pith. sign in

arxiv: 2504.02382 · v2 · submitted 2025-04-03 · 📡 eess.IV · cs.AI· cs.CV

Benchmark of Segmentation Techniques for Pelvic Fracture in CT and X-ray: Summary of the PENGWIN 2024 Challenge

Yudi Sang , Yanzhen Liu , Sutuke Yibulayimu , Yunning Wang , Benjamin D. Killeen , Mingxu Liu , Ping-Cheng Ku , Ole Johannsen
show 28 more authors
Karol Gotkowski Maximilian Zenk Klaus Maier-Hein Fabian Isensee Peiyan Yue Yi Wang Haidong Yu Zhaohong Pan Yutong He Xiaokun Liang Daiqi Liu Fuxin Fan Artur Jurgas Andrzej Skalski Yuxi Ma Jing Yang Szymon P{\l}otka Rafa{\l} Litka Gang Zhu Yingchun Song Mathias Unberath Mehran Armand Dan Ruan S. Kevin Zhou Qiyong Cao Chunpeng Zhao Xinbao Wu Yu Wang
This is my paper

Pith reviewed 2026-05-22 21:51 UTC · model grok-4.3

classification 📡 eess.IV cs.AIcs.CV
keywords pelvic fracture segmentationCT imagingX-ray imaginginstance segmentationDeepDRR simulationMICCAI challengetrauma surgeryfragment-wise IoU
0
0 comments X

The pith

CT segmentation reaches 0.93 fragment IoU on pelvic fractures while X-ray reaches only 0.77

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper benchmarks multiple algorithms on the task of separating individual bone fragments in pelvic fracture cases. It reports that the best CT methods reach an average fragment-wise intersection over union of 0.930 across a multi-center dataset, while the best X-ray methods reach 0.774 on images simulated from the same cases. A sympathetic reader would care because accurate fragment outlines are needed for surgical planning and for real-time guidance during operations, yet the lower X-ray score shows that projection overlap still limits reliability. The authors also note that different ways of representing fragments, such as primary-secondary splits versus boundary-core splits, produce distinct strategies and that uncertainties remain when fractures are incomplete.

Core claim

The top-performing CT algorithm achieved an average fragment-wise intersection over union of 0.930 while the best X-ray algorithm achieved 0.774. These scores indicate satisfactory accuracy for CT but insufficient performance for intra-operative X-ray decision-making. The challenge exposed that instance representation choices, such as primary-secondary classification versus boundary-core separation, lead to different segmentation strategies and that fragment definition carries inherent uncertainty in incomplete fractures.

What carries the argument

Fragment-wise intersection over union (IoU) computed on instance-level masks, used to compare algorithms that differ in how they represent fracture fragments.

If this is right

  • CT fragment segmentation is accurate enough to support pre-operative planning and post-operative assessment.
  • X-ray fragment segmentation is not yet reliable enough for intra-operative use because of overlap in projection views.
  • Instance representation choices, such as primary-secondary versus boundary-core, produce measurably different segmentation outcomes.
  • Uncertainties in defining incomplete fractures limit fully automatic methods and point toward interactive segmentation.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Real clinical X-rays may yield lower IoU than the simulated set because the simulation may understate certain artifacts.
  • A hybrid pipeline that starts with CT segmentation and then refines on X-ray could reduce the performance gap.
  • The released multi-center dataset and simulated X-rays can serve as a public testbed for future projection-aware segmentation research.

Load-bearing premise

The simulated X-ray images generated by DeepDRR capture the noise, artifacts, and positioning variations present in real clinical X-rays.

What would settle it

Apply the winning X-ray algorithms to a held-out set of real intraoperative X-ray images and measure whether fragment-wise IoU remains near 0.774 or drops substantially.

