Pith. sign in

REVIEW 4 major objections 6 minor 33 references

Learning Dense 2D-3D Correspondence for X-ray-to-CT Registration of Knee Bones

T0 review · 4 major / 6 minor · reviewed 2026-08-01 · deepseek-v4-flash

Pith's one-line read A single shared-weight network trained on 758 patients registers knee bones from one X-ray and the patient's CT, with pose recovered in closed form from a global, initialization-free PnP+RANSAC solve.

desk verdict Solid synthetic story and a sensible transmission-aware loss, but the real-radiograph section doesn't run the claimed closed-form PnP pipeline — that gap should be fixed before publication. read the letter →

arxiv 2607.22803 v1 pith:D566VDKN submitted 2026-07-24 eess.IV cs.CV

classification eess.IVcs.CV
keywords X-ray-to-CTregistration2D-3Dcorrespondenceposeestimationkneeself-supervisedlearningInfoNCElossPnPtransmission-aware
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 tries to establish that a single shared-weight neural network can register a patient's knee bones from one plain radiograph and their pre-operative CT, without any per-patient fitting, initialization, or test-time rendering. The key move is to learn a dense 2D–3D correspondence embedding: a 3D network embeds every CT bone point and a 2D network embeds every X-ray pixel into a common space, supervised only by the known projection geometry of synthetic X-rays. At test time, matching the two sets of features and solving PnP inside RANSAC yields the six-degree-of-freedom pose in closed form. On patients never seen in training the paper reports median mTRE of 6.5 mm for the femur and 4.7 mm for the patella in single view, with in-plane error sub-millimetre and the residual almost entirely depth ambiguity; a second calibrated view cuts femur error to 2.4 mm with 75% success. The authors also claim the embedding is semantically meaningful, carrying a label-free bone identity and landmark information, and they present real-radiograph results as a feasibility study with in-plane error still not clinical.

What carries the argument

The central mechanism is a two-encoder architecture trained jointly into a shared embedding space: a 3D U-Net over the CT density volume reads per-point features by trilinear sampling, and a 2D U-Net over the radiograph produces per-pixel features. A transmission-aware multi-positive InfoNCE loss pulls each pixel's feature toward the features of every bone point along its viewing ray, matching the additive physics of X-ray image formation. At test time, pixel features are matched to CT point features directly, and the pose is recovered by a global, initialization-free, render-free PnP+RANSAC solve — no pose-gradient optimization enters the loop.

What would settle it

On a held-out set of real radiographs paired with segmented CTs, measure the median in-plane mTRE after fine-tuning only the 2D encoder. If it does not approach the roughly 1.3 mm seen on synthetic DRRs and instead stays near the 9.4 mm pre-fine-tuning level, the claim that the correspondence space transfers to real films — the load-bearing transfer step — fails.

Watch

Extended reading notes

Core claim

The central claim is that dense, subject-agnostic 2D–3D correspondence between a single radiograph and a patient's segmented pre-operative CT can be learned purely from projection geometry across a large cohort, and that on unseen patients the 6-DoF pose follows from a closed-form PnP+RANSAC solve on matched pixel–point pairs — no per-patient optimization, no test-time rendering, no initialization. The correspondence is supervised with a multi-positive InfoNCE loss that treats all bone points along a ray as positive targets, reflecting X-ray transmission. The authors report that on held-out femurs the in-plane error is sub-millimetre while the remaining error is almost entirely depth, which

Load-bearing premise

The load-bearing premise is that a model trained on synthetic digitally reconstructed radiographs — with only the 2D pathway having domain randomization — embeds real X-rays into the same correspondence space as the CT features well enough that the matching and PnP solve works despite scatter, beam hardening, and overlying soft tissue.

Editorial extensions

If this is right

  • One shared-weight model per bone registers patients never seen during training, without per-patient optimization or initialization, directly supporting the subject-agnostic claim.
  • Pose is recovered in closed form at roughly 6 ms per registration, against about 1.9 s for a differentiable-rendering baseline, and is robust to any initialization offset instead of collapsing within a few degrees.
  • The residual single-view error is dominated by depth, not correspondence quality; a second calibrated view fused in a multi-view PnP solve lowers femur median mTRE from 6.5 mm to 2.4 mm and raises success rate to 75%.
  • Trained only to register, the embedding separates the knee's bones with 96% 2D–3D consistency on held-out patients and identifies anatomical landmarks across patients at 98% region accuracy, enabling CT segmentation transfer onto the radiograph.
  • On real plain radiographs registration transfers after light fine-tuning of the 2D encoder, cutting in-plane error from 9.4 mm to 4.6 mm, though the paper labels this a feasibility study rather than clinical-grade accuracy.

