Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

UltraBones100k: A reliable automated labeling method and large-scale dataset for ultrasound-based bone surface extraction

T0 review · 4 major / 5 minor · reviewed 2026-08-09 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Tracked CT models can label ultrasound bone surfaces at 100k-image scale, producing labels that beat manual labeling in low-intensity regions.

desk verdict The dataset and labeling method are real resources; the headline superiority-over-surgeon claim is not established because the evaluation uses the pipeline's own CT-derived labels as ground truth and the manual protocol excludes low-intensity regions. read the letter →

arxiv 2502.03783 v4 pith:YHMJSSPF submitted 2025-02-06 eess.IV cs.CV

Luohong Wu , Nicola A. Cavalcanti , Matthias Seibold , Giuseppe Loggia , Lisa Reissner , Jonas Hein , Silvan Beeler , Arnd Viehöfer
show 3 more authors
Stephan Wirth Lilian Calvet Philipp Fürnstahl
This is my paper · ORCID
classification eess.IVcs.CV
keywords ultrasoundbonesegmentationsurfaceextractionautomatedlabelingCT-ultrasoundregistrationopticaltrackingdeeplearningdatasetcomputer-assistedorthopedicsurgery
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper claims that bone surfaces in ultrasound can be labeled automatically and at large scale by superimposing a tracked CT bone model onto tracked ultrasound frames, without relying on costly expert manual annotation. The authors collected UltraBones100k, the largest known public dataset of over 100,000 ex-vivo human lower-limb ultrasound images with bone labels. A U-Net trained on these labels outperforms an expert surgeon on every reported metric, especially in low-intensity and anechoic bone regions where manual labeling is incomplete or impossible. If the claim holds, it would remove the main bottleneck to training and benchmarking ultrasound bone-segmentation models for computer-assisted orthopedic surgery.

What carries the argument

The load-bearing mechanism is the per-frame pose refinement in Eq. (2): small perturbations ($\boldsymbol{\theta}_\epsilon$, $\mathbf{d}_\epsilon$) are added to each ultrasound image pose so that the mean intensity of pixels lying on the CT-derived bone intersection is maximized, while a regularization term keeps the corrected pose close to the original optical tracking data. The optimization is solved with differential evolution, and failed optimizations are detected and discarded via a fiducial-correction limit. Bone-shadow regions are removed, and only bone pixels with incidence angle $\alpha \leq 85^\circ$ are retained, reflecting when ultrasound can actually return a surface echo. Evaluation uses distance-thresholded accuracy, completeness, and F1 score (Eqs. (3)-(6)), with high- and low-intensity regions separated by Otsu thresholding.

What would settle it

On the same 700 test frames, have several ultrasound-trained surgeons label all visible bone including low-intensity and anechoic regions following a protocol without the instruction to restrict to hyperechoic layers, and evaluate all labels against an independent ground truth such as micro-CT of the scanned specimen. If the model's completeness gain over manual labels at the 0.5 mm threshold drops substantially below the reported 320%, the central claim of superiority over manual labeling fails; alternatively, if an independent CT-ultrasound alignment on held-out specimens yields a mean surface distance larger than 2 mm for this pipeline's labels, the optimization's core error assumption is violated.

Watch

Extended reading notes

Core claim

The paper claims that automatically generated CT-derived labels, refined per frame by a local intensity-based optimization, are accurate enough to serve as ground truth for bone-surface segmentation and as training labels for deep models. After optical tracking and ultrasound calibration place each ultrasound pixel in CT space, the pipeline samples the CT bone surface, computes its intersection with the ultrasound image plane, and perturbs each frame's six-degree-of-freedom pose to maximize the mean intensity of intersected pixels while penalizing deviation from the recorded fiducial positions. After removing frames whose fiducial correction exceeds 1 mm (about 18.7% of frames), the remaining labels form UltraBones100k. A U-Net trained on this dataset achieves, at a 0.5 mm distance threshold, accuracy 0.900, completeness 0.895, and F1 0.893 across all bone regions; in low-intensity regions it improves completeness by 320% (0.190 to 0.801), accuracy by 27.4%, and F1 by 197% relative to an expert surgeon's manual labels.