Figures

Figures reproduced from arXiv: 2504.02382 by Andrzej Skalski, Artur Jurgas, Benjamin D. Killeen, Chunpeng Zhao, Daiqi Liu, Dan Ruan, Fabian Isensee, Fuxin Fan, Gang Zhu, Haidong Yu, Jing Yang, Karol Gotkowski, Klaus Maier-Hein, Mathias Unberath, Maximilian Zenk, Mehran Armand, Mingxu Liu, Ole Johannsen, Peiyan Yue, Ping-Cheng Ku, Qiyong Cao, Rafa{\l} Litka, S. Kevin Zhou, Sutuke Yibulayimu, Szymon P{\l}otka, Xiaokun Liang, Xinbao Wu, Yanzhen Liu, Yingchun Song, Yi Wang, Yudi Sang, Yunning Wang, Yutong He, Yu Wang, Yuxi Ma, Zhaohong Pan.

Figure 1
Figure 1. Figure 1: Example pelvic CT scan and the annotated segmentation rendered in [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 3
Figure 3. Figure 3: Example simulated X-ray image and the annotation. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Different fracture segmentation pipelines in the CT task. (a) Boundary-core decomposition scheme adopted by MIC-DKFZ, MedApp-AGH, and Sano. [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Different fracture segmentation pipelines in the X-ray task. (a) Bone extraction followed by primary-secondary segmentation by SMILE. (b) Direct [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Example CT segmentation results from the top-performing teams. Four representative cases are shown in (a)-(d), illustrating varying levels of complexity [PITH_FULL_IMAGE:figures/full_fig_p011_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Example X-ray segmentation results from the top-performing teams. Four representative cases are shown in (a)-(d), illustrating varying levels of [PITH_FULL_IMAGE:figures/full_fig_p012_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Significance matrices showing the p-values of the participating algorithms using one-sided Wilcoxon signed-rank tests. (a) Fragment size in CT (mm3) (b) Fragment count in CT (c) Fragment size in X-ray (pixel2) (d) Fragment count in X-ray [PITH_FULL_IMAGE:figures/full_fig_p013_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Impact of fragment size and count on the IoU-F metric for the top five performing algorithms in both tasks. In (a) and (c), fragment sizes are binned [PITH_FULL_IMAGE:figures/full_fig_p013_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Bootstrap-based stability analysis of team rankings in both tasks. Each dot represents the frequency with which a team achieved a specific rank [PITH_FULL_IMAGE:figures/full_fig_p014_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Example failure modes. The red and yellow boxes indicate false [PITH_FULL_IMAGE:figures/full_fig_p014_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Example reduction planning result using different segmentation [PITH_FULL_IMAGE:figures/full_fig_p015_12.png] view at source ↗
read the original abstract

The segmentation of pelvic fracture fragments in CT and X-ray images is crucial for trauma diagnosis, surgical planning, and intraoperative guidance. However, accurately and efficiently delineating the bone fragments remains a significant challenge due to complex anatomy and imaging limitations. The PENGWIN challenge, organized as a MICCAI 2024 satellite event, aimed to advance automated fracture segmentation by benchmarking state-of-the-art algorithms on these complex tasks. A diverse dataset of 150 CT scans was collected from multiple clinical centers, and a large set of simulated X-ray images was generated using the DeepDRR method. Final submissions from 16 teams worldwide were evaluated under a rigorous multi-metric testing scheme. The top-performing CT algorithm achieved an average fragment-wise intersection over union (IoU) of 0.930, demonstrating satisfactory accuracy. However, in the X-ray task, the best algorithm achieved an IoU of 0.774, which is promising but not yet sufficient for intra-operative decision-making, reflecting the inherent challenges of fragment overlap in projection imaging. Beyond the quantitative evaluation, the challenge revealed methodological diversity in algorithm design. Variations in instance representation, such as primary-secondary classification versus boundary-core separation, led to differing segmentation strategies. Despite promising results, the challenge also exposed inherent uncertainties in fragment definition, particularly in cases of incomplete fractures. These findings suggest that interactive segmentation approaches, integrating human decision-making with task-relevant information, may be essential for improving model reliability and clinical applicability.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 0 minor