Reading between the lines

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

  • Editorial inference: If the representation transfers across the knee as reported, the same correspondence-then-PnP recipe should extend to other joints with plain-film visibility (hip, wrist, spine) given segmented CTs, but the paper only demonstrates knee bones.
  • Editorial inference: The near-geometric depth limit suggests the clinically relevant deployment is the two-view configuration; a prospective study on real films with two calibrated views would test whether the reported 2.4 mm holds outside DRR evaluation.
  • Editorial inference: The label-free bone identity could be developed into a self-supervised segmentation-transfer tool, letting a CT parcellation be read off an X-ray without manual annotation, but the paper demonstrates this only qualitatively.
  • Editorial inference: Because the sim-to-real gap is confined to the 2D encoder by design, the method is compatible with future physics-based rendering that simulates scatter and beam hardening, which could close the remaining gap without retraining the correspondence space.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes an amortized, subject-agnostic dense 2D–3D correspondence method for registering knee bones (femur, patella) from a single calibrated radiograph to a patient's segmented CT. A 3D U-Net embeds CT points and a 2D U-Net embeds radiograph pixels into a shared space, trained with a multi-positive InfoNCE loss that accounts for X-ray transmission by treating all points along a ray as positives. At test time the authors propose to match pixel and point features and recover pose via a closed-form, initialization-free PnP+RANSAC solve. Experiments on synthetic DRRs from 758 training and 8 held-out patients report single-view median mTRE of 6.5 mm (femur) and 4.7 mm (patella), improving to 2.4 mm with two views. The paper also reports a capture-range/cost comparison against differentiable-rendering registration, a feasibility study on real radiographs (in-plane error 9.4 mm before and 4.6 mm after fine-tuning the 2D encoder), and analyses of emergent bone identity and anatomical landmark structure in the learned embeddings.

Significance. If the central claims are validated, the paper would make a useful contribution: it replaces per-patient iterative registration with an amortized, render-free PnP-based solve, introduces a transmission-aware correspondence target that is physically appropriate for X-ray, and provides an informative decomposition of error into in-plane and depth components. The capture-range comparison against differentiable rendering is well designed and the cost advantage (about 315×) is noteworthy. The emergence of bone identity and landmark information from a purely geometric training signal is a genuinely interesting result, probed with a frozen nearest-centroid classifier rather than an overfitted detector. The paper is also unusually candid about its limitations. However, the evidence base is narrow: only 8 held-out patients from a single institution, no confidence intervals, no code or data release, and—most importantly—the real-radiograph evaluation does not use the closed-form PnP pipeline that the abstract and introduction identify as the central methodological contribution.

major comments (4)
  1. [§4.5 and Abstract] The abstract and introduction state that pose follows in closed form from a global, initialization-free PnP+RANSAC solve for unseen patients. In the only real-radiograph evaluation, this pipeline is not used. §4.5 states that 'discrete correspondence+PnP, robust on synthetic data, is brittle on one real planar view' and instead recovers pose by 'optimizing the learned-feature cost in pose'—an iterative, per-patient, initialization-dependent procedure. The reported real-film in-plane error (9.4 mm before fine-tuning, 4.6 mm after) therefore does not validate the closed-form claim on the target modality. Either demonstrate that PnP+RANSAC works on real films after appropriate domain adaptation, or revise the central claim to be explicitly scoped to synthetic DRRs, presenting the real-film feature-space optimization as a separate fallback.
  2. [§4.3, Table 1] The cross-patient generalization claim is load-bearing but rests on only 8 held-out patients from a single institution, with DRRs rendered from the same CT the 3D encoder sees. The text asserts 'no significant generalization gap' without any confidence intervals, error bars, or significance tests. Given the heavy-tailed nature of mTRE, per-patient/per-pose distributions or bootstrap CIs are needed to support this claim. The absence of code or data also makes the headline numbers impossible to verify independently.
  3. [§4.5] The fine-tuning protocol is ambiguous and potentially invalid. The text says 'Fine-tuning only the 2D encoder on the real films, with the correspondence space frozen, cuts the held-out in-plane error from a median 9.4 to 4.6 mm.' It is not stated whether the fine-tuning set includes the same patients or views that are later evaluated as 'held-out.' If there is overlap, the improvement is a leak; if not, the split rule and number of fine-tuning subjects should be reported. Since this is the only real-data improvement, the details are essential.
  4. [§4.3, two-view result] The two-view PnP experiment, which reduces median mTRE from 6.5 to 2.4 mm on the femur, is demonstrated only on synthetic DRRs. Given that single-view PnP is brittle on real films, the claim that 'a second view removes depth' on actual radiographs is not established. The paper should either provide real two-view evidence or explicitly qualify the two-view result as synthetic-only in the abstract and conclusion.
minor comments (6)
  1. [§3.4, Eq. (1)] The temperature τ is a free parameter; its value and sensitivity are not discussed. Since InfoNCE performance can be sensitive to τ, please report how it was chosen.
  2. [Table 1] The 'Seen patients' rows report performance on training-set patients. This is not a generalization metric and is potentially misleading; consider removing it or labeling it explicitly as training-set fit.
  3. [§4.5] The phrase 'external test' for the real-radiograph cohort is imprecise: the cohort appears to be from the same institution. Use 'real-radiograph validation' or clarify the external nature.
  4. [Fig. 4] The horizontal axis 'Δ (deg and mm)' is ambiguous—report how rotation and translation offsets are combined or shown separately.
  5. [§3.4] The comparison '0.96 vs. 0.98 correspondence accuracy on a real femur' is confusing: the experiment is presumably on synthetic DRRs of a real femur geometry, not on a real radiograph. Clarify wording.
  6. [References] Reference formatting is inconsistent (e.g., [10] lacks page numbers, [12] lacks full venue details). Also, no ethics/IRB statement is included for the in-house CT and radiograph cohort; many journals require one.