Load-bearing premise

The evaluation treats the CT-derived labels produced by the paper's own pipeline as ground truth for both training and comparison, so the claim that the model beats manual labeling depends on those derived labels actually being the true bone surface.

Editorial extensions

If this is right

  • A segmentation model can be trained without expert annotation on a dataset roughly two orders of magnitude larger than typical manual-label sets, so model scale is no longer gated by labeling cost.
  • Because the model recovers bone contours in low-intensity and anechoic regions, downstream tasks such as 3D bone-surface reconstruction and CT-ultrasound registration can start from more complete 2D segmentations.
  • The public dataset and pretrained weights give the community a common benchmark, helping to standardize the more than 18 evaluation metrics currently used in ultrasound bone segmentation.
  • Qualitative results on pelvis, femur, and in-vivo fibula and tibia suggest the learned features transfer beyond the lower-leg ex-vivo training distribution, though the paper presents these results as qualitative rather than quantitative.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the per-frame optimization generalizes to other transducers and setups, the same tracking-plus-refinement recipe could generate labeled ultrasound datasets for anatomies like the spine or skull where manual labeling is even harder, while the paper's own limitation section warns that very different bone-soft-tissue structures may require fine-tuning.
  • The 18.7% frame-rejection rule acts as a quality filter; one testable extension is whether keeping only frames with small fiducial corrections is itself sufficient for training, without the intensity-maximization term.
  • The reported advantage over the surgeon may shrink if a neutral ground truth, such as an independent micro-CT scan of the same specimen, is used instead of the pipeline's own CT-derived labels, because the current evaluation compares manual labels against those same derived labels.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper introduces UltraBones100k, a public dataset of about 100k ex-vivo ultrasound images of human lower limbs with automated bone-surface labels. Labels are obtained by optically tracking the ultrasound probe and a cadaver marker, projecting a CT-derived bone model into each image, and then refining per-frame pose via an intensity-based optimization (Eq. 2). Bone-shadow and incidence-angle masks are applied. The authors report a two-rater clinical evaluation showing that the refined alignment improves over the initial tracking alignment and over a global CT-US registration method. They then train a U-Net on the generated labels and compare its predictions with manual annotations by an expert surgeon on 700 test images, reporting that the model outperforms manual labeling on accuracy, completeness, and F1-score, especially in low-intensity regions (e.g., 320% completeness improvement at 0.5 mm).

Significance. If the results hold, the paper would make a substantial contribution: it provides the largest public ultrasound bone-segmentation dataset to date, releases code and pretrained model weights, and demonstrates a practical pipeline that avoids costly expert labeling. The per-frame alignment refinement with a clinical rating study is a useful methodological step over previous global registration approaches. The claim that a model trained on automatically generated labels can segment bone surfaces more completely than an expert, particularly in low-intensity regions, is scientifically important but currently rests on an internal evaluation design that is not fully independent. The dataset release itself is valuable regardless of the outcome of the model-versus-manual comparison.