Summary. The manuscript summarizes the PENGWIN 2024 MICCAI challenge benchmarking automated segmentation of pelvic fracture fragments. It uses a multi-center dataset of 150 CT scans and corresponding simulated X-ray projections generated by DeepDRR. Sixteen independent team submissions were evaluated on held-out test data under a multi-metric protocol. The top CT method reached a fragment-wise IoU of 0.930; the top X-ray method reached 0.774. The paper additionally catalogs variations in instance-representation strategies across submissions and notes definitional uncertainty for incomplete fractures, concluding that interactive segmentation may be required for clinical reliability.

Significance. If the reported metrics hold, the work supplies a valuable, reproducible public benchmark for a clinically important task. The use of independent submissions evaluated on held-out multi-center data, together with a rigorous multi-metric scheme, strengthens the reliability of the quantitative claims. The CT result demonstrates that current methods can reach high fragment-wise accuracy; the X-ray result quantifies the additional difficulty of projection imaging. The explicit discussion of methodological diversity (primary-secondary vs. boundary-core representations) provides concrete guidance for future algorithm design. These empirical strengths are the primary contribution.

major comments (1)
  1. [Abstract] Abstract: the claim that an IoU of 0.774 on the X-ray task is 'promising but not yet sufficient for intra-operative decision-making' rests on the unvalidated assumption that DeepDRR simulations faithfully reproduce real clinical X-ray characteristics (scatter, detector noise, beam hardening, and variable fragment overlap arising from patient positioning). No quantitative comparison between simulated and real X-ray images is reported, making the clinical-sufficiency interpretation load-bearing yet unsupported by the presented evidence.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the positive evaluation of the work and for the constructive comment on the abstract. We address the point below.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim that an IoU of 0.774 on the X-ray task is 'promising but not yet sufficient for intra-operative decision-making' rests on the unvalidated assumption that DeepDRR simulations faithfully reproduce real clinical X-ray characteristics (scatter, detector noise, beam hardening, and variable fragment overlap arising from patient positioning). No quantitative comparison between simulated and real X-ray images is reported, making the clinical-sufficiency interpretation load-bearing yet unsupported by the presented evidence.

    Authors: We agree that the manuscript reports no quantitative comparison between DeepDRR-simulated projections and real clinical X-ray images, and that the interpretive claim regarding sufficiency for intra-operative decision-making therefore lacks direct supporting evidence. In the revised version we will remove this clinical-sufficiency statement from the abstract and replace it with a purely empirical observation on the performance gap between CT and X-ray tasks. revision: yes

Circularity Check

0 steps flagged

Empirical benchmark results with no circular derivations

full rationale

The paper reports performance metrics from a public challenge with independent team submissions evaluated on held-out test data. No equations, predictions, or first-principles derivations are present that reduce to fitted inputs or self-citations by construction. Central claims (CT IoU 0.930, X-ray IoU 0.774) are direct empirical outcomes. Any citations to DeepDRR or prior work are not load-bearing for the reported numbers and do not create self-definitional or fitted-input circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claims depend on the representativeness of the multi-center CT dataset and the fidelity of DeepDRR simulations for X-ray, along with consistent fragment annotations despite noted uncertainties in incomplete fractures.

axioms (1)
  • domain assumption Fragment-wise IoU is a suitable primary metric for assessing clinical utility of fracture segmentation algorithms.
    The paper adopts this without providing correlation studies to actual surgical outcomes or inter-observer variability in fragment labeling.

pith-pipeline@v0.9.0 · 5965 in / 1355 out tokens · 34965 ms · 2026-05-22T21:51:21.307371+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Vision Transformer-Conditioned UNet for Domain-Adaptive Semantic Segmentation

    cs.CV 2026-05 unverdicted novelty 5.0

    ViTC-UNet adapts frozen ViT representations to biomedical semantic segmentation by conditioning a UNet via learnable tokens and two-way attention decoding.

