spatial resolution and reduce anisotropy without increasing radiation exposure or acquisition time. While recent SRR developments have largely focused on machine learning-based approaches (Li et al., 2021 ; Qiu et al., 2023 ; Tang et al., 2025 ), the need for large, high-resolution training data to perfect their algorithms makes them impractical for researchers working with datasets such as retrospective clinical scans. In contrast, model-based SRR approaches present as a regularized optimization solution to estimate the most probable high-resolution image consistent with a series of low-resolution slices. Because the orthogonal slice stacks represent images of the same object, their geometric relationships can be approximated by a displacement matrix describing spatial shifts and orientation differences. Each set of low-resolution images is modeled as a geometrically transformed observation of the underlying high-resolution images with added noise (Moustafa et al., 2016 ; Tian & Ma, 2011 ). Solving the equation thus constitutes an inverse problem, in which the high-resolution volume is iteratively estimated by minimizing a cost function. The open-source NiftyMIC framework (Ebner et al., 2018a ; Ebner et al., 2018b ; Ebner et al., 2020 ), originally developed for neonatal brain MRI, implements such a model-based SRR pipeline to combine multiple orthogonal, low-resolution stacks into an isotropic, high-resolution reconstruction. The modular design of the pipeline and the need for an accessible and easily implementable tool to improve image resolution of archived clinical CT data in skeletal research makes it suitable for testing its applicability to the reconstruction of both 2D and 3D renderings of skeletal elements from low-resolution CT scans. This study investigates the utility of the model-based SRR framework implemented in NiftyMIC to reconstruct volumetric skeletal data from multiple anisotropic low-resolution thick-slice CT scan stacks (Fig. 1 ). We compare osteometric measurements across three model types: diagnostic PACS renderings, thick-slice reconstructions (“original”), and SRR-reconstructed models (“reconstructed”) to assess metric accuracy and conduct geometric comparisons between the thick-slice models and SRR-reconstructed models to evaluate the degree of geometric consistency. Results Table 1 summarizes the measurement agreement results. Intra-observer error rates were minimal for both observers (Observer 1: TEM = 0.610 mm, %TEM = 0.758%; Observer 2: TEM = 0.479 mm, %TEM = 0.505%). Inter-observer error was 1.243 mm (1.603%) for the “original” models and 0.439 mm (0.517%) for the “reconstructed” models, indicating slightly higher measurement consistency for the “reconstructed” models. Inter-rendering comparisons further demonstrated strong measurement reliability. The TEM between the 3D models was 1.357 mm (1.702%) for Observer 1 and 1.279 mm (1.619%) for Observer 2. Figure 2 shows the Bland-Altman plots showing inter-rendering differences in paired linear measurements. Measurement deviations between PACS renderings and “reconstructed” models are tightly centered around zero with narrower 95% limits of agreement. In contrast, the “original” models display slightly wider dispersion. The overall agreement in linear measurements between PACS renderings and “reconstructed” models (TEM = 0.828 mm, %TEM = 0.996%) is also better than between PACS renderings and “original” models (TEM = 1.411 mm, %TEM = 1.737%). Collectively, these results indicate that SRR reconstruction improves linear measurement accuracy relative to the thick-slice models while maintaining high intra- and inter-observer reliability. Table 1 Technical error of measurement (TEM) and relative technical error of measurement (%TEM) for inter-rendering, intra-observer, and inter-observer comparisons of linear osteometric measurements. Intra-observer error rates were minimal for both observers, while inter-observer and inter-rendering errors remained low. Inter-rendering comparison Observer 1 Observer 2 TEM (mm) %TEM TEM (mm) %TEM Original vs. reconstructed 1.357 1.702 1.279 1.619 Intra-observer comparison Observer 1 Observer 2 TEM (mm) %TEM TEM (mm) %TEM 0.610 0.758 0.479 0.505 Inter-observer comparison Original Reconstructed TEM (mm) %TEM TEM (mm) %TEM 1.243 1.603 0.439 0.517 Figure 3 shows the distributions of surface deviations between the “original” and “reconstructed” models, as measured by the signed Hausdorff distance. Generally, an average Hausdorff distance close to zero indicates a high level of agreement in model geometry. On average, the “reconstructed” model surfaces deviate between − 1 mm to 1 mm from the “original” model surfaces, indicating close geometric correspondence. However, more substantial local deviations were observed in some elements, with minimum and maximum Hausdorff distances exceeding − 10 mm in isolated regions. These larger deviations likely reflect partial separation of fusing epiphyses and localized surface depressions or gaps that were not fully resolved during segmentation and processing. Figure 3 also shows the distributions of the Dice similarity coefficient. Overall, the Dice similarity coefficients (0.682 to 0.978) suggest that the two model types have good geometric agreement. Representative examples of surface deviations are shown in Fig. 4 . For each long bone, one example was selected to provide a visual representation of surface similarity using heat maps of signed Hausdorff distance. The models presented in Fig. 4 are those with the greatest geometric discordance (lowest Dice similarity coefficients). The most pronounced deviations are generally localized at the metaphyseal surfaces at the bone ends, while diaphyseal surfaces generally show high levels of geometric agreement. ANOVA of the linear regression models confirmed that age significantly affects both the minimum and maximum surface deviations (signed Hausdorff distance) across all comparisons (p < 0.001 for both measures, Table 2 ). Regression coefficients showed that with increasing age, minimum deviation decreases (-0.495 mm per year) while maximum deviation increases (0.456 mm per year). Together, these results suggest, as individuals age and epiphyseal fusion progresses, the increasing difficulty in distinguishing fusing epiphyses from diaphyses introduces greater error and thus larger surface deviations, particularly at the metaphyseal surfaces. Table 2 Test of the effects of sex assigned at birth, age, and bone on the signed Hausdorff distance across all comparisons. Effect term df Sum of squares Mean square F p Minimum Hausdorff distance Age 1 306.42 306.42 28.18 < 0.001* Biological sex 1 6.46 6.46 0.59 0.4420 Bone 5 363.57 72.71 6.69 < 0.001* Age x bone 5 46.15 9.23 9,85 0.5169 Maximum Hausdorff distance Age 1 698.11 698.11 88.36 < 0.001* Biological sex 1 0.57 0.57 1.62 0.1567 Bone 5 64.08 12.82 0.07 0.7888 Age x bone 5 140.25 28.05 3.55 0.0044 The aligned “original” and “reconstructed” models corresponding to the examples shown in Fig. 4 are presented in Fig. 5 . “Original” models (shown in blue) are used as the reference, overlaid with “reconstructed” models (shown as red point clouds) to allow direct comparison. Qualitatively, the “reconstructed” models show smoother surfaces, particularly along curved surfaces of the diaphyses and at the metaphyseal surfaces. In contrast, the “original” models derived from thick-slice CT image stacks often exhibit visible voxelation and stepping artifacts. These visual comparisons highlight the advantages of the SRR framework in improving surface quality and recovering high-resolution skeletal morphology from clinically acquired low-resolution CT datasets. Discussion In this study, we evaluated the applicability of a model-based SRR pipeline to reconstruct high-resolution volumetric skeletal data from multiple stacks of low-resolution, thick-slice clinical CT scans using a non-adult sample. The results showed that the reconstructed volumes can be effectively segmented to produce 3D skeletal models that showed greater osteometric linear measurement accuracy than the thick-slice models, as demonstrated by lower technical error of measurement. Geometric comparisons confirmed that the overall geometry of the “reconstructed” models was broadly similar to the “original” models. Consistent with previous findings (Fu et al., 2017 ; Rathnayaka et al., 2011 ; Van den Broeck et al., 2014 ), localized surface deviations were most pronounced in the metaphyseal surfaces and epiphyseal regions, likely reflecting incomplete separation of fusing epiphyses or residual segmentation artifacts such as incompletely filled gaps in areas of thin cortical or sparse trabecular bone (Jeuthe et al., 2021 ). While these deviations may introduce localized errors in region-specific or landmark-based analyses, the overall geometry of the reconstructions is well-preserved. Qualitative visual assessments further showed that the “reconstructed” models exhibit smoother surfaces, particularly along curved diaphyseal regions and at metaphyseal or epiphyseal regions. These findings suggest that the NiftyMIC pipeline, with appropriate preprocessing steps, is a promising tool for enhancing 3D skeletal model quality derived from suboptimal clinical CT scan data. The “original” models, derived from thick-slice and lower-resolution clinical CT data, often exhibited surface artifacts such as voxelation and visible stepping between slices (Fig. 6 ). These artifacts are results of the discrete sampling (slicing) of a continuous anatomical structure (i.e., bone), a limitation inherent to medical image reconstructions. While over-smoothing resulting from the reconstruction pipeline may be a concern, over-smoothed surfaces tend to be systematically smaller due to the rounding of surface contours and the loss of fine details (Johnson et al., 2016 ; Ledig et al., 2017 ; Stock et al., 2020 ). No such systematic inward deviations were observed in our results. Instead, the overall deviations of the “reconstructed” models were between − 1 mm to 1 mm from the “original” model surface, showing no bias in the direction of surface deviations (inward or outward). These findings suggest that the visible improvements in model smoothness and surface quality are the results of improved resolution, and that the SRR algorithm mathematically estimates plausible surface geometry in regions where data were discontinuous or missing in the original stacks, producing models that are smoother and more geometrically consistent. The present study used a non-adult sample. This is particularly relevant to human skeletal variation research that relies on geometric morphometric approaches to obtain data for analysis. As demonstrated in previous studies, variation of different skeletal elements (e.g., Franklin et al., 2008 ; Huseynov et al., 2016 ; Noble & Hawks, 2025 ; Viðarsdóttir et al., 2002 ) during growth that persist as morphological differences in size and shape and into adulthood are not only general, but also localized to specific areas of bone, such as grooves, curves, and other anatomical reliefs. These characteristics should therefore be preserved as much as possible in virtual 3D renderings of juvenile bones for analyses to be reliable and accurate. The ability of the SRR pipeline to produce anatomically realistic, smoothly contoured skeletal surfaces may therefore be advantageous when access to dry bone samples is limited, as is the case for non-adult skeletons in anatomical collections (Albanese, 2003 ; Alemán et al., 2012 ; Cardoso, 2006 ; Stull & Corron, 2022 ). The postcranial elements used in this study also demonstrated the broad applicability of the SRR framework. Considering the majority of growth-related changes in the lower limb skeleton are related to the changes in the epiphyses or the relative shape of the diaphyses and epiphyses (Frelat & Mittereocker, 2011 ; Morimoto et al., 2018 ; Pujol et al., 2014 ; Pujol et al., 2016 ), a pattern likely shared by upper limb skeletons. Asymmetries in skeletal morphology as a result of environmental stress are also concentrated near bone ends, as shown in previous studies (Hong et al., 2021 ; Mopin et al., 2018 ). The results of this study showed that the most pronounced deviations across all model types were localized at the metaphyseal surfaces and epiphyseal regions, presumably because the improved image quality led to clearer separation of the unfused and fusing epiphyses during segmentation and better preservation of these regions in the “reconstructed” models. Preserving the geometry of intricate regions such as metaphyseal surfaces and epiphyses is crucial for geometric morphometric analyses of growth and development. Importantly, similar problems with epiphyses were observed in fused elements. Thus, while this study focused on non-adult skeletal elements, the SRR framework presented here may also be beneficial for analyses of adult bones by improving the resolution of features at bone ends. One limitation of the current study is the lack of “gold standards” (i.e., dry bones or their direct surface scans). Nevertheless, the use of osteometric measurements from PACS renderings serve as practical references since standardized osteometric data are consistent with their dry bone counterparts across 1:1 scaled medical imaging modalities (Lund et al., 2014 ; Stull et al., 2014b ). Importantly, comparisons showed improved measurement consistency and error rates for the “reconstructed” models compared to the “original” models. Based on the results presented here, it is clear that the SRR-based reconstruction framework is highly advantageous in generating high-quality 3D models from archived clinical CT scans for research and for teaching, as the high quality and likeness of the rendered surfaces could be 3D printed and used as a substitute for dry bones in classrooms where logistic constraints or ethical concerns regarding the use of dry skeletal remains can be a limiting factor (Carew et al., 2019 ; Shanley et al., 2024 ; Thomas et al., 2016 ). Another limitation of this study is the relatively small number of samples for each type of long bone. Future work using larger adult samples with fully fused epiphyses would allow for a more controlled evaluation of reconstruction accuracy, and subsequent testing across age groups could then clarify the broader applicability of this approach. In addition, a larger scale study focusing on comparing SRR-based reconstructions to models derived from high-resolution, isotropic thin-slice CT scans or dry bone surface scans could provide more robust validation (Li et al., 2021 ; Park et al., 2018 ; Tang et al., 2025 ). Despite this, improvements in surface quality and consistency are an important step toward the use of suboptimal clinical imaging for research. In summary, this study showed that the SRR framework implemented in NiftyMIC can be successfully adapted to generate high-quality 3D skeletal models from suboptimal, thick-slice clinical CT data. The “reconstructed” models show consistent osteometric measurements, improved surface smoothness, and clearer anatomical definition compared to models segmented directly from thick-slice scans. These proof-of-concept findings highlight SRR-based image registration as a practical approach for improving the utility of archived clinical imaging datasets, especially when access to high-resolution or physical skeletal samples is limited. Methods Archived clinical CT scans of long bones (humerus, radius, ulna, femur, tibia, or fibula) from individuals aged 0 to 16 years old of male or female sex assigned at birth were collected from National Taiwan University Hospital (NTUH) by a computerized search of the hospital PACS. Individuals were excluded if their CT scans did not include a complete long bone or if the long bone exhibited visible fractures or anomalies. The CT scans were originally performed between 2006 and 2017 as part of clinical diagnostic examinations using either a 256-slice or a 128-slice multidetector CT scanner. Imaging parameters varied across cases due to differences in clinical protocols but generally ranged from 80 to 140 kV and 50 to 330 mAs, with a collimated slice thickness ranging from 0.6 mm to 1.2 mm. MPR was performed following CT scanning, and axial, sagittal, and coronal reconstructions were stored as thick-section (1, 3, or 5 mm) image stacks. The reconstructed image stacks and the associated demographic information (sex assigned at birth and calendar age) were anonymized for this study. The final study sample consisted of 33 individuals and 61 long bones (Fig. 6 ). The collection, processing, and analysis of these CT scans as well as the use of associated demographic information were approved by the NTUH Research Ethics Committee (Institutional Review Board (IRB) case number: 201707024RINC). Informed consent was waived by the NTUH Research Ethics Committee due to the retrospective nature of anonymized clinical data. All methods were performed in accordance with the relevant guidelines and regulations. 