major comments (4)
  1. [Section 3.2 (Evaluation Metrics), Section 3.1 (Automatic labeling)] The central quantitative comparison uses the CT-derived labels as ground truth for evaluating both the trained model and the manual labels. The text states that the evaluation is based on the CT-derived labels (Section 3.2), and these labels are generated by the same tracked-CT projection and Eq. (2) optimization used to produce the training data. Consequently, the reported accuracy, completeness, and F1 scores measure agreement with the pipeline's own output, not anatomical correctness. In low-intensity and anechoic regions, where the image provides little or no bright echo, the optimization in Eq. (2) has no intensity evidence to constrain the solution, so the label reduces essentially to the tracked CT projection. The reported 320% completeness improvement in low-intensity regions therefore requires an independent geometric or manual reference before it can be interpreted as superiority of the model over human labeling.
  2. [Section 3.2 (Manual annotation)] The manual labeling protocol instructed the expert surgeon to identify the hyperechoic bone surface layer and its acoustic shadow, explicitly omitting low-intensity and anechoic bone regions. The Discussion concedes that manual labels mostly include only high-intensity bone regions. The dramatic completeness gap in low-intensity regions (0.190 manual vs. 0.801 model) is thus largely created by the labeling protocol rather than by an inherent limitation of human performance. A fair comparison requires either manual labels that also cover low-intensity regions (with a protocol allowing anatomical inference) or a separate evaluation restricted to regions where both manual and automatic labels are meaningful.
  3. [Section 3.1 (Validation step, frame discard)] Approximately 18.7% of frames are discarded after optimization because the mean fiducial correction exceeds 1 mm. This selection step removes frames where the tracking-based alignment and the intensity correction disagree, and the test set is drawn from the same post-filtered pipeline. The paper does not analyze the discarded frames or report how the model behaves on them. If the discarded frames are systematically different (e.g., curved bone regions, poor acoustic coupling), the reported segmentation performance may not generalize to the full acquisition stream, and the low-intensity-region claim could be affected by this selection bias.
  4. [Section 3.1 (Clinical Evaluation)] The clinical evaluation on 119 frames provides useful evidence that the optimized alignment is visually preferred over the initial and global-registration alignments, and the Wilcoxon test with Bonferroni correction supports that specific claim. However, this evaluation was designed to rate alignment quality using visible errors, and it does not independently validate the correctness of labels in anechoic or low-intensity regions, where the expert cannot directly see the bone surface. The p<0.001 result therefore does not by itself establish that the low-intensity labels used as ground truth are anatomically correct.
minor comments (5)
  1. [Section 3.1 (Clinical Evaluation)] Typo: 'tow expert physician' should be 'two expert physicians.'
  2. [Section 3.1 (Dataset collection)] The sentence 'all ultrasound scanning was conducted on fully thawed, all ultrasound scanning was performed on fully thawed' contains a duplicated and incomplete phrase; it should read 'all ultrasound scanning was performed on fully thawed, fresh-frozen human tissue.'
  3. [Section 3.2 (Manual annotation)] The description of the manual labeling protocol would benefit from reporting the actual time taken per image and any intra- or inter-rater variability, as the Limitations section itself notes that multiple-expert annotations are valuable.
  4. [Section 3.1 (Eq. 2)] The optimization parameter lambda is selected from a finite set by testing on randomly sampled frames, but the paper does not state how many frames were used for this selection or whether the selected lambda was fixed before the clinical evaluation. A brief description of the selection procedure would improve reproducibility.
  5. [Section 3.2 (Generalizability)] The generalizability figures (Figs. 8 and 9) are qualitative only; a sentence clarifying that these are illustrative and not quantitatively evaluated would prevent overinterpretation.

Circularity Check

2 steps flagged · score 7.0 of 10

The central model-vs-manual result is largely predetermined: the evaluation ground truth is the same CT-derived label set used to train the model, and the manual-labeling protocol excludes low-intensity regions by design, so the reported low-intensity completeness gap is built into the evaluation.

  1. self definitional [Section 3.2, Bone segmentation and Evaluation Metrics; Abstract/Results]
    "A U-net with ResNet encoders is trained on UltraBones100k to automatically segment bones in ultrasound images. ... Particularly, our evaluation is based on the CT-derived labels. ... The model trained on UltraBones100k consistently outperforms manual labeling in all metrics, particularly in low-intensity regions (at a distance threshold of 0.5 mm: 320% improvement in completeness, 27.4% improvement in accuracy, and 197% improvement in F1 score)."

    The model is trained to reproduce the CT-derived automatic labels, and the same CT-derived label set is then used as the ground truth in Eqs. (3)-(6) when comparing the model against manual labeling. The comparison therefore measures how closely each method reproduces the pipeline's own label generator. A model explicitly optimized to output those labels is expected to win; the 'superiority' is forced by the choice of evaluator and does not establish that the CT-derived labels are anatomically correct, especially in low-intensity regions.

  2. other [Section 3.2, Manual annotation; Section 5, Automatic vs. Manual Labeling]
    "The surgeon identified the hyperechoic bone surface layer and its corresponding acoustic shadow based on the following knowledge about ultrasound imaging and anatomical structures. ... Manual labels mostly include only high-intensity bone regions."