Reference graph

Works this paper leans on

36 extracted references · 36 canonical work pages · cited by 1 Pith paper

  1. [1]

    Basics of computer-assisted orthopaedic surgery,

    J. Kowal, F. Langlotz, and L.-P. Nolte, “Basics of computer-assisted orthopaedic surgery,” inNavigation and MIS in Orthopedic Surgery. Springer, 2007, pp. 2–8

    work page 2007

  2. [2]

    Computer-assisted orthopaedic surgery and robotic surgery in total hip arthroplasty,

    N. Sugano, “Computer-assisted orthopaedic surgery and robotic surgery in total hip arthroplasty,”Clin. Orthop. Surg., vol. 5, no. 1, pp. 1–9, 2013

    work page 2013

  3. [3]

    Robot-assisted autonomous reduction of a displaced pelvic fracture: a case report and brief literature review,

    Y . Ge, C. Zhao, Y . Wang, and X. Wu, “Robot-assisted autonomous reduction of a displaced pelvic fracture: a case report and brief literature review,”Journal of Clinical Medicine, vol. 11, no. 6, p. 1598, 2022

    work page 2022

  4. [4]

    Preoperative fracture reduction planning for image-guided pelvic trauma surgery: A comprehensive pipeline with learning,

    Y . Liu, S. Yibulayimu, Y . Sang, G. Zhu, C. Shi, C. Liang, Q. Cao, C. Zhao, X. Wu, and Y . Wang, “Preoperative fracture reduction planning for image-guided pelvic trauma surgery: A comprehensive pipeline with learning,”Medical Image Analysis, p. 103506, 2025

    work page 2025

  5. [5]

    Automatic path planning for pelvic fracture reduction with multi-degree-of-freedom,

    C. Shi, Q. Yang, Y . Wang, X. Zhao, S. Shi, L. Zhang, S. Yibulayimu, Y . Liu, C. Liang, Y . Wanget al., “Automatic path planning for pelvic fracture reduction with multi-degree-of-freedom,”Computer Methods and Programs in Biomedicine, vol. 261, p. 108591, 2025. 17

    work page 2025

  6. [6]

    Fracture reduction planning and guidance in orthopaedic trauma surgery via multi-body image registration,

    R. Han, A. Uneri, R. C. Vijayan, P. Wu, P. Vagdargi, N. Sheth, S. V ogt, G. Kleinszig, G. Osgood, and J. H. Siewerdsen, “Fracture reduction planning and guidance in orthopaedic trauma surgery via multi-body image registration,”Medical image analysis, vol. 68, p. 101917, 2021

    work page 2021

  7. [7]

    Automatic intraoperative CT-CBCT registration for image-guided pelvic fracture reduction,

    Y . Liu, Y . Sang, S. Yibulayimu, G. Zhu, C. Shi, C. Liang, J. Liu, Q. Yang, C. Zhao, Q. Caoet al., “Automatic intraoperative CT-CBCT registration for image-guided pelvic fracture reduction,” in2024 IEEE ISBI. IEEE, 2024, pp. 1–5

    work page 2024

  8. [8]

    DeepDRR–a catalyst for machine learning in fluoroscopy- guided procedures,

    M. Unberath, J.-N. Zaech, S. C. Lee, B. Bier, J. Fotouhi, M. Armand, and N. Navab, “DeepDRR–a catalyst for machine learning in fluoroscopy- guided procedures,” inProc. Med. Image. Comput. Comput. Assist. Interv.Springer, 2018, pp. 98–106

    work page 2018

  9. [9]

    Deep rib fracture instance segmentation and classification from CT on the RibFrac challenge,