3D-rendered “original” skeletal models were created from the CT stacks for each long bone following the segmentation protocol published by Stock et al. ( 2020 ). This protocol was developed with the Amira™ medical image visualization and analysis software (v.6.5.0, Thermo Fisher Scientific) on post-mortem multi-slice CT scans from a medical examiner’s office (Phillips Brilliance Big Bore 16-slice multi-detector scanner) that had comparable pre- and post-processing parameters to the current CT stack (512 by 512 pixel matrix, 1.0 mm slice thickness and 0.5 mm slice overlap). While only thick-slice CT MPR stacks were available for downstream segmentation and analysis, high-resolution volume renderings were also accessible via PACS. Accordingly, osteometric measurements were taken directly from these high-resolution volume renderings using a PACS DICOM viewer (Medixant, 2025 ). Although measurements from physical bone specimens would be ideal as the reference, such destructive processing is not feasible in clinical settings. Previously, Stull et al. ( 2014b ) demonstrated the mean differences in postcranial measurements from dry bone and CT volume rendering are generally less than 2%, supporting the use of PACS-based measurements as a practical reference. To create high-resolution 3D skeletal models from the same CT stacks, we implemented the algorithm in NiftyMIC ( https://github.com/gift-surg/NiftyMIC ). While summarized here, this algorithm is described in detail elsewhere (Ebner et al., 2018a ; Ebner et al., 2018b ; Ebner et al., 2020 ). The NiftyMIC workflow begins with automatic brain localization and extraction via two convolutional neural networks (CNNs), followed by slice-to-volume registration, aligning each slice to a common coordinate space using intensity- and boundary-based cost functions. The algorithm includes motion correction and bias field correction to account for slice misalignments and intensity inhomogeneities. The final high-resolution volume is reconstructed via a maximum a posteriori (MAP). In this study, we adapted NiftyMIC for skeletal data by inputting orthogonal stacks of thick-slice CT reconstructions (axial, sagittal, and coronal planes) that included each long bone. CT scans in DICOM formats were converted to NIfTI formats using the Amira™ software prior to input. The modular design of NiftyMIC allows us to bypass the machine learning-based, automatic segmentation step by performing manual, contour-based segmentations of the skeletal elements using ITK-SNAP (Yushkevich et al., 2006 ). These segmentations were used to create masks that defined the region of interest, allowing the pipeline to perform slice-to-volume registration and SRR focused on the target skeletal structure. The resulting high-resolution reconstructions were exported in NIfTI formats for downstream analysis. For each bone, two sets of 3D models were generated: one derived from the original low-resolution, thick-slice CT image stacks (“original”), and one from the reconstructed high-resolution volumes produced using the NiftyMIC pipeline (“reconstructed”). Linear osteometric measurements (Table 3), including diaphyseal length, proximal and distal breadth, and midshaft mediolateral (ML) breadth, were collected in the PACS DICOM viewer (Medixant, 2025 ) for PACS renderings and in the Amira™ software for the virtual models using the software’s measuring tool (0.01 mm resolution), following a published protocol for collecting osteometrics on virtual non-adult bone reconstructions (Stull & Corron, 2021 ; Stull et al., 2013 ). A total of 197 linear measurements (80 lengths and 117 breadths) were collected. All measurements were taken from the two sets of models per bone by two independent observers (A.D.Y. and L.K.C.). Intra-observer error was assessed by having each observer repeat all 197 measurements. To evaluate the reliability of the reconstruction processes and ensure that the pipeline did not introduce distortion, inter-rendering technical error of measurement (TEM) and relative technical error of measurement (%TEM) were calculated (Perini et al., 2005 ). Based on previous studies, we adapted acceptable ranges of < 2 mm for TEM and < 2% for %TEM when assessing measurement reliability (Colman et al., 2019 ; Langley et al., 2018 ; Stull et al., 2013 ). Bland-Altman plots were generated to visualize the agreement in osteometric measurements between PACS renderings and the two types of 3D models (Bland & Altman, 1986 ). Geometric comparison was performed exclusively between the “original” and “reconstructed” models in the 3D Slicer (Fedorov et al., 2012 ). Models were first aligned using the SlicerMorph platform extension (Rolfe et al., 2021 ), followed by quantification of geometric similarity using the signed Hausdorff distance (Huttenlocher et al., 1993 ) and Dice similarity coefficient (Dice, 1945 ). These metrics were computed using the Segment Comparison (Pinter et al., 2012 ), Model to Model Distance, and Mesh Statistics modules in 3D Slicer. The signed Hausdorff distance measures the directional deviation between the closest surface points on the models, with negative values indicating that the reconstructed surface lies within the reference surface, and positive values indicating it lies outside of the reference surface. The Dice similarity coefficient can range from 0 (no spatial overlap) to 1 (complete spatial overlap), providing a measure of volumetric agreement between models. Whereas the Hausdorff distance measures are reported in maximum, mean, and minimum values, the Dice similarity coefficient is one value per pair of models. For each pairwise comparison, a distance map was generated to show the perpendicular distance from the high-resolution “reconstructed” model to the corresponding low-resolution “original” models. Finally, to account for the potential ontogenetic and sex-related variation in bone morphology (Stull & Garvin, 2025 ; Stull et al., 2014a ), such as changes in size, curvature, and the degree of epiphyseal fusion that could influence reconstruction accuracy and surface deviation, linear regressions and ANOVA were performed to assess the effects of age, sex assigned at birth, bone type, and the age-bone interaction on minimum and maximum Hausdorff distance. All statistical analyses described above were performed in the R programming language (version 4.4.3, R Core Team, 2025 ). Data availability statement 3D model files supporting the findings of this study are openly available through the George Mason University Dataverse (doi: 10.13021/ORC2020/LFPH6M ). Declarations Funding This work is supported by the National Science Foundation (NSF BCS 1945797) and the Wenner-Gren Foundation for Anthropological Research (Dissertation Fieldwork Grant No. 9868). Author Contribution All authors contributed equally to the design of the study and revision of the manuscript. A.D.Y. acquired the funding, A.D.Y. and L.K.C. collected the data and contributed to data analysis and interpretation, and A.D.Y. drafted the manuscript. Data Availability 3D model files supporting the findings of this study are openly available through the George Mason University Dataverse (https://doi.org/10.13021/ORC2020/LFPH6M). References Albanese, J. Identified skeletal reference collections and the study of human variation. (Doctor of Philosophy), McMaster University, Hamilton, Ontario. (2003). Retrieved from http://hdl.handle.net/11375/7531 Alemán, I. et al. Brief communication: the Granada osteological collection of identified infants and young children. Am. J. Phys. Anthropol. 149 (4), 606–610 (2012). Bade, R., Haase, J. P. & Preim, B. Comparison of Fundamental Mesh Smoothing Algorithms for Medical Surface Models. Paper presented at the Simulation und Visualisierung, Magdeburg. (2006). Bland, J. M. & Altman, D. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 327 (8476), 307–310. 10.1016/S0140-6736(86)90837-8 (1986). Brough, A. L., Bennett, J., Morgan, B., Black, S. & Rutty, G. N. Anthropological Measurement of the Juvenile Clavicle Using Multi-Detector Computed Tomography—Affirming Reliability. J. Forensic Sci. 58 (4), 946–951. https://doi.org/10.1111/1556-4029.12126 (2013). Cardoso, H. F. V. Brief communication: The collection of identified human skeletons housed at the Bocage Museum (National Museum of Natural History), Lisbon, Portugal. Am. J. Phys. Anthropol. 129 (2), 173–176. https://doi.org/10.1002/ajpa.20228 (2006). Carew, R. M., Morgan, R. M. & Rando, C. A Preliminary Investigation into the Accuracy of 3D Modeling and 3D Printing in Forensic Anthropology Evidence Reconstruction. J. Forensic Sci. 64 (2), 342–352. https://doi.org/10.1111/1556-4029.13917 (2019). Christensen, A. M., Smith, M. A., Gleiber, D. S., Cunningham, D. L. & Wescott, D. J. The use of X-ray computed tomography technologies in forensic anthropology. Forensic Anthropol. 1 (2), 124–124. https://doi.org/10.5744/fa.2018.0013 (2018). Colman, K. L. et al. The geometrical precision of virtual bone models derived from clinical computed tomography data for forensic anthropology. Int. J. Legal Med. 131 (4), 1155–1163. 10.1007/s00414-017-1548-z (2017). Colman, K. L. et al. Virtual forensic anthropology: The accuracy of osteometric analysis of 3D bone models derived from clinical computed tomography (CT) scans. Forensic Sci. Int. 304 10.1016/j.forsciint.2019.109963 (2019). Copes, L. E., Lucas, L. M., Thostenson, J. O., Hoekstra, H. E. & Boyer, D. M. A collection of non-human primate computed tomography scans housed in MorphoSource, a repository for 3D data. Sci. Data . 3 (1), 160001. 10.1038/sdata.2016.1 (2016). Corron, L., Marchal, F., Condemi, S., Chaumoître, K. & Adalian, P. Evaluating the consistency, repeatability, and reproducibility of osteometric data on dry bone surfaces, scanned dry bone surfaces, and scanned bone surfaces obtained from living individuals. BMSAP 29 (1), 33–53. 10.1007/s13219-016-0172-7 (2017). Dice, L. R. Measures of the Amount of Ecologic Association Between Species. Ecology 26 (3), 297–302. 10.2307/1932409 (1945). Ebert, L. C. et al. A review of visualization techniques of post-mortem computed tomography data for forensic death investigations. Int. J. Legal Med. 135 (5), 1855–1867. 10.1007/s00414-021-02581-4 (2021). Ebner, M. et al. Volumetric reconstruction from printed films: Enabling 30 year longitudinal analysis in MR neuroimaging. Neuroimage 165 , 238–250. https://doi.org/10.1016/j.neuroimage.2017.09.056 (2018a). Ebner, M. et al. An Automated Localization, Segmentation and Reconstruction Framework for Fetal Brain MRI , Cham. (2018b). Ebner, M. et al. An automated framework for localization, segmentation and super-resolution reconstruction of fetal brain MRI. Neuroimage 206 , 116324. https://doi.org/10.1016/j.neuroimage.2019.116324 (2020). Edgar, H. J. H. et al. New Mexico Decedent Image Database . (2020). Fedorov, A. et al. 3D Slicer as an image computing platform for the Quantitative Imaging Network. Magn. Reson. Imaging . 30 (9), 1323–1341. https://doi.org/10.1016/j.mri.2012.05.001 (2012). Fischer, M. C. M. Database of segmentations and surface models of bones of the entire lower body created from cadaver CT scans. Sci. Data . 10 (1), 763. 10.1038/s41597-023-02669-z (2023). Franklin, D. et al. Concordance of traditional osteometric and volume-rendered MSCT interlandmark cranial measurements. Int. J. Legal Med. 127 (2), 505–520. 10.1007/s00414-012-0772-9 (2013). Franklin, D., Cardini, A., O’Higgins, P., Oxnard, C. E. & Dadour, I. Mandibular morphology as an indicator of human subadult age: geometric morphometric approaches. Forensic Sci. Med. Pathol. 4 (2), 91–99. 10.1007/s12024-007-9015-7 (2008). Frelat, M. A. & Mittereocker, P. Postnatal ontogeny of tibia and femur form in two human populations: a multivariate morphometric analysis. Am. J. Hum. Biology . 23 (6), 796–804. https://doi.org/10.1002/ajhb.21217 (2011). Fu, Y., Liu, S., Li, H. H. & Yang, D. Automatic and hierarchical segmentation of the human skeleton in CT images. Phys. Med. Biol. 62 (7), 2812. 10.1088/1361-6560/aa6055 (2017). Garvin, H. M. & Stock, M. K. The utility of advanced imaging in forensic anthropology. Acad. Forensic Pathol. 6 (3), 499–516. 10.23907/2016.050 (2016). Gascho, D., Thali, M. J. & Niemann, T. Post-mortem computed tomography: Technical principles and recommended parameter settings for high-resolution imaging. Med. Sci. Law . 58 (1), 70–82. 10.1177/0025802417747167 (2018). Geijer, M. & El-Khoury, G. Y. MDCT in the evaluation of skeletal trauma: principles, protocols, and clinical applications. Emerg. Radiol. 13 (1), 7–18. 10.1007/s10140-006-0509-5 (2006). Goldman, L. W. Principles of CT: Radiation Dose and Image Quality. J. Nucl. Med. Technol. 35 (4), 213–225. 10.2967/jnmt.106.037846 (2007). Gunz, P., Mitteroecker, P., Neubauer, S., Weber, G. W. & Bookstein, F. L. Principles for the virtual reconstruction of hominin crania. J. Hum. Evol. 57 (1), 48–62. https://doi.org/10.1016/j.jhevol.2009.04.004 (2009). Hong, E., Kwak, D. S. & Kim, I. B. Morphometric Evaluation of Detailed Asymmetry for the Proximal Humerus in Korean Population. Symmetry 13 (5), 862 (2021). Huseynov, A. et al. Developmental evidence for obstetric adaptation of the human female pelvis. Proc. Natl. Acad. Sci. U.S.A. 113 (19), 5227–5232. 10.1073/pnas.1517085113 (2016). Huttenlocher, D. P., Klanderman, G. A. & Rucklidge, W. J. Comparing images using the Hausdorff distance. IEEE Trans. Pattern Anal. Mach. Intell. 15 (9), 850–863. 10.1109/34.232073 (1993). Jeuthe, J. et al. SEMI-AUTOMATED 3D. Radiat. Prot. Dosimetry . 195 (3–4), 172–176. 10.1093/rpd/ncab073 (2021). Johnson, J., Alahi, A. & Fei-Fei, L. Perceptual Losses for Real-Time Style Transfer and Super-Resolution , Cham. (2016). Kistler, M., Bonaretti, S., Pfahrer, M., Niklaus, R. & Büchler, P. The Virtual Skeleton Database: An Open Access Repository for Biomedical Research and Collaboration. J. Med. Internet. Res. 15 (11), e245. 10.2196/jmir.2930 (2013). Langley, N. R. et al. Error quantification of osteometric data in forensic anthropology. Forensic Sci. Int. 287 , 183–189. https://doi.org/10.1016/j.forsciint.2018.04.004 (2018). Lasek, J. & Piórkowski, A. CT Scan Transformation from a Sharp to a Soft Reconstruction Kernel Using Filtering Techniques , Singapore. (2021). Lautenschlager, S. Reconstructing the past: methods and techniques for the digital restoration of fossils. Royal Soc. Open. Sci. 3 (10), 160342. 10.1098/rsos.160342 (2016). Ledig, C. et al. Photo-realistic single image super-resolution using a generative adversarial network. Paper presented at the Proceedings of the IEEE conference on computer vision and pattern recognition. (2017). Lee, K. H. et al. Managing the CT Data Explosion: Initial Experiences of Archiving Volumetric Datasets in a Mini-PACS. J. Digit. Imaging . 18 (3), 188–195. 10.1007/s10278-005-5163-z (2005). Li, Y., Sixou, B. & Peyrin, F. A Review of the Deep Learning Methods for Medical Images Super Resolution Problems. IRBM 42 (2), 120–133. https://doi.org/10.1016/j.irbm.2020.08.004 (2021). Lund, H., Gröndahl, K. & Gröndahl, H. G. Accuracy and precision of linear measurements in cone beam computed tomography Accuitomo® tomograms obtained with different reconstruction techniques. Dentomaxillofacial Radiol. 38 (6), 379–386. 10.1259/dmfr/15022357 (2014). Medixant. RadiAnt DICOM Viewer (Version 2025.2). (2025). Retrieved from https://www.radiantviewer.com Mopin, C., Chaumoître, K., Signoli, M. & Adalian, P. Developmental stability and environmental stress: A geometric morphometrics analysis of asymmetry in the human femur. Am. J. Phys. Anthropol. 