    The manual labeling protocol instructed the surgeon to label only the hyperechoic bone surface layer and its acoustic shadow, so low-intensity and anechoic bone regions were excluded from manual labels by design. Since the CT-derived labels, which include low-intensity regions, are used as the completeness denominator in Eq. (5), manual completeness in the low-intensity partition is near zero by protocol rather than by impartial measurement. The reported 320% completeness gain is therefore largely a consequence of differing labeling definitions, not a validated superiority over the surgeon.

full rationale

The paper's automatic labeling pipeline is mostly self-contained: tracked CT models, ultrasound calibration, and the intensity-based optimization of Eq. (2) are presented with measurable calibration error and are supported by an independent expert alignment-rating experiment (Table 2). That part is not circular. The circularity concerns the headline segmentation claim. The same CT-derived label set is both the training target and the evaluation ground truth ('our evaluation is based on the CT-derived labels'), and the manual protocol excludes low-intensity regions, so the model-vs-manual comparison is not an independent test of label quality. The magnitude of the reported low-intensity improvement is predetermined by the evaluation design more than by measured anatomical correctness. The existence of the expert alignment ratings and the qualitative generalizability results prevent this from being a complete 10, but the central quantitative claim is substantially circular.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

The central contribution is a dataset and an alignment pipeline, not a new physical entity; no invented particles, forces, or modalities. The ledger instead shows five hand-set parameters and several domain assumptions about ultrasound physics and CT accuracy. The most consequential items are the use of the pipeline's own CT-derived labels as ground truth and the asymmetric manual labeling protocol, which jointly create the circularity in the main comparison.

free parameters (5)
  • gamma (intersection thickness) = 0.3 mm
    Defines how close a CT model point must be to a projected ultrasound point to count as bone surface; chosen empirically in Section 3.1 (Initial Bone Labeling).
  • lambda (regularization weight) = selected from [0.01, 0.1, 1, 10, 100, 1000]
    Balances image intensity term against deviation from tracked fiducials in Eq. (2); optimal value selected by testing random frames in Section 3.1 (Optimization Step).
  • incidence angle cutoff alpha = 85 degrees
    Bone pixels with incidence angle <= 85 degrees are kept to avoid boundary cases; chosen to account for ultrasound physics in Section 3.1.
  • frame discard threshold = 1 mm mean fiducial correction; 18.7% frames discarded
    Frames whose optimized pose correction exceeds 1 mm are removed as failed optimizations in Section 3.1, which can bias the dataset toward well-tracked frames.
  • Otsu intensity threshold = computed from label pixel intensities
    Used to split labels into high- and low-intensity regions for the evaluation in Section 3.2; the split is data-dependent and determines how the completeness gap is measured.
assumptions (6)
  • domain assumption CT bone models derived from Mimics segmentation are anatomically accurate and serve as the reference for bone location.
    The entire labeling pipeline projects CT bone surfaces into ultrasound images; no independent validation of the CT segmentation is reported.
  • domain assumption Bones and attached K-wires/markers do not move between CT and ultrasound acquisition and during scanning.
    K-wires were used to stabilize bones, but a single rigid transformation is assumed to link CT and tracked ultrasound for each specimen.
  • domain assumption Ultrasound appearance of bone is governed by the stated physics: high intensity when perpendicular, shadow beneath, intensity decreasing with incidence angle beyond 85 degrees.
    Used in optimization objective and exclusion of bone shadow/oblique regions; plausible but not independently measured in this paper.
  • domain assumption Initial CT-ultrasound alignment error is below 2 mm and mostly caused by tracking/calibration errors.
    Stated as assumption (iii) in Section 3.1, supported by a small random sample (mean 1.13 +/- 0.606 mm) rather than a systematic evaluation.
  • ad hoc to paper CT-derived labels are a valid ground truth for evaluating both the trained model and manual labels.
    Section 3.2 Evaluation Metrics states evaluation is based on CT-derived labels; this is the circularity that underwrites the model-vs-manual comparison.
  • ad hoc to paper Manual labeling protocol is comparable to automatic labeling despite being restricted to the hyperechoic bone surface layer and acoustic shadow.
    Manual annotation instructions in Section 3.2 exclude low-intensity/anechoic bone regions, making the low-intensity completeness comparison one-sided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of UltraBones100k: A reliable automated labeling method and large-scale dataset for ultrasound-based bone surface extraction." pith.science (2026). https://pith.science/paper/YHMJSSPF