    J. Yang, R. Shi, L. Jin, X. Huang, K. Kuang, D. Wei, S. Gu, J. Liu, P. Liu, Z. Chaiet al., “Deep rib fracture instance segmentation and classification from CT on the RibFrac challenge,”arXiv preprint:2402.09372, 2024

    work page arXiv 2024

  10. [10]

    Assessing the performance of models from the 2022 RSNA cervical spine fracture detection competition at a level i trauma center,

    Z. Hu, M. Patel, R. L. Ball, H. M. Lin, L. M. Prevedello, M. Naseri, S. Mathur, R. Moreland, J. Wilson, C. Witiwet al., “Assessing the performance of models from the 2022 RSNA cervical spine fracture detection competition at a level i trauma center,”Radiology: Artificial Intelligence, vol. 6, no. 6, p. e230550, 2024

    work page 2022

  11. [11]

    Fractured bone detection challenge,

    A. Khanal, K. Santosh, and S. K. Bajracharya, “Fractured bone detection challenge,” https://kaggle.com/competitions/ fractured-bone-detection-challenge, 2023, kaggle

    work page 2023

  12. [12]

    Fully automatic and fast segmen- tation of the femur bone from 3D-CT images with no shape prior,

    M. Kr ˇcah, G. Sz ´ekely, and R. Blanc, “Fully automatic and fast segmen- tation of the femur bone from 3D-CT images with no shape prior,” in 2011 IEEE ISBI. IEEE, 2011, pp. 2087–2090

    work page 2011

  13. [13]

    Recognition and segmen- tation of individual bone fragments with a deep learning approach in CT scans of complex intertrochanteric fractures: a retrospective study,

    L. Yang, S. Gao, P. Li, J. Shi, and F. Zhou, “Recognition and segmen- tation of individual bone fragments with a deep learning approach in CT scans of complex intertrochanteric fractures: a retrospective study,” Journal of Digital Imaging, vol. 35, no. 6, pp. 1681–1689, 2022

    work page 2022

  14. [14]

    Automatic segmentation of inconstant fractured fragments for tibia/fibula from CT images using deep learning,

    H. Kim, Y . D. Jeon, K. B. Park, H. Cha, M.-S. Kim, J. You, S.-W. Lee, S.-H. Shin, Y .-G. Chung, S. B. Kanget al., “Automatic segmentation of inconstant fractured fragments for tibia/fibula from CT images using deep learning,”Scientific Reports, vol. 13, no. 1, p. 20431, 2023

    work page 2023

  15. [15]

    Accu- racy and reliability analysis of a machine learning based segmentation tool for intertrochanteric femoral fracture CT,

    D. Wang, Z. Wu, G. Fan, H. Liu, X. Liao, Y . Chen, and H. Zhang, “Accu- racy and reliability analysis of a machine learning based segmentation tool for intertrochanteric femoral fracture CT,”Frontiers in Surgery, vol. 9, p. 913385, 2022

    work page 2022

  16. [16]

    Pelvic fracture segmentation using a multi-scale distance-weighted neural network,

    Y . Liu, S. Yibulayimu, Y . Sang, G. Zhu, Y . Wang, C. Zhao, and X. Wu, “Pelvic fracture segmentation using a multi-scale distance-weighted neural network,” inProc. Med. Image. Comput. Comput. Assist. Interv. Springer, 2023, pp. 312–321

    work page 2023

  17. [17]

    Fragment distance- guided dual-stream learning for automatic pelvic fracture segmentation,

    B. Zeng, H. Wang, L. Joskowicz, and X. Chen, “Fragment distance- guided dual-stream learning for automatic pelvic fracture segmentation,” Comput. Med. Imaging Graph, vol. 116, p. 102412, 2024

    work page 2024

  18. [18]

    Synergistically segmenting and reducing fracture bones via whole-to- whole deep dense matching,