167 (1), 144–160. https://doi.org/10.1002/ajpa.23613 (2018). Morimoto, N., Nakatsukasa, M., Ponce de León, M. S. & Zollikofer, C. P. E. Femoral ontogeny in humans and great apes and its implications for their last common ancestor. Scientific Reports, 8 (1), 1930. (2018). 10.1038/s41598-018-20410-4 Moustafa, M., Ebied, H. M., Helmy, A., Nazamy, T. M. & Tolba, M. F. Optimization Methods for Medical Image Super Resolution Reconstruction. In (eds Hassanien, A. E., Grosan, C. & Fahmy Tolba, M.) Applications of Intelligent Optimization in Biology and Medicine: Current Trends and Open Problems (135–157). Cham: Springer International Publishing. (2016). Noble, E. & Hawks, J. A Geometric Morphometric Study of Scapular Ontogeny in Modern Humans. Am. J. Biol. Anthropol. 187 (3), e70090. https://doi.org/10.1002/ajpa.70090 (2025). Park, J. et al. Computed tomography super-resolution using deep convolutional neural network. Phys. Med. Biol. 63 (14), 145011. 10.1088/1361-6560/aacdd4 (2018). Park, S. C., Park, M. K. & Kang, M. G. Super-resolution image reconstruction: a technical overview. IEEE. Signal. Process. Mag. 20 (3), 21–36. 10.1109/MSP.2003.1203207 (2003). Perini, T. A., Oliveira, G. L., d., Ornellas, J. S. & Oliveira, F. P. d. Cálculo do erro técnico de medição em antropometria. Revista Brasileira de Medicina do Esporte, 11 . (2005). Pinter, C., Lasso, A., Wang, A., Jaffray, D. & Fichtinger, G. SlicerRT: Radiation therapy research toolkit for 3D Slicer. Med. Phys. 39 (10), 6332–6338. https://doi.org/10.1118/1.4754659 (2012). Pujol, A., Rissech, C., Ventura, J., Badosa, J. & Turbón, D. Ontogeny of the female femur: geometric morphometric analysis applied on current living individuals of a Spanish population. J. Anat. 225 (3), 346–357. https://doi.org/10.1111/joa.12209 (2014). Pujol, A., Rissech, C., Ventura, J. & Turbón, D. Ontogeny of the male femur: Geometric morphometric analysis applied to a contemporary Spanish population. Am. J. Phys. Anthropol. 159 (1), 146–163. https://doi.org/10.1002/ajpa.22846 (2016). Qiu, D., Cheng, Y. & Wang, X. Medical image super-resolution reconstruction algorithms based on deep learning: A survey. Comput. Methods Programs Biomed. 238 , 107590. https://doi.org/10.1016/j.cmpb.2023.107590 (2023). R Core Team. R: A language and environment for statistical computing (R Foundation for Statistical Computing, 2025). Rathnayaka, K., Sahama, T., Schuetz, M. A. & Schmutz, B. Effects of CT image segmentation methods on the accuracy of long bone 3D reconstructions. Med. Eng. Phys. 33 (2), 226–233. https://doi.org/10.1016/j.medengphy.2010.10.002 (2011). Rolfe, S. et al. SlicerMorph: An open and extensible platform to retrieve, visualize and analyse 3D morphology. Methods Ecol. Evol. 12 (10), 1816–1825. https://doi.org/10.1111/2041-210X.13669 (2021). Sawdayee, H., Vaxman, A. & Bermano, A. H. Orex: Object reconstruction from planar cross-sections using neural fields. Paper presented at the Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. (2023). Scherf, H. Computed tomography in paleoanthropology — an overview. Archaeol. Anthropol. Sci. 5 (3), 205–214. 10.1007/s12520-013-0128-5 (2013). Scherf, H. & Tilgner, R. A new high-resolution computed tomography (CT) segmentation method for trabecular bone architectural analysis. Am. J. Phys. Anthropol. 140 (1), 39–51. https://doi.org/10.1002/ajpa.21033 (2009). Shanley, C., Obertová, Z. & Franklin, D. Dimensional accuracy of structured light scans and 3D prints of various human skeletal elements. Forensic Sci. Int. 361 , 112138. https://doi.org/10.1016/j.forsciint.2024.112138 (2024). Simmons-Ehrhardt, T. Open osteology: Medical imaging databases as skeletal collections. Forensic Imaging . 26 , 200462. https://doi.org/10.1016/j.fri.2021.200462 (2021). Stock, M. K. et al. The importance of processing procedures and threshold values in CT scan segmentation of skeletal elements: An example using the immature os coxa. Forensic Sci. Int. 309 10.1016/j.forsciint.2020.110232 (2020). Stull, K. E. & Corron, L. K. Subadult Virtual Anthropology Database (SVAD) Data Collection Protocol: Epiphyseal Fusion, Diaphyseal Dimensions, Dental Development Stages, Vertebral Neural Canal Dimensions 2nd edn (Zenodo, 2021). Stull, K. E. & Corron, L. K. The Subadult Virtual Anthropology Database (SVAD): An Accessible Repository of Contemporary Subadult Reference Data. Forensic Sci. 2 (1), 20–36. https://doi.org/10.3390/forensicsci2010003 (2022). Stull, K. E. & Garvin, H. M. A Visual Atlas of Skeletal Growth and Development 1st edn (Taylor & Francis Group, 2025). Stull, K. E., L'Abbé, E. N. & Ousley, S. D. Using multivariate adaptive regression splines to estimate subadult age from diaphyseal dimensions. Am. J. Phys. Anthropol. 154 (3), 376–386. 10.1002/ajpa.22522 (2014a). Stull, K. E., L'abbé, E. N. & Steiner, S. Measuring distortion of skeletal elements in Lodox Statscan-generated images. Clin. Anat. 26 (6), 780–786. https://doi.org/10.1002/ca.22225 (2013). Stull, K. E., Tise, M. L., Ali, Z. & Fowler, D. R. Accuracy and reliability of measurements obtained from computed tomography 3D volume rendered images. Forensic Sci. Int. 238 , 133–140. 10.1016/j.forsciint.2014.03.005 (2014b). Tang, Z., Xing, X., Wang, G. & Yang, G. Enhancing Super-Resolution Network Efficacy in CT Imaging: Cost-Effective Simulation of Training Data. IEEE Open. J. Eng. Med. Biology . 6 , 576–583. 10.1109/OJEMB.2025.3610160 (2025). Thomas, D. B., Hiscox, J. D., Dixon, B. J. & Potgieter, J. 3D scanning and printing skeletal tissues for anatomy education. J. Anat. 229 (3), 473–481. https://doi.org/10.1111/joa.12484 (2016). Tian, J. & Ma, K. K. A survey on super-resolution imaging. Signal. Image Video Process. 5 (3), 329–342. 10.1007/s11760-010-0204-6 (2011). Uldin, T. Virtual Anthropology – A Brief Review of the Literature and History of Computed Tomography. Forensic Sci. Res. 2 (4), 165–173. 10.1080/20961790.2017.1369621 (2017). Van den Broeck, J., Vereecke, E., Wirix-Speetjens, R. & Vander Sloten, J. Segmentation accuracy of long bones. Med. Eng. Phys. 36 (7), 949–953. https://doi.org/10.1016/j.medengphy.2014.03.016 (2014). Viðarsdóttir, U. S., O'higgins, P. & Stringer, C. A geometric morphometric study of regional differences in the ontogeny of the modern human facial skeleton. J. Anat. 201 , 211–229. https://doi.org/10.1046/j.1469-7580.2002.00092.x (2002). Weber, G. W. Virtual Anthropology. Am. J. Phys. Anthropol. 156 (S59), 22–42. https://doi.org/10.1002/ajpa.22658 (2015). Yoshinobu, T. et al. Data Management Solution for Large-Volume Computed Tomography in an Existing Picture Archiving and Communication System (PACS). J. Digit. Imaging . 24 (1), 107–113. 10.1007/s10278-009-9251-3 (2011). Yushkevich, P. A. et al. User-guided 3D active contour segmentation of anatomical structures: Significantly improved efficiency and reliability. Neuroimage 31 (3), 1116–1128. https://doi.org/10.1016/j.neuroimage.2006.01.015 (2006). Zhang, S. & Xia, Y. CT image reconstruction algorithms: A comprehensive survey. Concurrency Computation: Pract. Experience . 33 (8), e5506. https://doi.org/10.1002/cpe.5506 (2021). Zollikofer, C. P. E. & de Ponce, M. S. Visualizing patterns of craniofacial shape variation in Homo sapiens. Proceedings of the Royal Society of London. Series B: Biological Sciences, 269 (1493), 801–807. doi: (2002). 10.1098/rspb.2002.1960 Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Reviewers agreed at journal 13 Apr, 2026 Reviewers agreed at journal 24 Mar, 2026 Reviewers invited by journal 30 Nov, 2025 Editor assigned by journal 30 Nov, 2025 Editor invited by journal 28 Nov, 2025 Submission checks completed at journal 26 Nov, 2025 First submitted to journal 26 Nov, 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-8195787","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":563481837,"identity":"887ff555-0bfe-44a2-a416-04b2bea44a1f","order_by":0,"name":"An-Di Yim","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAr0lEQVRIiWNgGAWjYFACxgbmHxUQpgTxWhjOkKaFgYGZsY0ULfwzkls3F86rjTY4wHzwNg8xWiRuJLbdnrnteO6GA2zJ1kRpYQBqucG77RhQC4+ZNFFa5MFa5oC08H8jTosByGG8DTUgW9iI02J45mHbzRnHDuTOPMxmbDmHGC1yx9Of3fhQU5fbd7z54Y03xGhhEEgAkYeBsUOUchDgPwAi64hWPwpGwSgYBSMQAADrsDpA7uiHlAAAAABJRU5ErkJggg==","orcid":"","institution":"George Mason University","correspondingAuthor":true,"prefix":"","firstName":"An-Di","middleName":"","lastName":"Yim","suffix":""},{"id":563481838,"identity":"1f90537d-adf4-4dd8-bf87-ef09dac6192b","order_by":1,"name":"Louise K. Corron","email":"","orcid":"","institution":"University of Nevada Reno","correspondingAuthor":false,"prefix":"","firstName":"Louise","middleName":"K.","lastName":"Corron","suffix":""},{"id":563481839,"identity":"f5ebccd3-e71f-4551-8a66-1baeddf2a8f1","order_by":2,"name":"Kyra E. Stull","email":"","orcid":"","institution":"University of Nevada Reno","correspondingAuthor":false,"prefix":"","firstName":"Kyra","middleName":"E.","lastName":"Stull","suffix":""}],"badges":[],"createdAt":"2025-11-24 17:08:05","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8195787/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8195787/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":100373889,"identity":"1ab24e05-e6ad-4175-ad86-db16daebf41a","added_by":"auto","created_at":"2026-01-16 08:20:02","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":850534,"visible":true,"origin":"","legend":"","description":"","filename":"yimSRRclinicalCT112625.docx","url":"https://assets-eu.researchsquare.com/files/rs-8195787/v1/bb9b4c413d93306db84ed18a.docx"},{"id":100342138,"identity":"9b9cd65b-81e1-4ceb-b780-c241ec8b3df9","added_by":"auto","created_at":"2026-01-16 00:04:52","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5247,"visible":true,"origin":"","legend":"","description":"","filename":"0316d5473f4146ca9db13cbfea7501e9.json","url":"https://assets-eu.researchsquare.com/files/rs-8195787/v1/90cdebfa42e8359db5cdbb3a.json"},{"id":100373277,"identity":"d74296c3-02ef-42c8-ab0b-7e0148f1cf4b","added_by":"auto","created_at":"2026-01-16 08:13:59","extension":"xml","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":162365,"visible":true,"origin":"","legend":"","description":"","filename":"0316d5473f4146ca9db13cbfea7501e91enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8195787/v1/24d8268d3b64e8c681f83b2a.xml"},{"id":100342152,"identity":"ed13918d-fbd6-42fa-9c12-9039530cd46c","added_by":"auto","created_at":"2026-01-16 00:04:52","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":90576,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8195787/v1/4ae08d95c6e73ab1266d0988.png"},{"id":100373066,"identity":"fe968d11-215b-400b-9a5f-fca3bd4211be","added_by":"auto","created_at":"2026-01-16 08:13:34","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":35578,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8195787/v1/9146f54986cadbf996ff1c28.png"},{"id":100373663,"identity":"aebc5950-95f2-4f41-acb4-3c5160902abd","added_by":"auto","created_at":"2026-01-16 08:16:35","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":47933,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8195787/v1/aa4b35b977482525f31ad005.png"},{"id":100379679,"identity":"e252b4aa-db0d-426f-b13c-0ce8ec1b510b","added_by":"auto","created_at":"2026-01-16 09:16:30","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":50515,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8195787/v1/5735e81c0791a9416537b739.png"},{"id":100373139,"identity":"45b85dcb-26c1-4719-8498-18dd8da96094","added_by":"auto","created_at":"2026-01-16 08:13:43","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":63934,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-8195787/v1/392ae9cab688ff91dee0bcd3.png"},{"id":100373274,"identity":"6b264bae-38bb-4248-9ad4-12b679e011b6","added_by":"auto","created_at":"2026-01-16 08:13:58","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":18558,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-8195787/v1/661708bec0038430b52c5963.png"},{"id":100342154,"identity":"d46d0141-295e-4739-8f9b-e9e38c06d62b","added_by":"auto","created_at":"2026-01-16 00:04:52","extension":"xml","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":160862,"visible":true,"origin":"","legend":"","description":"","filename":"0316d5473f4146ca9db13cbfea7501e91structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8195787/v1/42ff19469c90d9490a34404f.xml"},{"id":100342153,"identity":"e00ded2f-bd79-4ecb-8c91-12658db35eb4","added_by":"auto","created_at":"2026-01-16 00:04:52","extension":"html","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":173469,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8195787/v1/aec989a5bf42ff12eec2be5a.html"},{"id":100342140,"identity":"19f02904-30b1-4aa9-9d06-da76790f478a","added_by":"auto","created_at":"2026-01-16 00:04:52","extension":"jpeg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":206460,"visible":true,"origin":"","legend":"\u003cp\u003eWorkflow illustrating the NiftyMIC-based super-resolution reconstruction (SRR) process for clinical CT data. Anisotropic, thick-slice CT image stacks were converted from DICOM to NIfTI format and manually masked to isolate the target bone. The NiftyMIC pipeline performed slice-to-volume registration and MAP-based SRR to combine the orthogonal stacks into an isotropic, high-resolution 3D volume. Segmented 3D skeletal models were generated from both the thick-slice CT data (“original”) and the SRR-reconstructed volume (“reconstructed”). A high-resolution PACS view is included for comparison purposes.\u003c/p\u003e","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8195787/v1/a9efb55010ebc2cbe3b2eafd.jpeg"},{"id":100373135,"identity":"b7c6a4db-7968-448e-9b32-c007ce73ba3f","added_by":"auto","created_at":"2026-01-16 08:13:42","extension":"jpeg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":132127,"visible":true,"origin":"","legend":"\u003cp\u003eBland-Altman plots showing inter-rendering agreement in linear osteometric measurements between PACS rendering and “original” and “reconstructed” models. Dashed lines indicate overall bias and the 95% range of bias (overall bias ± 2SD). For comparison between PACS rendering and “reconstructed” models, the overall bias is centered around zero with narrower 95% range indicating a higher level of consistency and no systematic size distortion introduced by reconstruction or registration.\u003c/p\u003e","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8195787/v1/7bad419c9ade7120a8a40865.jpeg"},{"id":100342141,"identity":"7fcffc0e-d9f2-4e02-bb32-bb05bedceb05","added_by":"auto","created_at":"2026-01-16 00:04:52","extension":"jpeg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":135496,"visible":true,"origin":"","legend":"\u003cp\u003eBox plots showing distributions of surface deviations between “original” and “reconstructed” models. Deviations were quantified using signed Hausdorff distances (A: mean, B: maximum, C: minimum; in mm) and Dice similarity coefficients (D). Each box represents the distribution of values for a given long bone type. Overall, mean Hausdorff distances fall within ±1 mm, indicating close geometric correspondence between models. Larger maximum and minimum deviations, particularly at the metaphyseal regions, likely result from incomplete separation of fusing epiphyses and localized surface depressions or gaps not fully resolved during segmentation and processing.\u003c/p\u003e","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8195787/v1/0c8b2d2ca689777b7c41d3de.jpeg"},{"id":100372696,"identity":"bd65da79-f321-49e2-93d4-c42f8a3c61c8","added_by":"auto","created_at":"2026-01-16 08:12:58","extension":"jpeg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":142601,"visible":true,"origin":"","legend":"\u003cp\u003eRepresentative examples of surface deviation heat maps for the unfused diaphyses of humerus (A), radius (B), ulna (C), femur (D), tibia (E), and fibula (F). Deviations are measured by the signed Hausdorff distance (mm), with red indicating outward deviation (“reconstructed” surface larger than “original”) and green indicating inward deviation (“reconstructed” surface smaller than “original”). Most notable deviations occur at the metaphyseal surfaces. Color bar scales differ across bones to optimize visualization of local variation.