@misc{pith2026250203783,
  author       = {Pith},
  title        = {Pith review of: UltraBones100k: A reliable automated labeling method and large-scale dataset for ultrasound-based bone surface extraction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YHMJSSPF}},
  note         = {Machine review of arXiv:2502.03783}
}
read the original abstract

Ultrasound-based bone surface segmentation is crucial in computer-assisted orthopedic surgery. However, ultrasound images have limitations, including a low signal-to-noise ratio, and acoustic shadowing, which make interpretation difficult. Existing deep learning models for bone segmentation rely primarily on costly manual labeling by experts, limiting dataset size and model generalizability. Additionally, the complexity of ultrasound physics and acoustic shadow makes the images difficult for humans to interpret, leading to incomplete labels in anechoic regions and limiting model performance. To advance ultrasound bone segmentation and establish effective model benchmarks, larger and higher-quality datasets are needed. We propose a methodology for collecting ex-vivo ultrasound datasets with automatically generated bone labels, including anechoic regions. The proposed labels are derived by accurately superimposing tracked bone CT models onto the tracked ultrasound images. These initial labels are refined to account for ultrasound physics. A clinical evaluation is conducted by an expert physician specialized on orthopedic sonography to assess the quality of the generated bone labels. A neural network for bone segmentation is trained on the collected dataset and its predictions are compared to expert manual labels, evaluating accuracy, completeness, and F1-score. We collected the largest known dataset of 100k ultrasound images of human lower limbs with bone labels, called UltraBones100k. A Wilcoxon signed-rank test with Bonferroni correction confirmed that the bone alignment after our method significantly improved the quality of bone labeling (p < 0.001). The model trained on UltraBones100k consistently outperforms manual labeling in all metrics, particularly in low-intensity regions (320% improvement in completeness at a distance threshold of 0.5 mm).

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. SonoGym: High Performance Simulation for Challenging Surgical Tasks with Robotic Ultrasound

    cs.RO 2025-07 conditional novelty 6.0 of 10

    SonoGym provides parallel, realistic ultrasound simulation for training RL and imitation-learning agents on robotic orthopedic tasks including navigation, reconstruction, and surgery.

Reference graph

Works this paper leans on

41 extracted references · 34 canonical work pages · cited by 1 Pith paper

  1. [12]

    van der Zee, M

    J.M. van der Zee, M. Fitski, M.A.J. van de Sande, M.A.D. Buser, M.A.J. Hiep, C.E.J. Terwisscha van Scheltinga, C.C.C. Hulsker, C.H. van den Bosch, C.P. van de Ven, L. van der Heijden, G.M.J. Bökkerink, M.H.W.A. Wijnen, F.J. Siepel, A.F.W. van der Steeg, Tracked ultrasound registration for intraoperative navigation during pediatric bone tumor resections wi...

    work page 2023

  2. [1]

    Blankstein, Ultrasound in the diagnosis of clinical orthopedics: The orthopedic stethoscope

    A. Blankstein, Ultrasound in the diagnosis of clinical orthopedics: The orthopedic stethoscope. World j orthop 2: 13–24, 2011

    work page 2011

  3. [2]

    Khoury, E

    V. Khoury, E. Cardinal, N.J. Bureau, Musculoskeletal sonography: a dynamic tool for usual and unusual disorders, Am. J. Roentgenol. 188 (1) (2007) W63–W73

    work page 2007

  4. [3]

    Garcia, W.J

    T. Garcia, W.J. Hornof, M.F. Insana, On the ultrasonic properties of tendon, Ultrasound Med. Biol. 29 (12) (2003) 1787–1797

    work page 2003

  5. [4]

    Hacihaliloglu, 3D ultrasound for orthopedic interventions, Intell

    I. Hacihaliloglu, 3D ultrasound for orthopedic interventions, Intell. Orthop.: Artif. Intell. Smart Image- Guid. Technol. Orthop. (2018) 113–129

    work page 2018

  6. [5]