Circularity Check

0 steps flagged · score 1.0 of 10

No circularity in the derivation chain; self-citations are not load-bearing, and the real-X-ray feasibility gap is an external-validity issue, not a definitional reduction.

full rationale

The central training objective (Eq. 1) is a multi-positive InfoNCE loss supervised by the known projection pose T* at training time only; at inference the pose is obtained from nearest-neighbour matching followed by a global PnP+RANSAC solve. No test pose is fit or back-propagated, and held-out patients are evaluated. The claimed emergent properties are probed on frozen representations: the landmark classifier is leave-one-patient-out nearest-centroid, and the bone-identity metric is measured directly from the matcher, not from a detector trained on labels. The only self-citations ([3], [4]) appear in motivation/related work and are corroborated by independent references ([26] and geometric arguments); they do not supply any premise of the derivation. Sec 4.5 honestly reports that the closed-form PnP pipeline is 'brittle on one real planar view' and instead evaluates an iterative feature-cost optimization after fine-tuning the 2D encoder; this is a validation/overclaim gap for the abstract's central claim, not circularity, because those experiments do not define the method's output on synthetic held-out patients. Overall, no equation or fitted parameter reduces to the claimed result.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The paper introduces no new physical entities. It proposes a representation (transmission-aware correspondence embedding) and a learned feature space; all inputs are standard patient CTs and radiographs. The free parameters are standard training/inference hyperparameters; the largest trusted inputs are the calibration, segmentation, and DRR physics assumptions.

free parameters (3)
  • InfoNCE temperature τ = not reported
    Appears in Eq. (1); controls feature-space sharpness and matching distribution; a hyperparameter chosen by hand, not evaluated for sensitivity.
  • RANSAC inlier threshold and iterations = not reported
    Sec. 3.5; directly determines how many candidate matches become pose inliers and thus the reported success rates.
  • Embedding dimension d and U-Net capacities = not reported
    Sec. 3.2; authors state capacity changes how often the solve succeeds (Sec. 4.3), so these choices are load-bearing for the quantitative results.
assumptions (5)
  • domain assumption Known, calibrated cone-beam projection geometry with intrinsics K for every radiograph
    Sec. 3.1; both supervision and PnP assume calibration is correct.
  • domain assumption Accurate per-bone CT segmentation from the in-house tool
    Sec. 4.1; the method registers segmented bone volumes, so segmentation errors directly corrupt training labels and test inputs.
  • domain assumption DRR rendering is a sufficient approximation of X-ray image formation for learning transferable 2D embeddings
    Sec. 4.1 and Sec. 5.2; scatter, beam hardening, and soft tissue are omitted; only domain randomization on the 2D tower addresses them.
  • domain assumption Multi-positive InfoNCE with log-sum-exp over ray points yields a correspondence embedding usable by nearest-neighbor matching
    Eq. (1), Sec. 3.4; a modeling assumption validated only empirically on synthetic DRRs.
  • standard math PnP with RANSAC can recover pose from the dense, outlier-contaminated match set
    Sec. 3.5; classical geometric result assuming a sufficient inlier ratio and calibrated intrinsics.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Learning Dense 2D-3D Correspondence for X-ray-to-CT Registration of Knee Bones." pith.science (2026). https://pith.science/paper/D566VDKN