\u003c/p\u003e","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8195787/v1/2ffcfec0d466f2f1f10f94a8.jpeg"},{"id":100342142,"identity":"aa4c91b6-e97a-4f92-ac03-4ba1cee2d1a4","added_by":"auto","created_at":"2026-01-16 00:04:52","extension":"jpeg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":111210,"visible":true,"origin":"","legend":"\u003cp\u003eRepresentative examples of “reconstructed” models overlaid on “original” models for humerus (A), radius (B), ulna (C), femur (D), tibia (E), and fibula (F). “Original” models are shown in blue, while “reconstructed” models are shown as red point clouds. “Reconstructed” models generated through the super-resolution registration pipeline show smoother surfaces, particularly along curved diaphyseal and metaphyseal margins. In contrast, “original” models derived from thick-slice CT data often appear pixelated or voxelated.\u003c/p\u003e","description":"","filename":"floatimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8195787/v1/323d2ed98ef48b9c6f6dab9e.jpeg"},{"id":100342150,"identity":"6b957fec-58b4-4751-b7de-40b45d174872","added_by":"auto","created_at":"2026-01-16 00:04:52","extension":"jpeg","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":108080,"visible":true,"origin":"","legend":"\u003cp\u003eSummary of sample composition showing the age distribution and total number of long bones (humerus, radius, ulna, femur, tibia, and fibula).\u003c/p\u003e","description":"","filename":"floatimage6.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8195787/v1/7e1e36215a0a88173a11d7c6.jpeg"},{"id":100405670,"identity":"045b8af2-2497-4363-8d84-bd43198d0ea8","added_by":"auto","created_at":"2026-01-16 12:11:52","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1595871,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8195787/v1/ce34fb04-1212-4a14-8d12-d42ce3402a0e.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Adapting super-resolution reconstruction for skeletal analysis of clinical computed tomography data","fulltext":[{"header":"Introduction","content":"\u003cp\u003eIn recent years, advances in medical imaging and processing techniques have opened new avenues for biological anthropology and anatomy research. In particular, the use of computed tomography (CT) scans as data sources has become increasingly common in skeletal analysis literature (Simmons-Ehrhardt, \u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Uldin, \u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). CT imaging offers several benefits, including strong contrast between dense objects (such as bones) and surrounding soft tissues (Weber, \u003cspan citationid=\"CR76\" class=\"CitationRef\"\u003e2015\u003c/span\u003e), and the preservation of internal structures such as trabecular bone architecture, bony cavities, and their internal surfaces (Christensen et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Scherf, \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Scherf \u0026amp; Tilgner, \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2009\u003c/span\u003e). These capabilities extend the application of CT data beyond traditional osteometric analysis. CT scan-based datasets have been used to build virtual skeletal collections (e.g., Copes et al., \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Edgar et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Fischer, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Kistler et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Stull \u0026amp; Corron, \u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), reconstruct incomplete or fragmented specimens (e.g., Gunz et al., \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; Lautenschlager, \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2016\u003c/span\u003e), perform advanced morphometric shape analysis (e.g., Huseynov et al., \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Zollikofer \u0026amp; Ponce de Le\u0026oacute;n, 2002), as well as develop new methods and document case findings in forensic sciences (Christensen et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Garvin \u0026amp; Stock, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2016\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eAnatomical sciences typically rely on two primary sources of CT scan-based data. First, CT scans directly obtained from fossils, archaeological artifacts, cadavers, or dry skeletal elements can be conducted in research settings under controlled conditions. These scans are characterized by higher radiation doses (Goldman, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2007\u003c/span\u003e), minimal soft tissue, and stationary conditions to reduce noise and improve imaging quality (Gascho et al., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). However, this approach is limited by access to equipment and trained personnel. Second, clinical CT scans, generated in the hospital setting by radiologists or trained technicians as part of routine diagnostic practice, can be anonymized and repurposed to provide accessible and valuable retrospective data for clinical and non-clinical research. Previous studies (Colman et al., \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Colman et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) have demonstrated that 3D skeletal models reconstructed from clinical CT scans are accurate and precise representations of physical skeletal elements \u003cem\u003ein situ\u003c/em\u003e or \u003cem\u003eex situ\u003c/em\u003e (e.g., Brough et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Corron et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Franklin et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). Measurements taken from these rendered 3D models showed no significant difference from those collected directly from dry bones or dry bone surface scans, with an average intra-observer error of less than 1 mm and an average inter-observer error of less than 3 mm, considerably less than some errors produced via traditional osteometric analysis on dry bones (Langley et al., \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Surface deviations between 3D surfaces derived from clinical CT scans and dry bone surface scans were also minimal (Colman et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), supporting their use in subsequent morphometric and shape analyses.\u003c/p\u003e \u003cp\u003eIn most clinical settings, CT scans are acquired as isotropic or near-isotropic volumetric datasets (equal voxel dimensions in all three spatial direction) and subsequently reformatted and resliced at three anatomical planes (axial, sagittal, and coronal) through a process known as multiplanar reconstruction (MPR) (Geijer \u0026amp; El-Khoury, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2006\u003c/span\u003e). Radiologists and physicians routinely use MPR to assess anatomical structures and pathologies from multiple perspectives (Ebert et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). While sufficient for diagnostic purposes, the MRP images present challenges for researchers. These reconstructions typically consist of thicker slices (3 to 5 mm slice thickness) with anisotropic voxel dimensions (unequal voxel dimensions) and are therefore of lower resolution. For researchers, an MPR dataset may be high-resolution in one dimension but distorted in others, leading to inaccurate measurement and uneven surfaces in 3D reconstructions. Unfortunately, instead of the original high-resolution volumetric datasets, it is often these MPR images that are distributed and stored in the hospital picture archiving and communication system (PACS) (e.g., Lee et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2005\u003c/span\u003e; Yoshinobu et al., \u003cspan citationid=\"CR77\" class=\"CitationRef\"\u003e2011\u003c/span\u003e). As a result, researchers in human skeletal biology conducting retrospective data collection from clinical CT scans may be limited to these suboptimal, thick-slice reconstructions, which can result in poorly visible morphological features on CT slices and varying qualities of virtual reconstructions of skeletal elements after image segmentation, depending on the pre- and post-processing protocols and segmentation parameters set by the user (Stock et al., \u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eTo reduce voxelation and stepping artifacts, smoothing algorithms can be applied to fill the gaps between voxels via interpolation (Bade et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2006\u003c/span\u003e; Zhang \u0026amp; Xia, \u003cspan citationid=\"CR79\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), improving the visual appearance of bone surfaces (Lasek \u0026amp; Pi\u0026oacute;rkowski, \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Sawdayee et al., \u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Stock et al., \u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). However, when applied to thick-slice CT data, such smoothing rarely eliminates voxelation completely, and researchers risk over-smoothing, altering surface morphology and erasing anatomical features (Carew et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Johnson et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Ledig et al., \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2017\u003c/span\u003e), which can lead to inaccurate or unreliable osteological data obtained from these reconstructions (Stock et al., \u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). This inherent limitation highlights the need for reconstruction approaches that enhance resolution without relying entirely on smoothing algorithms.\u003c/p\u003e \u003cp\u003eSuper-resolution reconstruction (SRR) is a group of techniques that produces high-resolution images from low-resolution inputs (Park et al., \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2003\u003c/span\u003e). In medical imaging processing, SRR has been applied to modalities such as CT or magnetic resonance imaging (MRI) to enhance spatial resolution and reduce anisotropy without increasing radiation exposure or acquisition time. While recent SRR developments have largely focused on machine learning-based approaches (Li et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Qiu et al., \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Tang et al., \u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e2025\u003c/span\u003e), the need for large, high-resolution training data to perfect their algorithms makes them impractical for researchers working with datasets such as retrospective clinical scans. In contrast, model-based SRR approaches present as a regularized optimization solution to estimate the most probable high-resolution image consistent with a series of low-resolution slices. Because the orthogonal slice stacks represent images of the same object, their geometric relationships can be approximated by a displacement matrix describing spatial shifts and orientation differences. Each set of low-resolution images is modeled as a geometrically transformed observation of the underlying high-resolution images with added noise (Moustafa et al., \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Tian \u0026amp; Ma, \u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e2011\u003c/span\u003e). Solving the equation thus constitutes an inverse problem, in which the high-resolution volume is iteratively estimated by minimizing a cost function. The open-source NiftyMIC framework (Ebner et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2018a\u003c/span\u003e; Ebner et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2018b\u003c/span\u003e; Ebner et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2020\u003c/span\u003e), originally developed for neonatal brain MRI, implements such a model-based SRR pipeline to combine multiple orthogonal, low-resolution stacks into an isotropic, high-resolution reconstruction. The modular design of the pipeline and the need for an accessible and easily implementable tool to improve image resolution of archived clinical CT data in skeletal research makes it suitable for testing its applicability to the reconstruction of both 2D and 3D renderings of skeletal elements from low-resolution CT scans.\u003c/p\u003e \u003cp\u003eThis study investigates the utility of the model-based SRR framework implemented in NiftyMIC to reconstruct volumetric skeletal data from multiple anisotropic low-resolution thick-slice CT scan stacks (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). We compare osteometric measurements across three model types: diagnostic PACS renderings, thick-slice reconstructions (\u0026ldquo;original\u0026rdquo;), and SRR-reconstructed models (\u0026ldquo;reconstructed\u0026rdquo;) to assess metric accuracy and conduct geometric comparisons between the thick-slice models and SRR-reconstructed models to evaluate the degree of geometric consistency.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e summarizes the measurement agreement results. Intra-observer error rates were minimal for both observers (Observer 1: TEM\u0026thinsp;=\u0026thinsp;0.610 mm, %TEM\u0026thinsp;=\u0026thinsp;0.758%; Observer 2: TEM\u0026thinsp;=\u0026thinsp;0.479 mm, %TEM\u0026thinsp;=\u0026thinsp;0.505%). Inter-observer error was 1.243 mm (1.603%) for the \u0026ldquo;original\u0026rdquo; models and 0.439 mm (0.517%) for the \u0026ldquo;reconstructed\u0026rdquo; models, indicating slightly higher measurement consistency for the \u0026ldquo;reconstructed\u0026rdquo; models. Inter-rendering comparisons further demonstrated strong measurement reliability. The TEM between the 3D models was 1.357 mm (1.702%) for Observer 1 and 1.279 mm (1.619%) for Observer 2. Figure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e shows the Bland-Altman plots showing inter-rendering differences in paired linear measurements. Measurement deviations between PACS renderings and \u0026ldquo;reconstructed\u0026rdquo; models are tightly centered around zero with narrower 95% limits of agreement. In contrast, the \u0026ldquo;original\u0026rdquo; models display slightly wider dispersion. The overall agreement in linear measurements between PACS renderings and \u0026ldquo;reconstructed\u0026rdquo; models (TEM\u0026thinsp;=\u0026thinsp;0.828 mm, %TEM\u0026thinsp;=\u0026thinsp;0.996%) is also better than between PACS renderings and \u0026ldquo;original\u0026rdquo; models (TEM\u0026thinsp;=\u0026thinsp;1.411 mm, %TEM\u0026thinsp;=\u0026thinsp;1.737%). Collectively, these results indicate that SRR reconstruction improves linear measurement accuracy relative to the thick-slice models while maintaining high intra- and inter-observer reliability.\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\u003eTechnical error of measurement (TEM) and relative technical error of measurement (%TEM) for inter-rendering, intra-observer, and inter-observer comparisons of linear osteometric measurements. Intra-observer error rates were minimal for both observers, while inter-observer and inter-rendering errors remained low.