    Boctor, M.A

    E.M. Boctor, M.A. Choti, E.C. Burdette, R.J. Webster Iii, Three-dimensional ultrasound-guided robotic needle placement: an experimental evaluation, Int. J. Med. Robot. Comput. Assist. Surg. 4 (2) (2008) 180–191

    work page 2008

  7. [6]

    Hacihaliloglu, P

    I. Hacihaliloglu, P. Guy, A.J. Hodgson, R. Abugharbieh, Automatic extraction of bone surfaces from 3D ultrasound images in orthopaedic trauma cases, Int. J. Comput. Assist. Radiol. Surg. 10 (2015) 1279–1287

    work page 2015

  8. [7]

    R. Li, A. Davoodi, Y. Cai, K. Niu, G. Borghesan, N. Cavalcanti, A. Massalimova, F. Carrillo, C.J. Laux, M. Farshad, P. Fürnstahl, E.V. Poorten, Robot-assisted ultrasound reconstruction for spine surgery: from bench-top to pre-clinical study, Int. J. Comput. Assist. Radiol. Surg. 18 (9) (2023) 1613–1623, URL http://dx. doi.org/10.1007/s11548-023-02932-z

    doi:10.1007/s11548-023-02932-z 2023

Show all 41 references
  1. [8]

    Salehi, R

    M. Salehi, R. Prevost, J.-L. Moctezuma, N. Navab, W. Wein, Precise ultrasound bone registration with learning-based segmentation and speed of sound calibra- tion, 2017, Paper presented at Medical Image Computing and Computer-Assisted Intervention - MICCAI

    2017

  2. [9]

    R. Li, A. Davoodi, Y. Cai, K. Niu, G. Borghesan, N. Cavalcanti, A. Massalimova, F. Carrillo, C.J. Laux, M. Farshad, et al., Robot-assisted ultrasound reconstruction for spine surgery: from bench-top to pre-clinical study, Int. J. Comput. Assist. Radiol. Surg. 18 (9) (2023) 1613–1623

    2023

  3. [10]

    Ciganovic, F

    M. Ciganovic, F. Ozdemir, F. Pean, P. Fuernstahl, C. Tanner, O. Goksel, Regis- tration of 3D freehand ultrasound to a bone model for orthopedic procedures of the forearm, Int. J. Comput. Assist. Radiol. Surg. 13 (6) (2018) 827–836, URL http://dx.doi.org/10.1007/s11548-018-1756-0

    2018 doi

  4. [11]

    Gebhardt, L

    C. Gebhardt, L. Göttling, L. Buchberger, C. Ziegler, F. Endres, Q. Wuermeling, B.M. Holzapfel, W. Wein, F. Wagner, O. Zettinig, Femur reconstruction in 3D ultrasound for orthopedic surgery planning, Int. J. Comput. Assist. Radiol. Surg. 18 (6) (2023) 1001–1008, URL http://dx.d...

    2023 doi

  5. [13]

    H. Chen, L. Kumaralingam, S. Zhang, S. Song, F. Zhang, H. Zhang, T.-T. Pham, K. Punithakumar, E.H. Lou, Y. Zhang, L.H. Le, R. Zheng, Neural implicit surface reconstruction of freehand 3D ultrasound volume with geometric constraints, Med. Image Anal. 98 (2024) 103305, URL https...

    2024

  6. [14]

    Pandey, P

    P. Pandey, P. Guy, A.J. Hodgson, R. Abugharbieh, Fast and automatic bone segmentation and registration of 3D ultrasound to CT for the full pelvic anatomy: a comparative study, Int. J. Comput. Assist. Radiol. Surg. 13 (10) (2018) 1515–1524, URL http://dx.doi.org/10.1007/s11548-...