    Z. Deng, J. Jiang, R. Huang, W. Zhang, Z. Chen, K. He, and Q. Yao, “Synergistically segmenting and reducing fracture bones via whole-to- whole deep dense matching,”Computers & Graphics, vol. 116, pp. 404– 417, 2023

    work page 2023

  19. [19]

    A survey of fracture detection techniques in bone X-ray images,

    D. Joshi and T. P. Singh, “A survey of fracture detection techniques in bone X-ray images,”Artificial Intelligence Review, vol. 53, no. 6, pp. 4475–4517, 2020

    work page 2020

  20. [20]

    Application of a deep learning algorithm for detection and visualization of hip fractures on plain pelvic radiographs,

    C.-T. Cheng, T.-Y . Ho, T.-Y . Lee, C.-C. Chang, C.-C. Chou, C.-C. Chen, I. Chung, C.-H. Liaoet al., “Application of a deep learning algorithm for detection and visualization of hip fractures on plain pelvic radiographs,” European radiology, vol. 29, no. 10, pp. 5469–5477, 2019

    work page 2019

  21. [21]

    Osteoporotic vertebral fracture classification in X-rays based on a multi-modal semantic consistency network,

    Y . Wang, T. Bai, T. Li, and L. Huang, “Osteoporotic vertebral fracture classification in X-rays based on a multi-modal semantic consistency network,”J. Bionic. Eng., vol. 19, no. 6, pp. 1816–1829, 2022

    work page 2022

  22. [22]

    Analysis on leg bone fracture detection and classification using X-ray images,

    W. W. Myint, K. S. Tun, H. M. Tun, and H. Myint, “Analysis on leg bone fracture detection and classification using X-ray images,”Machine Learning Research, vol. 3, no. 3, pp. 49–59, 2018

    work page 2018

  23. [23]

    Local- entropy based approach for X-ray image segmentation and fracture detection,

    F. Hr ˇzi´c, I. ˇStajduhar, S. Tschauner, E. Sorantin, and J. Lerga, “Local- entropy based approach for X-ray image segmentation and fracture detection,”Entropy, vol. 21, no. 4, p. 338, 2019

    work page 2019

  24. [24]

    Fracture detection in wrist X- ray images using deep learning-based object detection models,

    F. Hardalac ¸, F. Uysal, O. Peker, M. C ¸ ic ¸eklida˘g, T. Tolunay, N. Tokg ¨oz, U. Kutbay, B. Demirciler, and F. Mert, “Fracture detection in wrist X- ray images using deep learning-based object detection models,”Sensors, vol. 22, no. 3, p. 1285, 2022

    work page 2022

  25. [25]

    Deep learning-based localization and segmentation of wrist fractures on X-ray radiographs,

    D. Joshi, T. P. Singh, and A. K. Joshi, “Deep learning-based localization and segmentation of wrist fractures on X-ray radiographs,”Neural Computing and Applications, vol. 34, no. 21, pp. 19 061–19 077, 2022

    work page 2022

  26. [26]

    The impact of implementing backbone architectures on fracture segmentation in X-ray images,

    S. Turk, O. Bingol, A. Coskuncay, and T. Aydin, “The impact of implementing backbone architectures on fracture segmentation in X-ray images,”Eng. Sci. Technol. Int. J., vol. 59, p. 101883, 2024

    work page 2024

  27. [27]

    Deep-learning-based pelvic automatic segmentation in pelvic fractures,

    J. M. Lee, J. Y . Park, Y . J. Kim, and K. G. Kim, “Deep-learning-based pelvic automatic segmentation in pelvic fractures,”Scientific Reports, vol. 14, no. 1, p. 12258, 2024

    work page 2024

  28. [28]

    Automatic detection and segmentation of lumbar vertebrae from X-ray images for compression fracture evaluation,