\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=\"left\" 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=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eInter-rendering comparison\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"4\" nameend=\"c5\" namest=\"c2\"\u003e \u003cp\u003eObserver 1\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"3\" nameend=\"c8\" namest=\"c6\"\u003e \u003cp\u003eObserver 2\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003eTEM (mm)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e%TEM\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003eTEM (mm)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003e%TEM\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOriginal vs. reconstructed\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003e1.357\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.702\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003e1.279\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1.619\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eIntra-observer comparison\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"4\" nameend=\"c5\" namest=\"c2\"\u003e \u003cp\u003e\u003cb\u003eObserver 1\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c8\" namest=\"c6\"\u003e \u003cp\u003e\u003cb\u003eObserver 2\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\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003e\u003cb\u003eTEM (mm)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e%TEM\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003e\u003cb\u003eTEM (mm)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e\u003cb\u003e%TEM\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\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003e0.610\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.758\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003e0.479\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.505\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eInter-observer comparison\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c5\" namest=\"c3\"\u003e \u003cp\u003e\u003cb\u003eOriginal\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c8\" namest=\"c6\"\u003e \u003cp\u003e\u003cb\u003eReconstructed\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003eTEM (mm)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e \u003cp\u003e\u003cb\u003e%TEM\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003eTEM (mm)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c8\" namest=\"c7\"\u003e \u003cp\u003e\u003cb\u003e%TEM\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.243\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e \u003cp\u003e1.603\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.439\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c8\" namest=\"c7\"\u003e \u003cp\u003e0.517\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\u003e \u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e shows the distributions of surface deviations between the \u0026ldquo;original\u0026rdquo; and \u0026ldquo;reconstructed\u0026rdquo; models, as measured by the signed Hausdorff distance. Generally, an average Hausdorff distance close to zero indicates a high level of agreement in model geometry. On average, the \u0026ldquo;reconstructed\u0026rdquo; model surfaces deviate between \u0026minus;\u0026thinsp;1 mm to 1 mm from the \u0026ldquo;original\u0026rdquo; model surfaces, indicating close geometric correspondence. However, more substantial local deviations were observed in some elements, with minimum and maximum Hausdorff distances exceeding \u0026minus;\u0026thinsp;10 mm in isolated regions. These larger deviations likely reflect partial separation of fusing epiphyses and localized surface depressions or gaps that were not fully resolved during segmentation and processing. Figure\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e also shows the distributions of the Dice similarity coefficient. Overall, the Dice similarity coefficients (0.682 to 0.978) suggest that the two model types have good geometric agreement.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eRepresentative examples of surface deviations are shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e. For each long bone, one example was selected to provide a visual representation of surface similarity using heat maps of signed Hausdorff distance. The models presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e are those with the greatest geometric discordance (lowest Dice similarity coefficients). The most pronounced deviations are generally localized at the metaphyseal surfaces at the bone ends, while diaphyseal surfaces generally show high levels of geometric agreement. ANOVA of the linear regression models confirmed that age significantly affects both the minimum and maximum surface deviations (signed Hausdorff distance) across all comparisons (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001 for both measures, Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). Regression coefficients showed that with increasing age, minimum deviation decreases (-0.495 mm per year) while maximum deviation increases (0.456 mm per year). Together, these results suggest, as individuals age and epiphyseal fusion progresses, the increasing difficulty in distinguishing fusing epiphyses from diaphyses introduces greater error and thus larger surface deviations, particularly at the metaphyseal surfaces.\u003c/p\u003e \u003cp\u003e \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\u003eTest of the effects of sex assigned at birth, age, and bone on the signed Hausdorff distance across all comparisons.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\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=\"left\" 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=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEffect term\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003edf\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSum of squares\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eMean square\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eF\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cem\u003ep\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colspan=\"6\" nameend=\"c6\" namest=\"c1\"\u003e \u003cp\u003eMinimum Hausdorff distance\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e306.42\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e306.42\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e28.18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001*\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBiological sex\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e6.46\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e6.46\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.59\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.4420\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBone\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e363.57\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e72.71\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e6.69\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001*\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge x bone\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e46.15\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e9.23\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e9,85\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.5169\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"6\" nameend=\"c6\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eMaximum Hausdorff distance\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e698.11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e698.11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e88.36\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001*\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBiological sex\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.57\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.57\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.62\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.1567\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBone\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e64.08\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e12.82\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.07\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.7888\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge x bone\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e140.25\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e28.05\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3.55\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.0044\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\u003e \u003c/p\u003e \u003cp\u003eThe aligned \u0026ldquo;original\u0026rdquo; and \u0026ldquo;reconstructed\u0026rdquo; models corresponding to the examples shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e are presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e. \u0026ldquo;Original\u0026rdquo; models (shown in blue) are used as the reference, overlaid with \u0026ldquo;reconstructed\u0026rdquo; models (shown as red point clouds) to allow direct comparison. Qualitatively, the \u0026ldquo;reconstructed\u0026rdquo; models show smoother surfaces, particularly along curved surfaces of the diaphyses and at the metaphyseal surfaces. In contrast, the \u0026ldquo;original\u0026rdquo; models derived from thick-slice CT image stacks often exhibit visible voxelation and stepping artifacts. These visual comparisons highlight the advantages of the SRR framework in improving surface quality and recovering high-resolution skeletal morphology from clinically acquired low-resolution CT datasets.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this study, we evaluated the applicability of a model-based SRR pipeline to reconstruct high-resolution volumetric skeletal data from multiple stacks of low-resolution, thick-slice clinical CT scans using a non-adult sample. The results showed that the reconstructed volumes can be effectively segmented to produce 3D skeletal models that showed greater osteometric linear measurement accuracy than the thick-slice models, as demonstrated by lower technical error of measurement. Geometric comparisons confirmed that the overall geometry of the \u0026ldquo;reconstructed\u0026rdquo; models was broadly similar to the \u0026ldquo;original\u0026rdquo; models. Consistent with previous findings (Fu et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Rathnayaka et al., \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Van den Broeck et al., \u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e2014\u003c/span\u003e), localized surface deviations were most pronounced in the metaphyseal surfaces and epiphyseal regions, likely reflecting incomplete separation of fusing epiphyses or residual segmentation artifacts such as incompletely filled gaps in areas of thin cortical or sparse trabecular bone (Jeuthe et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). While these deviations may introduce localized errors in region-specific or landmark-based analyses, the overall geometry of the reconstructions is well-preserved. Qualitative visual assessments further showed that the \u0026ldquo;reconstructed\u0026rdquo; models exhibit smoother surfaces, particularly along curved diaphyseal regions and at metaphyseal or epiphyseal regions. These findings suggest that the NiftyMIC pipeline, with appropriate preprocessing steps, is a promising tool for enhancing 3D skeletal model quality derived from suboptimal clinical CT scan data.\u003c/p\u003e \u003cp\u003eThe \u0026ldquo;original\u0026rdquo; models, derived from thick-slice and lower-resolution clinical CT data, often exhibited surface artifacts such as voxelation and visible stepping between slices (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e). These artifacts are results of the discrete sampling (slicing) of a continuous anatomical structure (i.e., bone), a limitation inherent to medical image reconstructions. While over-smoothing resulting from the reconstruction pipeline may be a concern, over-smoothed surfaces tend to be systematically smaller due to the rounding of surface contours and the loss of fine details (Johnson et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Ledig et al., \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Stock et al., \u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). No such systematic inward deviations were observed in our results. Instead, the overall deviations of the \u0026ldquo;reconstructed\u0026rdquo; models were between \u0026minus;\u0026thinsp;1 mm to 1 mm from the \u0026ldquo;original\u0026rdquo; model surface, showing no bias in the direction of surface deviations (inward or outward). These findings suggest that the visible improvements in model smoothness and surface quality are the results of improved resolution, and that the SRR algorithm mathematically estimates plausible surface geometry in regions where data were discontinuous or missing in the original stacks, producing models that are smoother and more geometrically consistent.\u003c/p\u003e \u003cp\u003eThe present study used a non-adult sample. This is particularly relevant to human skeletal variation research that relies on geometric morphometric approaches to obtain data for analysis. As demonstrated in previous studies, variation of different skeletal elements (e.g., Franklin et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2008\u003c/span\u003e; Huseynov et al., \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Noble \u0026amp; Hawks, \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Vi\u0026eth;arsd\u0026oacute;ttir et al., \u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e2002\u003c/span\u003e) during growth that persist as morphological differences in size and shape and into adulthood are not only general, but also localized to specific areas of bone, such as grooves, curves, and other anatomical reliefs. These characteristics should therefore be preserved as much as possible in virtual 3D renderings of juvenile bones for analyses to be reliable and accurate. The ability of the SRR pipeline to produce anatomically realistic, smoothly contoured skeletal surfaces may therefore be advantageous when access to dry bone samples is limited, as is the case for non-adult skeletons in anatomical collections (Albanese, \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2003\u003c/span\u003e; Alem\u0026aacute;n et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Cardoso, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2006\u003c/span\u003e; Stull \u0026amp; Corron, \u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe postcranial elements used in this study also demonstrated the broad applicability of the SRR framework. Considering the majority of growth-related changes in the lower limb skeleton are related to the changes in the epiphyses or the relative shape of the diaphyses and epiphyses (Frelat \u0026amp; Mittereocker, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Morimoto et al., \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Pujol et al., \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Pujol et al., \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2016\u003c/span\u003e), a pattern likely shared by upper limb skeletons. Asymmetries in skeletal morphology as a result of environmental stress are also concentrated near bone ends, as shown in previous studies (Hong et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Mopin et al., \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). The results of this study showed that the most pronounced deviations across all model types were localized at the metaphyseal surfaces and epiphyseal regions, presumably because the improved image quality led to clearer separation of the unfused and fusing epiphyses during segmentation and better preservation of these regions in the \u0026ldquo;reconstructed\u0026rdquo; models. Preserving the geometry of intricate regions such as metaphyseal surfaces and epiphyses is crucial for geometric morphometric analyses of growth and development. Importantly, similar problems with epiphyses were observed in fused elements. Thus, while this study focused on non-adult skeletal elements, the SRR framework presented here may also be beneficial for analyses of adult bones by improving the resolution of features at bone ends.