    2018 doi

  7. [15]

    Quader, A.J

    N. Quader, A.J. Hodgson, K. Mulpuri, E. Schaeffer, R. Abugharbieh, Automatic evaluation of scan adequacy and dysplasia metrics in 2-d ultrasound images of the neonatal hip, Ultrasound Med. Biol. 43 (6) (2017) 1252–1262, URL http://dx.doi.org/10.1016/j.ultrasmedbio.2017.01.012

    2017 doi

  8. [16]

    N. Baka, S. Leenstra, T. van Walsum, Ultrasound aided vertebral level localization for lumbar surgery, IEEE Trans. Med. Imaging 36 (10) (2017) 2138–2147, URL http://dx.doi.org/10.1109/TMI.2017.2738612

    2017

  9. [17]

    T. Ungi, H. Greer, K.R. Sunderland, V. Wu, Z.M.C. Baum, C. Schlenger, M. Oetgen, K. Cleary, S.R. Aylward, G. Fichtinger, Automatic spine ultrasound segmentation for scoliosis visualization and measurement, IEEE Trans. Biomed. Eng. 67 (11) (2020) 3234–3241

    2020

  10. [18]

    El-Hariri, K

    H. El-Hariri, K. Mulpuri, A. Hodgson, R. Garbi, Comparative evaluation of hand- engineered and deep-learned features for neonatal hip bone segmentation in ultrasound, in: Medical Image Computing and Computer Assisted Intervention – MICCAI 2019, Springer International Publishin...

    2019

  11. [19]

    Villa, G

    M. Villa, G. Dardenne, M. Nasan, H. Letissier, C. Hamitouche, E. Stindel, FCN- based approach for the automatic segmentation of bone surfaces in ultrasound images, Int. J. Comput. Assist. Radiol. Surg. 13 (11) (2018) 1707–1716, URL http://dx.doi.org/10.1007/s11548-018-1856-x

    2018 doi

  12. [20]

    Alsinan, V.M

    A.Z. Alsinan, V.M. Patel, I. Hacihaliloglu, Automatic segmentation of bone surfaces from ultrasound using a filter-layer-guided CNN, Int. J. Comput. Assist. Radiol. Surg. 14 (5) (2019) 775–783, URL http://dx.doi.org/10.1007/s11548- 019-01934-0

    2019 doi

  13. [21]

    Pandey, N

    P.U. Pandey, N. Quader, P. Guy, R. Garbi, A.J. Hodgson, Ultrasound bone seg- mentation: A scoping review of techniques and validation practices., Ultrasound Med. Biol. (2020) URL https://api.semanticscholar.org/CorpusID:210924028

    2020

  14. [22]

    Hohlmann, P

    B. Hohlmann, P. Broessner, K. Radermacher, Ultrasound-based 3D bone mod- elling in computer assisted orthopedic surgery - a review and future challenges, Comput. Assist. Surg. ( Abingdon, England) 29 (2024) 2276055

    2024

  15. [23]

    W. Wein, S. Brunke, A. Khamene, M.R. Callstrom, N. Navab, Automatic CT- ultrasound registration for diagnostic imaging and image-guided intervention, Med. Image Anal. 12 (5) (2008) 577–585

    2008

  16. [24]

    Hacihaliloglu, R

    I. Hacihaliloglu, R. Abugharbieh, A.J. Hodgson, R.N. Rohling, Bone surface localization in ultrasound using image phase-based features, Ultrasound Med. Biol. 35 (9) (2009) 1475–1487

    2009

  17. [25]

    Hohlmann, P

    B. Hohlmann, P. Broessner, L. Phlippen, T. Rohde, K. Radermacher, Knee bone models from ultrasound, IEEE Trans. Ultrason. Ferroelectr. Freq. Control 70 (9) (2023) 1054–1063, URL http://dx.doi.org/10.1109/TUFFC.2023.3286287

    2023

  18. [26]

    Liebmann, D

    F. Liebmann, D. Stütz, D. Suter, S. Jecklin, J.G. Snedeker, M. Farshad, P. Fürnstahl, H. Esfandiari, SpineDepth: A multi-modal data collection approach for automatic labelling and intraoperative spinal shape reconstruction based on RGB-d data, J. Imaging 7 (9) (2021) 164, URL ...

    2021

  19. [27]

    J. Hein, N. Cavalcanti, D. Suter, L. Zingg, F. Carrillo, L. Calvet, M. Farshad, M. Pollefeys, N. Navab, P. Fürnstahl, Next-generation surgical navigation: Marker- less multi-view 6dof pose estimation of surgical instruments, 2023, URL https: //arxiv.org/abs/2305.03535

    2023 arXiv

  20. [28]

    Cignoni, M

    P. Cignoni, M. Callieri, M. Corsini, M. Dellepiane, F. Ganovelli, G. Ranzuglia, MeshLab: an Open-Source Mesh Processing Tool, in: V. Scarano, R.D. Chiara, U. Erra (Eds.), Eurographics Italian Chapter Conference, The Eurograph- ics Association, 2008, http://dx.doi.org/10.2312/L...