    K. C. Kim, H. C. Cho, T. J. Jang, J. M. Choi, and J. K. Seo, “Automatic detection and segmentation of lumbar vertebrae from X-ray images for compression fracture evaluation,”Computer Methods and Programs in Biomedicine, vol. 200, p. 105833, 2021

    work page 2021

  29. [29]

    Deep learning to segment pelvic bones: large-scale CT datasets and baseline models,

    P. Liu, H. Han, Y . Du, H. Zhu, Y . Li, F. Gu, H. Xiao, J. Li, C. Zhao, L. Xiaoet al., “Deep learning to segment pelvic bones: large-scale CT datasets and baseline models,”International Journal of Computer Assisted Radiology and Surgery, vol. 16, pp. 749–756, 2021

    work page 2021

  30. [30]

    Synthetic data accelerates the development of gen- eralizable learning-based algorithms for X-ray image analysis,

    C. Gao, B. D. Killeen, Y . Hu, R. B. Grupp, R. H. Taylor, M. Armand, and M. Unberath, “Synthetic data accelerates the development of gen- eralizable learning-based algorithms for X-ray image analysis,”Nature Machine Intelligence, vol. 5, no. 3, pp. 294–308, 2023

    work page 2023

  31. [31]

    Mitigating bias in radiology machine learning: 3. performance metrics,

    S. Faghani, B. Khosravi, K. Zhang, M. Moassefi, J. M. Jagtap, F. Nugen, S. Vahdati, S. P. Kuanar, S. M. Rassoulinejad-Mousavi, Y . Singhet al., “Mitigating bias in radiology machine learning: 3. performance metrics,” Radiology: Artificial Intelligence, vol. 4, no. 5, p. e220061, 2022

    work page 2022

  32. [32]

    Totalsegmen- tator: robust segmentation of 104 anatomic structures in CT images,

    J. Wasserthal, H.-C. Breit, M. T. Meyer, M. Pradella, D. Hinck, A. W. Sauter, T. Heye, D. T. Boll, J. Cyriac, S. Yanget al., “Totalsegmen- tator: robust segmentation of 104 anatomic structures in CT images,” Radiology: Artificial Intelligence, vol. 5, no. 5, p. e230024, 2023

    work page 2023

  33. [33]

    Methods and open-source toolkit for analyzing and visualizing challenge results,

    M. Wiesenfarth, A. Reinke, B. A. Landman, M. Eisenmann, L. A. Saiz, M. J. Cardoso, L. Maier-Hein, and A. Kopp-Schneider, “Methods and open-source toolkit for analyzing and visualizing challenge results,” Scientific reports, vol. 11, no. 1, p. 2369, 2021

    work page 2021

  34. [34]

    Pose estimation of pe- riacetabular osteotomy fragments with intraoperative X-ray navigation,

    R. B. Grupp, R. A. Hegeman, R. J. Murphy, C. P. Alexander, Y . Otake, B. A. McArthur, M. Armand, and R. H. Taylor, “Pose estimation of pe- riacetabular osteotomy fragments with intraoperative X-ray navigation,” IEEE Trans. Biomed. Eng., vol. 67, no. 2, pp. 441–452, 2019

    work page 2019

  35. [35]

    Fracformer: Fracture reduction planning with transformer-based shape restoration and fracture data simulation,

    S. Yibulayimu, Y . Liu, Y . Sang, J. Qin, C. Shi, C. Liang, G. Zhu, Y . Wang, C. Zhao, and X. Wu, “Fracformer: Fracture reduction planning with transformer-based shape restoration and fracture data simulation,” IEEE Transactions on Medical Imaging, 2025

    work page 2025

  36. [36]

    Epidemiology and burden of pelvic fractures: Results from the global burden of disease study 2019,

    S. Hu, J. Guo, B. Zhu, Y . Dong, and F. Li, “Epidemiology and burden of pelvic fractures: Results from the global burden of disease study 2019,” Injury, vol. 54, no. 2, pp. 589–597, 2023

    work page 2019