\u003c/p\u003e \u003cp\u003eOne limitation of the current study is the lack of \u0026ldquo;gold standards\u0026rdquo; (i.e., dry bones or their direct surface scans). Nevertheless, the use of osteometric measurements from PACS renderings serve as practical references since standardized osteometric data are consistent with their dry bone counterparts across 1:1 scaled medical imaging modalities (Lund et al., \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Stull et al., \u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e2014b\u003c/span\u003e). Importantly, comparisons showed improved measurement consistency and error rates for the \u0026ldquo;reconstructed\u0026rdquo; models compared to the \u0026ldquo;original\u0026rdquo; models. Based on the results presented here, it is clear that the SRR-based reconstruction framework is highly advantageous in generating high-quality 3D models from archived clinical CT scans for research and for teaching, as the high quality and likeness of the rendered surfaces could be 3D printed and used as a substitute for dry bones in classrooms where logistic constraints or ethical concerns regarding the use of dry skeletal remains can be a limiting factor (Carew et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Shanley et al., \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Thomas et al., \u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e2016\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eAnother limitation of this study is the relatively small number of samples for each type of long bone. Future work using larger adult samples with fully fused epiphyses would allow for a more controlled evaluation of reconstruction accuracy, and subsequent testing across age groups could then clarify the broader applicability of this approach. In addition, a larger scale study focusing on comparing SRR-based reconstructions to models derived from high-resolution, isotropic thin-slice CT scans or dry bone surface scans could provide more robust validation (Li et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Park et al., \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Tang et al., \u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Despite this, improvements in surface quality and consistency are an important step toward the use of suboptimal clinical imaging for research.\u003c/p\u003e \u003cp\u003eIn summary, this study showed that the SRR framework implemented in NiftyMIC can be successfully adapted to generate high-quality 3D skeletal models from suboptimal, thick-slice clinical CT data. The \u0026ldquo;reconstructed\u0026rdquo; models show consistent osteometric measurements, improved surface smoothness, and clearer anatomical definition compared to models segmented directly from thick-slice scans. These proof-of-concept findings highlight SRR-based image registration as a practical approach for improving the utility of archived clinical imaging datasets, especially when access to high-resolution or physical skeletal samples is limited.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003eArchived clinical CT scans of long bones (humerus, radius, ulna, femur, tibia, or fibula) from individuals aged 0 to 16 years old of male or female sex assigned at birth were collected from National Taiwan University Hospital (NTUH) by a computerized search of the hospital PACS. Individuals were excluded if their CT scans did not include a complete long bone or if the long bone exhibited visible fractures or anomalies. The CT scans were originally performed between 2006 and 2017 as part of clinical diagnostic examinations using either a 256-slice or a 128-slice multidetector CT scanner. Imaging parameters varied across cases due to differences in clinical protocols but generally ranged from 80 to 140 kV and 50 to 330 mAs, with a collimated slice thickness ranging from 0.6 mm to 1.2 mm. MPR was performed following CT scanning, and axial, sagittal, and coronal reconstructions were stored as thick-section (1, 3, or 5 mm) image stacks. The reconstructed image stacks and the associated demographic information (sex assigned at birth and calendar age) were anonymized for this study. The final study sample consisted of 33 individuals and 61 long bones (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e). The collection, processing, and analysis of these CT scans as well as the use of associated demographic information were approved by the NTUH Research Ethics Committee (Institutional Review Board (IRB) case number: 201707024RINC). Informed consent was waived by the NTUH Research Ethics Committee due to the retrospective nature of anonymized clinical data. All methods were performed in accordance with the relevant guidelines and regulations.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e3D-rendered \u0026ldquo;original\u0026rdquo; skeletal models were created from the CT stacks for each long bone following the segmentation protocol published by Stock et al. (\u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). This protocol was developed with the Amira\u0026trade; medical image visualization and analysis software (v.6.5.0, Thermo Fisher Scientific) on post-mortem multi-slice CT scans from a medical examiner\u0026rsquo;s office (Phillips Brilliance Big Bore 16-slice multi-detector scanner) that had comparable pre- and post-processing parameters to the current CT stack (512 by 512 pixel matrix, 1.0 mm slice thickness and 0.5 mm slice overlap). While only thick-slice CT MPR stacks were available for downstream segmentation and analysis, high-resolution volume renderings were also accessible via PACS. Accordingly, osteometric measurements were taken directly from these high-resolution volume renderings using a PACS DICOM viewer (Medixant, \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Although measurements from physical bone specimens would be ideal as the reference, such destructive processing is not feasible in clinical settings. Previously, Stull et al. (\u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e2014b\u003c/span\u003e) demonstrated the mean differences in postcranial measurements from dry bone and CT volume rendering are generally less than 2%, supporting the use of PACS-based measurements as a practical reference.\u003c/p\u003e \u003cp\u003eTo create high-resolution 3D skeletal models from the same CT stacks, we implemented the algorithm in NiftyMIC (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/gift-surg/NiftyMIC\u003c/span\u003e\u003cspan address=\"https://github.com/gift-surg/NiftyMIC\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). While summarized here, this algorithm is described in detail elsewhere (Ebner et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2018a\u003c/span\u003e; Ebner et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2018b\u003c/span\u003e; Ebner et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). The NiftyMIC workflow begins with automatic brain localization and extraction via two convolutional neural networks (CNNs), followed by slice-to-volume registration, aligning each slice to a common coordinate space using intensity- and boundary-based cost functions. The algorithm includes motion correction and bias field correction to account for slice misalignments and intensity inhomogeneities. The final high-resolution volume is reconstructed via a maximum a posteriori (MAP).\u003c/p\u003e \u003cp\u003eIn this study, we adapted NiftyMIC for skeletal data by inputting orthogonal stacks of thick-slice CT reconstructions (axial, sagittal, and coronal planes) that included each long bone. CT scans in DICOM formats were converted to NIfTI formats using the Amira\u0026trade; software prior to input. The modular design of NiftyMIC allows us to bypass the machine learning-based, automatic segmentation step by performing manual, contour-based segmentations of the skeletal elements using ITK-SNAP (Yushkevich et al., \u003cspan citationid=\"CR78\" class=\"CitationRef\"\u003e2006\u003c/span\u003e). These segmentations were used to create masks that defined the region of interest, allowing the pipeline to perform slice-to-volume registration and SRR focused on the target skeletal structure. The resulting high-resolution reconstructions were exported in NIfTI formats for downstream analysis.\u003c/p\u003e \u003cp\u003eFor each bone, two sets of 3D models were generated: one derived from the original low-resolution, thick-slice CT image stacks (\u0026ldquo;original\u0026rdquo;), and one from the reconstructed high-resolution volumes produced using the NiftyMIC pipeline (\u0026ldquo;reconstructed\u0026rdquo;).\u003c/p\u003e \u003cp\u003eLinear osteometric measurements (Table\u0026nbsp;3), including diaphyseal length, proximal and distal breadth, and midshaft mediolateral (ML) breadth, were collected in the PACS DICOM viewer (Medixant, \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) for PACS renderings and in the Amira\u0026trade; software for the virtual models using the software\u0026rsquo;s measuring tool (0.01 mm resolution), following a published protocol for collecting osteometrics on virtual non-adult bone reconstructions (Stull \u0026amp; Corron, \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Stull et al., \u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). A total of 197 linear measurements (80 lengths and 117 breadths) were collected. All measurements were taken from the two sets of models per bone by two independent observers (A.D.Y. and L.K.C.). Intra-observer error was assessed by having each observer repeat all 197 measurements. To evaluate the reliability of the reconstruction processes and ensure that the pipeline did not introduce distortion, inter-rendering technical error of measurement (TEM) and relative technical error of measurement (%TEM) were calculated (Perini et al., \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2005\u003c/span\u003e). Based on previous studies, we adapted acceptable ranges of \u0026lt;\u0026thinsp;2 mm for TEM and \u0026lt;\u0026thinsp;2% for %TEM when assessing measurement reliability (Colman et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Langley et al., \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Stull et al., \u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). Bland-Altman plots were generated to visualize the agreement in osteometric measurements between PACS renderings and the two types of 3D models (Bland \u0026amp; Altman, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e1986\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eGeometric comparison was performed exclusively between the \u0026ldquo;original\u0026rdquo; and \u0026ldquo;reconstructed\u0026rdquo; models in the 3D Slicer (Fedorov et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). Models were first aligned using the SlicerMorph platform extension (Rolfe et al., \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), followed by quantification of geometric similarity using the signed Hausdorff distance (Huttenlocher et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e1993\u003c/span\u003e) and Dice similarity coefficient (Dice, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e1945\u003c/span\u003e). These metrics were computed using the Segment Comparison (Pinter et al., \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e2012\u003c/span\u003e), Model to Model Distance, and Mesh Statistics modules in 3D Slicer. The signed Hausdorff distance measures the directional deviation between the closest surface points on the models, with negative values indicating that the reconstructed surface lies within the reference surface, and positive values indicating it lies outside of the reference surface. The Dice similarity coefficient can range from 0 (no spatial overlap) to 1 (complete spatial overlap), providing a measure of volumetric agreement between models. Whereas the Hausdorff distance measures are reported in maximum, mean, and minimum values, the Dice similarity coefficient is one value per pair of models. For each pairwise comparison, a distance map was generated to show the perpendicular distance from the high-resolution \u0026ldquo;reconstructed\u0026rdquo; model to the corresponding low-resolution \u0026ldquo;original\u0026rdquo; models. Finally, to account for the potential ontogenetic and sex-related variation in bone morphology (Stull \u0026amp; Garvin, \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Stull et al., \u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e2014a\u003c/span\u003e), such as changes in size, curvature, and the degree of epiphyseal fusion that could influence reconstruction accuracy and surface deviation, linear regressions and ANOVA were performed to assess the effects of age, sex assigned at birth, bone type, and the age-bone interaction on minimum and maximum Hausdorff distance. All statistical analyses described above were performed in the R programming language (version 4.4.3, R Core Team, \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e\n\u003ch3\u003eData availability statement\u003c/h3\u003e\n\u003cp\u003e3D model files supporting the findings of this study are openly available through the George Mason University Dataverse (doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.13021/ORC2020/LFPH6M\u003c/span\u003e\u003cspan address=\"10.13021/ORC2020/LFPH6M\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e).\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eFunding\u003c/h2\u003e \u003cp\u003eThis work is supported by the National Science Foundation (NSF BCS 1945797) and the Wenner-Gren Foundation for Anthropological Research (Dissertation Fieldwork Grant No. 9868).\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eAll authors contributed equally to the design of the study and revision of the manuscript. A.D.Y. acquired the funding, A.D.Y. and L.K.C. collected the data and contributed to data analysis and interpretation, and A.D.Y. drafted the manuscript.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003e3D model files supporting the findings of this study are openly available through the George Mason University Dataverse (https://doi.org/10.13021/ORC2020/LFPH6M).\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAlbanese, J. \u003cem\u003eIdentified skeletal reference collections and the study of human variation.\u003c/em\u003e (Doctor of Philosophy), McMaster University, Hamilton, Ontario. (2003). Retrieved from \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://hdl.handle.net/11375/7531\u003c/span\u003e\u003cspan address=\"http://hdl.handle.net/11375/7531\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlem\u0026aacute;n, I. et al. Brief communication: the Granada osteological collection of identified infants and young children. \u003cem\u003eAm. J. Phys. Anthropol.\u003c/em\u003e \u003cb\u003e149\u003c/b\u003e (4), 606\u0026ndash;610 (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBade, R., Haase, J. P. \u0026amp; Preim, B. \u003cem\u003eComparison of Fundamental Mesh Smoothing Algorithms for Medical Surface Models.\u003c/em\u003e Paper presented at the Simulation und Visualisierung, Magdeburg. (2006).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBland, J. M. \u0026amp; Altman, D. Statistical methods for assessing agreement between two methods of clinical measurement. \u003cem\u003eLancet\u003c/em\u003e \u003cb\u003e327\u003c/b\u003e (8476), 307\u0026ndash;310. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/S0140-6736(86)90837-8\u003c/span\u003e\u003cspan address=\"10.1016/S0140-6736(86)90837-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (1986).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBrough, A. L., Bennett, J., Morgan, B., Black, S. \u0026amp; Rutty, G. N. Anthropological Measurement of the Juvenile Clavicle Using Multi-Detector Computed Tomography\u0026mdash;Affirming Reliability. \u003cem\u003eJ. Forensic Sci.\u003c/em\u003e \u003cb\u003e58\u003c/b\u003e (4), 946\u0026ndash;951. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/1556-4029.12126\u003c/span\u003e\u003cspan address=\"10.1111/1556-4029.12126\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCardoso, H. F. V. Brief communication: The collection of identified human skeletons housed at the Bocage Museum (National Museum of Natural History), Lisbon, Portugal. \u003cem\u003eAm. J. Phys. Anthropol.\u003c/em\u003e \u003cb\u003e129\u003c/b\u003e (2), 173\u0026ndash;176. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/ajpa.20228\u003c/span\u003e\u003cspan address=\"10.1002/ajpa.20228\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2006).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCarew, R. M., Morgan, R. M. \u0026amp; Rando, C. A Preliminary Investigation into the Accuracy of 3D Modeling and 3D Printing in Forensic Anthropology Evidence Reconstruction. \u003cem\u003eJ. Forensic Sci.\u003c/em\u003e \u003cb\u003e64\u003c/b\u003e (2), 342\u0026ndash;352. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/1556-4029.13917\u003c/span\u003e\u003cspan address=\"10.1111/1556-4029.13917\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChristensen, A. M., Smith, M. A., Gleiber, D. S., Cunningham, D. L. \u0026amp; Wescott, D. J. The use of X-ray computed tomography technologies in forensic anthropology. \u003cem\u003eForensic Anthropol.