    2008 doi

  21. [29]

    Besl, N.D

    P.J. Besl, N.D. McKay, Method for registration of 3-D shapes, in: Sensor Fusion IV: Control Paradigms and Data Structures, Vol. 1611, Spie, 1992, pp. 586–606

    1992

  22. [30]

    Mozaffari, W.-S

    M.H. Mozaffari, W.-S. Lee, Freehand 3-D ultrasound imaging: A systematic review, Ultrasound Med. Biol. 43 (10) (2017) 2099–2124, URL https://www. sciencedirect.com/science/article/pii/S0301562917302776

    2017

  23. [31]

    R. Li, Y. Cai, K. Niu, E.V. Poorten, Comparative quantitative analysis of robotic ultrasound image calibration methods, in: 2021 20th International Conference on Advanced Robotics, ICAR, 2021, pp. 511–516

    2021

  24. [32]

    Storn, K

    R. Storn, K. Price, Differential evolution – a simple and efficient heuristic for global optimization over continuous spaces, J. Global Optim. 11 (4) (1997) 341–359, URL http://dx.doi.org/10.1023/A:1008202821328

    1997 doi

  25. [33]

    Ronneberger, P

    O. Ronneberger, P. Fischer, T. Brox, U-net: Convolutional networks for biomedi- cal image segmentation, in: Medical Image Computing and Computer-Assisted Intervention–MICCAI 2015: 18th International Conference, Munich, Germany, October 5-9, 2015, Proceedings, Part III 18, Spri...

    2015

  26. [34]

    Iakubovskii, Segmentation models pytorch, 2019, https://github.com/qubvel/ segmentation_models.pytorch,

    P. Iakubovskii, Segmentation models pytorch, 2019, https://github.com/qubvel/ segmentation_models.pytorch,

    2019

  27. [35]

    Otsu, et al., A threshold selection method from gray-level histograms, Autom

    N. Otsu, et al., A threshold selection method from gray-level histograms, Autom. 11 (285–296) (1975) 23–27

    1975

  28. [36]

    Holland, L

    J.P. Holland, L. Waugh, A. Horgan, V. Paleri, D.J. Deehan, Cadaveric hands- on training for surgical specialties: is this back to the future for surgical skills development? J. Surg. Educ. 68 (2) (2011) 110–116

    2011

  29. [37]

    H. Yang, C. Shan, T. Tan, A.F. Kolen, P.H. de With, Transferring from ex-vivo to in-vivo: Instrument localization in 3d cardiac ultrasound using pyramid-unet with hybrid loss, in: Medical Image Computing and Computer Assisted Intervention– MICCAI 2019: 22nd International Confe...

    2019

  30. [38]

    De Rosa, S

    L. De Rosa, S. L’Abbate, C. Kusmic, F. Faita, Applications of deep learning algorithms to ultrasound imaging analysis in preclinical studies on in vivo animals, Life 13 (8) (2023) 1759

    2023

  31. [39]

    Isensee, P.F

    F. Isensee, P.F. Jaeger, S.A. Kohl, J. Petersen, K.H. Maier-Hein, Nnu-net: a self-configuring method for deep learning-based biomedical image segmentation, Nature Methods 18 (2) (2021) 203–211

    2021

  32. [40]

    G. Chen, L. Li, Y. Dai, J. Zhang, M.H. Yap, AAU-net: an adaptive attention U-net for breast lesions segmentation in ultrasound images, IEEE Trans. Med. Imaging 42 (5) (2022) 1289–1300

    2022

  33. [41]

    G. Chen, L. Li, J. Zhang, Y. Dai, Rethinking the unpretentious U-net for medical ultrasound image segmentation, Pattern Recognit. 142 (2023) 109728. Computers in Biology and Medicine 194 (2025) 110435 11

    2023

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.