\u003c/em\u003e \u003cb\u003e1\u003c/b\u003e (2), 124\u0026ndash;124. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.5744/fa.2018.0013\u003c/span\u003e\u003cspan address=\"10.5744/fa.2018.0013\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eColman, K. L. et al. The geometrical precision of virtual bone models derived from clinical computed tomography data for forensic anthropology. \u003cem\u003eInt. J. Legal Med.\u003c/em\u003e \u003cb\u003e131\u003c/b\u003e (4), 1155\u0026ndash;1163. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s00414-017-1548-z\u003c/span\u003e\u003cspan address=\"10.1007/s00414-017-1548-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eColman, K. L. et al. Virtual forensic anthropology: The accuracy of osteometric analysis of 3D bone models derived from clinical computed tomography (CT) scans. \u003cem\u003eForensic Sci. Int.\u003c/em\u003e \u003cb\u003e304\u003c/b\u003e \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.forsciint.2019.109963\u003c/span\u003e\u003cspan address=\"10.1016/j.forsciint.2019.109963\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCopes, L. E., Lucas, L. M., Thostenson, J. O., Hoekstra, H. E. \u0026amp; Boyer, D. M. A collection of non-human primate computed tomography scans housed in MorphoSource, a repository for 3D data. \u003cem\u003eSci. Data\u003c/em\u003e. \u003cb\u003e3\u003c/b\u003e (1), 160001. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/sdata.2016.1\u003c/span\u003e\u003cspan address=\"10.1038/sdata.2016.1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCorron, L., Marchal, F., Condemi, S., Chaumo\u0026icirc;tre, K. \u0026amp; Adalian, P. Evaluating the consistency, repeatability, and reproducibility of osteometric data on dry bone surfaces, scanned dry bone surfaces, and scanned bone surfaces obtained from living individuals. \u003cem\u003eBMSAP\u003c/em\u003e \u003cb\u003e29\u003c/b\u003e (1), 33\u0026ndash;53. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s13219-016-0172-7\u003c/span\u003e\u003cspan address=\"10.1007/s13219-016-0172-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDice, L. R. Measures of the Amount of Ecologic Association Between Species. \u003cem\u003eEcology\u003c/em\u003e \u003cb\u003e26\u003c/b\u003e (3), 297\u0026ndash;302. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.2307/1932409\u003c/span\u003e\u003cspan address=\"10.2307/1932409\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (1945).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEbert, L. C. et al. A review of visualization techniques of post-mortem computed tomography data for forensic death investigations. \u003cem\u003eInt. J. Legal Med.\u003c/em\u003e \u003cb\u003e135\u003c/b\u003e (5), 1855\u0026ndash;1867. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s00414-021-02581-4\u003c/span\u003e\u003cspan address=\"10.1007/s00414-021-02581-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEbner, M. et al. Volumetric reconstruction from printed films: Enabling 30 year longitudinal analysis in MR neuroimaging. \u003cem\u003eNeuroimage\u003c/em\u003e \u003cb\u003e165\u003c/b\u003e, 238\u0026ndash;250. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.neuroimage.2017.09.056\u003c/span\u003e\u003cspan address=\"10.1016/j.neuroimage.2017.09.056\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2018a).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEbner, M. et al. \u003cem\u003eAn Automated Localization, Segmentation and Reconstruction Framework for Fetal Brain MRI\u003c/em\u003e, Cham. (2018b).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEbner, M. et al. An automated framework for localization, segmentation and super-resolution reconstruction of fetal brain MRI. \u003cem\u003eNeuroimage\u003c/em\u003e \u003cb\u003e206\u003c/b\u003e, 116324. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.neuroimage.2019.116324\u003c/span\u003e\u003cspan address=\"10.1016/j.neuroimage.2019.116324\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEdgar, H. J. H. et al. \u003cem\u003eNew Mexico Decedent Image Database\u003c/em\u003e. (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFedorov, A. et al. 3D Slicer as an image computing platform for the Quantitative Imaging Network. \u003cem\u003eMagn. Reson. Imaging\u003c/em\u003e. \u003cb\u003e30\u003c/b\u003e (9), 1323\u0026ndash;1341. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.mri.2012.05.001\u003c/span\u003e\u003cspan address=\"10.1016/j.mri.2012.05.001\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFischer, M. C. M. Database of segmentations and surface models of bones of the entire lower body created from cadaver CT scans. \u003cem\u003eSci. Data\u003c/em\u003e. \u003cb\u003e10\u003c/b\u003e (1), 763. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41597-023-02669-z\u003c/span\u003e\u003cspan address=\"10.1038/s41597-023-02669-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFranklin, D. et al. Concordance of traditional osteometric and volume-rendered MSCT interlandmark cranial measurements. \u003cem\u003eInt. J. Legal Med.\u003c/em\u003e \u003cb\u003e127\u003c/b\u003e (2), 505\u0026ndash;520. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s00414-012-0772-9\u003c/span\u003e\u003cspan address=\"10.1007/s00414-012-0772-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFranklin, D., Cardini, A., O\u0026rsquo;Higgins, P., Oxnard, C. E. \u0026amp; Dadour, I. Mandibular morphology as an indicator of human subadult age: geometric morphometric approaches. \u003cem\u003eForensic Sci. Med. Pathol.\u003c/em\u003e \u003cb\u003e4\u003c/b\u003e (2), 91\u0026ndash;99. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s12024-007-9015-7\u003c/span\u003e\u003cspan address=\"10.1007/s12024-007-9015-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2008).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFrelat, M. A. \u0026amp; Mittereocker, P. Postnatal ontogeny of tibia and femur form in two human populations: a multivariate morphometric analysis. \u003cem\u003eAm. J. Hum. Biology\u003c/em\u003e. \u003cb\u003e23\u003c/b\u003e (6), 796\u0026ndash;804. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/ajhb.21217\u003c/span\u003e\u003cspan address=\"10.1002/ajhb.21217\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2011).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFu, Y., Liu, S., Li, H. H. \u0026amp; Yang, D. Automatic and hierarchical segmentation of the human skeleton in CT images. \u003cem\u003ePhys. Med. Biol.\u003c/em\u003e \u003cb\u003e62\u003c/b\u003e (7), 2812. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1088/1361-6560/aa6055\u003c/span\u003e\u003cspan address=\"10.1088/1361-6560/aa6055\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGarvin, H. M. \u0026amp; Stock, M. K. The utility of advanced imaging in forensic anthropology. \u003cem\u003eAcad. Forensic Pathol.\u003c/em\u003e \u003cb\u003e6\u003c/b\u003e (3), 499\u0026ndash;516. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.23907/2016.050\u003c/span\u003e\u003cspan address=\"10.23907/2016.050\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGascho, D., Thali, M. J. \u0026amp; Niemann, T. Post-mortem computed tomography: Technical principles and recommended parameter settings for high-resolution imaging. \u003cem\u003eMed. Sci. Law\u003c/em\u003e. \u003cb\u003e58\u003c/b\u003e (1), 70\u0026ndash;82. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1177/0025802417747167\u003c/span\u003e\u003cspan address=\"10.1177/0025802417747167\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGeijer, M. \u0026amp; El-Khoury, G. Y. MDCT in the evaluation of skeletal trauma: principles, protocols, and clinical applications. \u003cem\u003eEmerg. Radiol.\u003c/em\u003e \u003cb\u003e13\u003c/b\u003e (1), 7\u0026ndash;18. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s10140-006-0509-5\u003c/span\u003e\u003cspan address=\"10.1007/s10140-006-0509-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2006).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGoldman, L. W. Principles of CT: Radiation Dose and Image Quality. \u003cem\u003eJ. Nucl. Med. Technol.\u003c/em\u003e \u003cb\u003e35\u003c/b\u003e (4), 213\u0026ndash;225. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.2967/jnmt.106.037846\u003c/span\u003e\u003cspan address=\"10.2967/jnmt.106.037846\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2007).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGunz, P., Mitteroecker, P., Neubauer, S., Weber, G. W. \u0026amp; Bookstein, F. L. Principles for the virtual reconstruction of hominin crania. \u003cem\u003eJ. Hum. Evol.\u003c/em\u003e \u003cb\u003e57\u003c/b\u003e (1), 48\u0026ndash;62. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.jhevol.2009.04.004\u003c/span\u003e\u003cspan address=\"10.1016/j.jhevol.2009.04.004\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2009).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHong, E., Kwak, D. S. \u0026amp; Kim, I. B. Morphometric Evaluation of Detailed Asymmetry for the Proximal Humerus in Korean Population. \u003cem\u003eSymmetry\u003c/em\u003e \u003cb\u003e13\u003c/b\u003e (5), 862 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuseynov, A. et al. Developmental evidence for obstetric adaptation of the human female pelvis. \u003cem\u003eProc. Natl. Acad. Sci. U.S.A.\u003c/em\u003e \u003cb\u003e113\u003c/b\u003e (19), 5227\u0026ndash;5232. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1073/pnas.1517085113\u003c/span\u003e\u003cspan address=\"10.1073/pnas.1517085113\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuttenlocher, D. P., Klanderman, G. A. \u0026amp; Rucklidge, W. J. Comparing images using the Hausdorff distance. \u003cem\u003eIEEE Trans. Pattern Anal. Mach. Intell.\u003c/em\u003e \u003cb\u003e15\u003c/b\u003e (9), 850\u0026ndash;863. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1109/34.232073\u003c/span\u003e\u003cspan address=\"10.1109/34.232073\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (1993).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJeuthe, J. et al. SEMI-AUTOMATED 3D. \u003cem\u003eRadiat. Prot. Dosimetry\u003c/em\u003e. \u003cb\u003e195\u003c/b\u003e (3\u0026ndash;4), 172\u0026ndash;176. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/rpd/ncab073\u003c/span\u003e\u003cspan address=\"10.1093/rpd/ncab073\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJohnson, J., Alahi, A. \u0026amp; Fei-Fei, L. \u003cem\u003ePerceptual Losses for Real-Time Style Transfer and Super-Resolution\u003c/em\u003e, Cham. (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKistler, M., Bonaretti, S., Pfahrer, M., Niklaus, R. \u0026amp; B\u0026uuml;chler, P. The Virtual Skeleton Database: An Open Access Repository for Biomedical Research and Collaboration. \u003cem\u003eJ. Med. Internet. Res.\u003c/em\u003e \u003cb\u003e15\u003c/b\u003e (11), e245. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.2196/jmir.2930\u003c/span\u003e\u003cspan address=\"10.2196/jmir.2930\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLangley, N. R. et al. Error quantification of osteometric data in forensic anthropology. \u003cem\u003eForensic Sci. Int.\u003c/em\u003e \u003cb\u003e287\u003c/b\u003e, 183\u0026ndash;189. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.forsciint.2018.04.004\u003c/span\u003e\u003cspan address=\"10.1016/j.forsciint.2018.04.004\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLasek, J. \u0026amp; Pi\u0026oacute;rkowski, A. \u003cem\u003eCT Scan Transformation from a Sharp to a Soft Reconstruction Kernel Using Filtering Techniques\u003c/em\u003e, Singapore. (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLautenschlager, S. Reconstructing the past: methods and techniques for the digital restoration of fossils. \u003cem\u003eRoyal Soc. Open. Sci.\u003c/em\u003e \u003cb\u003e3\u003c/b\u003e (10), 160342. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1098/rsos.160342\u003c/span\u003e\u003cspan address=\"10.1098/rsos.160342\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLedig, C. et al. \u003cem\u003ePhoto-realistic single image super-resolution using a generative adversarial network.\u003c/em\u003e Paper presented at the Proceedings of the IEEE conference on computer vision and pattern recognition. (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee, K. H. et al. Managing the CT Data Explosion: Initial Experiences of Archiving Volumetric Datasets in a Mini-PACS. \u003cem\u003eJ. Digit. Imaging\u003c/em\u003e. \u003cb\u003e18\u003c/b\u003e (3), 188\u0026ndash;195. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s10278-005-5163-z\u003c/span\u003e\u003cspan address=\"10.1007/s10278-005-5163-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2005).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi, Y., Sixou, B. \u0026amp; Peyrin, F. A Review of the Deep Learning Methods for Medical Images Super Resolution Problems. \u003cem\u003eIRBM\u003c/em\u003e \u003cb\u003e42\u003c/b\u003e (2), 120\u0026ndash;133. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.irbm.2020.08.004\u003c/span\u003e\u003cspan address=\"10.1016/j.irbm.2020.08.004\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLund, H., Gr\u0026ouml;ndahl, K. \u0026amp; Gr\u0026ouml;ndahl, H. G. Accuracy and precision of linear measurements in cone beam computed tomography Accuitomo\u0026reg; tomograms obtained with different reconstruction techniques. \u003cem\u003eDentomaxillofacial Radiol.\u003c/em\u003e \u003cb\u003e38\u003c/b\u003e (6), 379\u0026ndash;386. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1259/dmfr/15022357\u003c/span\u003e\u003cspan address=\"10.1259/dmfr/15022357\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMedixant. RadiAnt DICOM Viewer (Version 2025.2). (2025). Retrieved from \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.radiantviewer.com\u003c/span\u003e\u003cspan address=\"https://www.radiantviewer.com\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMopin, C., Chaumo\u0026icirc;tre, K., Signoli, M. \u0026amp; Adalian, P. Developmental stability and environmental stress: A geometric morphometrics analysis of asymmetry in the human femur. \u003cem\u003eAm. J. Phys. Anthropol.\u003c/em\u003e \u003cb\u003e167\u003c/b\u003e (1), 144\u0026ndash;160. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/ajpa.23613\u003c/span\u003e\u003cspan address=\"10.1002/ajpa.23613\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMorimoto, N., Nakatsukasa, M., Ponce de Le\u0026oacute;n, M. S. \u0026amp; Zollikofer, C. P. E. Femoral ontogeny in humans and great apes and its implications for their last common ancestor. \u003cem\u003eScientific Reports, 8\u003c/em\u003e(1), 1930. (2018). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41598-018-20410-4\u003c/span\u003e\u003cspan address=\"10.1038/s41598-018-20410-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMoustafa, M., Ebied, H. M., Helmy, A., Nazamy, T. M. \u0026amp; Tolba, M. F. Optimization Methods for Medical Image Super Resolution Reconstruction. In (eds Hassanien, A. E., Grosan, C. \u0026amp; Fahmy Tolba, M.) Applications of Intelligent Optimization in Biology and Medicine: Current Trends and Open Problems (135\u0026ndash;157). Cham: Springer International Publishing. (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNoble, E. \u0026amp; Hawks, J. A Geometric Morphometric Study of Scapular Ontogeny in Modern Humans. \u003cem\u003eAm. J. Biol. Anthropol.\u003c/em\u003e \u003cb\u003e187\u003c/b\u003e (3), e70090. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/ajpa.70090\u003c/span\u003e\u003cspan address=\"10.1002/ajpa.70090\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePark, J. et al. Computed tomography super-resolution using deep convolutional neural network. \u003cem\u003ePhys. Med. Biol.\u003c/em\u003e \u003cb\u003e63\u003c/b\u003e (14), 145011. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1088/1361-6560/aacdd4\u003c/span\u003e\u003cspan address=\"10.1088/1361-6560/aacdd4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePark, S. C., Park, M. K. \u0026amp; Kang, M. G. Super-resolution image reconstruction: a technical overview. \u003cem\u003eIEEE. Signal. Process. Mag.\u003c/em\u003e \u003cb\u003e20\u003c/b\u003e (3), 21\u0026ndash;36. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1109/MSP.2003.1203207\u003c/span\u003e\u003cspan address=\"10.1109/MSP.2003.1203207\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2003).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePerini, T. A., Oliveira, G. L., d., Ornellas, J. S. \u0026amp; Oliveira, F. P. d. C\u0026aacute;lculo do erro t\u0026eacute;cnico de medi\u0026ccedil;\u0026atilde;o em antropometria. \u003cem\u003eRevista Brasileira de Medicina do Esporte, 11\u003c/em\u003e. (2005).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePinter, C., Lasso, A., Wang, A., Jaffray, D. \u0026amp; Fichtinger, G. SlicerRT: Radiation therapy research toolkit for 3D Slicer. \u003cem\u003eMed. Phys.\u003c/em\u003e \u003cb\u003e39\u003c/b\u003e (10), 6332\u0026ndash;6338. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1118/1.4754659\u003c/span\u003e\u003cspan address=\"10.1118/1.4754659\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePujol, A., Rissech, C., Ventura, J., Badosa, J. \u0026amp; Turb\u0026oacute;n, D. Ontogeny of the female femur: geometric morphometric analysis applied on current living individuals of a Spanish population. \u003cem\u003eJ. Anat.\u003c/em\u003e \u003cb\u003e225\u003c/b\u003e (3), 346\u0026ndash;357. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/joa.12209\u003c/span\u003e\u003cspan address=\"10.1111/joa.12209\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePujol, A., Rissech, C., Ventura, J. \u0026amp; Turb\u0026oacute;n, D. Ontogeny of the male femur: Geometric morphometric analysis applied to a contemporary Spanish population. \u003cem\u003eAm. J. Phys. Anthropol.\u003c/em\u003e \u003cb\u003e159\u003c/b\u003e (1), 146\u0026ndash;163. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/ajpa.22846\u003c/span\u003e\u003cspan address=\"10.1002/ajpa.22846\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQiu, D., Cheng, Y. \u0026amp; Wang, X. Medical image super-resolution reconstruction algorithms based on deep learning: A survey. \u003cem\u003eComput. Methods Programs Biomed.\u003c/em\u003e \u003cb\u003e238\u003c/b\u003e, 107590. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.cmpb.2023.107590\u003c/span\u003e\u003cspan address=\"10.1016/j.cmpb.2023.107590\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eR Core Team. \u003cem\u003eR: A language and environment for statistical computing\u003c/em\u003e (R Foundation for Statistical Computing, 2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRathnayaka, K., Sahama, T., Schuetz, M. A. \u0026amp; Schmutz, B. Effects of CT image segmentation methods on the accuracy of long bone 3D reconstructions. \u003cem\u003eMed. Eng. Phys.\u003c/em\u003e \u003cb\u003e33\u003c/b\u003e (2), 226\u0026ndash;233. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.medengphy.2010.10.002\u003c/span\u003e\u003cspan address=\"10.1016/j.medengphy.2010.10.002\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2011).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRolfe, S. et al. SlicerMorph: An open and extensible platform to retrieve, visualize and analyse 3D morphology. \u003cem\u003eMethods Ecol. Evol.\u003c/em\u003e \u003cb\u003e12\u003c/b\u003e (10), 1816\u0026ndash;1825. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/2041-210X.13669\u003c/span\u003e\u003cspan address=\"10.1111/2041-210X.13669\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSawdayee, H., Vaxman, A. \u0026amp; Bermano, A. H. \u003cem\u003eOrex: Object reconstruction from planar cross-sections using neural fields.\u003c/em\u003e Paper presented at the Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eScherf, H. Computed tomography in paleoanthropology \u0026mdash; an overview. \u003cem\u003eArchaeol. Anthropol. Sci.\u003c/em\u003e \u003cb\u003e5\u003c/b\u003e (3), 205\u0026ndash;214. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s12520-013-0128-5\u003c/span\u003e\u003cspan address=\"10.1007/s12520-013-0128-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eScherf, H. \u0026amp; Tilgner, R. A new high-resolution computed tomography (CT) segmentation method for trabecular bone architectural analysis. \u003cem\u003eAm. J. Phys. Anthropol.\u003c/em\u003e \u003cb\u003e140\u003c/b\u003e (1), 39\u0026ndash;51. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/ajpa.21033\u003c/span\u003e\u003cspan address=\"10.1002/ajpa.21033\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2009).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShanley, C., Obertov\u0026aacute;, Z. \u0026amp; Franklin, D. Dimensional accuracy of structured light scans and 3D prints of various human skeletal elements. \u003cem\u003eForensic Sci. Int.\u003c/em\u003e \u003cb\u003e361\u003c/b\u003e, 112138. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.forsciint.2024.112138\u003c/span\u003e\u003cspan address=\"10.1016/j.forsciint.2024.112138\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSimmons-Ehrhardt, T. Open osteology: Medical imaging databases as skeletal collections. \u003cem\u003eForensic Imaging\u003c/em\u003e. \u003cb\u003e26\u003c/b\u003e, 200462. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.fri.2021.200462\u003c/span\u003e\u003cspan address=\"10.1016/j.fri.2021.200462\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStock, M. K. et al. The importance of processing procedures and threshold values in CT scan segmentation of skeletal elements: An example using the immature os coxa. \u003cem\u003eForensic Sci. Int.\u003c/em\u003e \u003cb\u003e309\u003c/b\u003e \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.forsciint.2020.110232\u003c/span\u003e\u003cspan address=\"10.1016/j.forsciint.2020.110232\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStull, K. E. \u0026amp; Corron, L. K. \u003cem\u003eSubadult Virtual Anthropology Database (SVAD) Data Collection Protocol: Epiphyseal Fusion, Diaphyseal Dimensions, Dental Development Stages, Vertebral Neural Canal Dimensions\u003c/em\u003e 2nd edn (Zenodo, 2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStull, K. E. \u0026amp; Corron, L. K. The Subadult Virtual Anthropology Database (SVAD): An Accessible Repository of Contemporary Subadult Reference Data. \u003cem\u003eForensic Sci.\u003c/em\u003e \u003cb\u003e2\u003c/b\u003e (1), 20\u0026ndash;36. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/forensicsci2010003\u003c/span\u003e\u003cspan address=\"10.3390/forensicsci2010003\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStull, K. E. \u0026amp; Garvin, H. M. \u003cem\u003eA Visual Atlas of Skeletal Growth and Development\u003c/em\u003e 1st edn (Taylor \u0026amp; Francis Group, 2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStull, K. E., L'Abb\u0026eacute;, E. N. \u0026amp; Ousley, S. D. Using multivariate adaptive regression splines to estimate subadult age from diaphyseal dimensions. \u003cem\u003eAm. J. Phys. Anthropol.\u003c/em\u003e \u003cb\u003e154\u003c/b\u003e (3), 376\u0026ndash;386. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1002/ajpa.22522\u003c/span\u003e\u003cspan address=\"10.1002/ajpa.22522\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2014a).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStull, K. E., L'abb\u0026eacute;, E. N. \u0026amp; Steiner, S. Measuring distortion of skeletal elements in Lodox Statscan-generated images. \u003cem\u003eClin. Anat.\u003c/em\u003e \u003cb\u003e26\u003c/b\u003e (6), 780\u0026ndash;786. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/ca.22225\u003c/span\u003e\u003cspan address=\"10.1002/ca.22225\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStull, K. E., Tise, M. L., Ali, Z. \u0026amp; Fowler, D. R. Accuracy and reliability of measurements obtained from computed tomography 3D volume rendered images. \u003cem\u003eForensic Sci. Int.\u003c/em\u003e \u003cb\u003e238\u003c/b\u003e, 133\u0026ndash;140. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.forsciint.2014.03.005\u003c/span\u003e\u003cspan address=\"10.1016/j.forsciint.2014.03.005\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2014b).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTang, Z., Xing, X., Wang, G. \u0026amp; Yang, G. Enhancing Super-Resolution Network Efficacy in CT Imaging: Cost-Effective Simulation of Training Data. \u003cem\u003eIEEE Open. J. Eng. Med. Biology\u003c/em\u003e. \u003cb\u003e6\u003c/b\u003e, 576\u0026ndash;583. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1109/OJEMB.2025.3610160\u003c/span\u003e\u003cspan address=\"10.1109/OJEMB.2025.3610160\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eThomas, D. B., Hiscox, J. D., Dixon, B. J. \u0026amp; Potgieter, J. 3D scanning and printing skeletal tissues for anatomy education. \u003cem\u003eJ. Anat.\u003c/em\u003e \u003cb\u003e229\u003c/b\u003e (3), 473\u0026ndash;481. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/joa.12484\u003c/span\u003e\u003cspan address=\"10.1111/joa.12484\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTian, J. \u0026amp; Ma, K. K. A survey on super-resolution imaging. \u003cem\u003eSignal. Image Video Process.\u003c/em\u003e \u003cb\u003e5\u003c/b\u003e (3), 329\u0026ndash;342. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s11760-010-0204-6\u003c/span\u003e\u003cspan address=\"10.1007/s11760-010-0204-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2011).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUldin, T. Virtual Anthropology \u0026ndash; A Brief Review of the Literature and History of Computed Tomography. \u003cem\u003eForensic Sci. Res.\u003c/em\u003e \u003cb\u003e2\u003c/b\u003e (4), 165\u0026ndash;173. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1080/20961790.2017.1369621\u003c/span\u003e\u003cspan address=\"10.1080/20961790.2017.1369621\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVan den Broeck, J., Vereecke, E., Wirix-Speetjens, R. \u0026amp; Vander Sloten, J. Segmentation accuracy of long bones. \u003cem\u003eMed. Eng. Phys.\u003c/em\u003e \u003cb\u003e36\u003c/b\u003e (7), 949\u0026ndash;953. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.medengphy.2014.03.016\u003c/span\u003e\u003cspan address=\"10.1016/j.medengphy.2014.03.016\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVi\u0026eth;arsd\u0026oacute;ttir, U. S., O'higgins, P. \u0026amp; Stringer, C. A geometric morphometric study of regional differences in the ontogeny of the modern human facial skeleton. \u003cem\u003eJ. Anat.\u003c/em\u003e \u003cb\u003e201\u003c/b\u003e, 211\u0026ndash;229. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1046/j.1469-7580.2002.00092.x\u003c/span\u003e\u003cspan address=\"10.1046/j.1469-7580.2002.00092.x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2002).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWeber, G. W. Virtual Anthropology. \u003cem\u003eAm. J. Phys. Anthropol.\u003c/em\u003e \u003cb\u003e156\u003c/b\u003e (S59), 22\u0026ndash;42. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/ajpa.22658\u003c/span\u003e\u003cspan address=\"10.1002/ajpa.22658\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYoshinobu, T. et al. Data Management Solution for Large-Volume Computed Tomography in an Existing Picture Archiving and Communication System (PACS). \u003cem\u003eJ. Digit. Imaging\u003c/em\u003e. \u003cb\u003e24\u003c/b\u003e (1), 107\u0026ndash;113. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s10278-009-9251-3\u003c/span\u003e\u003cspan address=\"10.1007/s10278-009-9251-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2011).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYushkevich, P. A. et al. User-guided 3D active contour segmentation of anatomical structures: Significantly improved efficiency and reliability. \u003cem\u003eNeuroimage\u003c/em\u003e \u003cb\u003e31\u003c/b\u003e (3), 1116\u0026ndash;1128. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.neuroimage.2006.01.015\u003c/span\u003e\u003cspan address=\"10.1016/j.neuroimage.2006.01.015\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2006).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang, S. \u0026amp; Xia, Y. CT image reconstruction algorithms: A comprehensive survey. \u003cem\u003eConcurrency Computation: Pract. Experience\u003c/em\u003e. \u003cb\u003e33\u003c/b\u003e (8), e5506. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/cpe.5506\u003c/span\u003e\u003cspan address=\"10.1002/cpe.5506\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZollikofer, C. P. E. \u0026amp; de Ponce, M. S. Visualizing patterns of craniofacial shape variation in \u0026lt;\u0026thinsp;i\u0026thinsp;\u0026gt;\u0026thinsp;Homo sapiens. \u003cem\u003eProceedings of the Royal Society of London. Series B: Biological Sciences, 269\u003c/em\u003e(1493), 801\u0026ndash;807. doi: (2002). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1098/rspb.2002.1960\u003c/span\u003e\u003cspan address=\"10.1098/rspb.2002.1960\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":true,"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":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Computed tomography (CT), morphometrics, super-resolution reconstruction, skeletal biology","lastPublishedDoi":"10.21203/rs.3.rs-8195787/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8195787/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eClinical computed tomography (CT) datasets are increasingly common in skeletal research, yet archived retrospective datasets with thick-slice, anisotropic reconstructions are more commonly available to researchers than the original high-resolution scans. Models reconstructed from these suboptimal scans can produce distorted measurements and irregular surfaces. This study evaluates a super-resolution reconstruction (SRR) framework for generating high-resolution skeletal models from multiple orthogonal, thick-slice CT stacks. Archived CT scans of long bones from 33 individuals (0\u0026ndash;16 years) were collected from National Taiwan University Hospital. For each individual, 3D models were generated from both the original thick-slice stacks and SRR-processed volumes. Linear measurements were compared with those taken directly from high-resolution picture archiving and communication system (PACS) renderings, and geometric similarity was quantified by signed Hausdorff distance and Dice similarity coefficient. SRR-reconstructed models showed lower measurement error and greater agreement with the PACS rendering than the thick-slice models. Surface geometry was generally consistent across model types, with localized deviations concentrated at metaphyseal and epiphyseal regions. SRR processing also produced smoother, more anatomically plausible surfaces with a clearer separation of fusing elements. These results demonstrate that SRR can improve virtual skeletal model quality from suboptimal clinical imaging, particularly for applications where anatomically precise surfaces are beneficial.\u003c/p\u003e","manuscriptTitle":"Adapting super-resolution reconstruction for skeletal analysis of clinical computed tomography data","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-01-16 00:04:47","doi":"10.21203/rs.3.rs-8195787/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewerAgreed","content":"17485742015493238320637783939937662970","date":"2026-04-13T19:24:51+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"242408707376166776525884088971888043317","date":"2026-03-25T02:15:50+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-11-30T07:25:48+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-11-30T07:23:14+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-11-28T15:12:27+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-11-26T13:28:46+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2025-11-26T13:22:19+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"c5313fdf-158b-45c6-8180-d1e97878cbf6","owner":[],"postedDate":"January 16th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-01-16T00:04:47+00:00","versionOfRecord":[],"versionCreatedAt":"2026-01-16 00:04:47","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8195787","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8195787","identity":"rs-8195787","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}