@misc{pith2026260722803,
  author       = {Pith},
  title        = {Pith review of: Learning Dense 2D-3D Correspondence for X-ray-to-CT Registration of Knee Bones},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/D566VDKN}},
  note         = {Machine review of arXiv:2607.22803}
}
read the original abstract

Recovering the 6-DoF pose of the knee bones from a plain radiograph, given the patient's segmented pre-operative CT, turns a routine low-dose image into a quantitative measurement of joint geometry, without the added dose of a repeat CT or a fixed biplanar rig. Classic solutions align a rendered bone silhouette to image edges; recent alternatives refine pose by backpropagating an image-similarity loss through a differentiable X-ray renderer. Both operate one patient at a time and are fragile under a single view. Silhouettes are depth-ambiguous, and differentiable-rendering refinement has a narrow capture range at substantial per-iteration cost. We instead learn an amortized, subject-agnostic dense 2D-3D correspondence, supervised solely by projection geometry. One shared-weight model per bone, trained across 758 patients, registers patients unseen during training. The pose then follows in closed form from a global, initialization-free, render-free PnP+RANSAC solve. Because X-ray formation is transmissive, our correspondence target is transmission-aware rather than tied to a single surface. Though trained only to register, the representation is anatomically semantic: a simple classifier reads a landmark's anatomical region from its embedding across held-out patients, and the same features separate the knee's bones into a 2D-3D-consistent identity learned without any bone label. On a large single-institution cohort the model generalizes well to held-out patients.

Figures

Figures reproduced from arXiv: 2607.22803 by the authors.

Figure 1
Figure 1. Method overview. We learn an amortized dense 2D–3D correspondence be￾tween a patient’s CT and a single radiograph. A 3D U-Net Φθ embeds the CT and a 2D U-Net Ψϕ embeds the radiograph into a shared feature space. At training time the known pose T ⋆ projects each 3D point to its pixel, and a multi-positive InfoNCE loss (Eq. (1)) pulls a pixel’s feature toward the features of all bone points along its ray (X￾ray is tra… view at source ↗
Figure 2
Figure 2. The same shared-weight, mesh-free recipe on the patella, three held-out patients (easy/typical/hard). Left: the input DRR; middle: the 2D-encoder embedding (3D￾UMAP→RGB, masked to the patella); right: the recovered-pose silhouette (red) over ground truth (green). The embedding is smooth and consistent across patients on this near-circular bone that has no usable contour, and the recovered silhouette overlaps ground … view at source ↗
Figure 3
Figure 3. Cross-patient embedding transfer by one shared-weight, mesh-free model on three held-out patients (easy/typical/hard). Left: the input DRR; middle: the 2D￾encoder embedding (3D-UMAP→RGB, masked to the femur); right: the recovered￾pose silhouette (red) over ground truth (green). On bones never seen in training the embedding is smooth, informative and anatomically coherent. This shows that the en￾coder keys on transfe… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Capture range and cost. Success rate (left) and median mTRE (right) vs. initialization offset ∆ (deg and mm), for differentiable-rendering registration (DiffDRR, red) and our init-free solve (blue, dashed). DiffDRR is most accurate at ∆= 0 but its success collapses wit…
Figure 5
Figure 5. Figure 5: Emergent, label-free bone identity. The whole-knee model is trained with no bone-identity label, only 2D–3D dense correspondence. (a) the AP radiograph; (b) the learned 2D-encoder embedding (3D-UMAP→RGB), a continuous, label-free feature map; (c) each foreground pixel …
Figure 6
Figure 6. Figure 6: The registration-only embedding is anatomically semantic. [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 6 canonical work pages

  1. [1]

    IEEE Transactions on Biomedical Engineering43(6), 638–649 (1996).https://doi.org/10.1109/10

    Banks, S.A., Hodge, W.A.: Accurate measurement of three-dimensional knee replacement kinematics using single-plane fluoroscopy. IEEE Transactions on Biomedical Engineering43(6), 638–649 (1996).https://doi.org/10.1109/10. 495283

  2. [2]

    Springer (2018), arXiv:1803.08608

    Bier, B., Unberath, M., Zaech, J.N., Fotouhi, J., Armand, M., Osgood, G., Navab, N., Maier, A.: X-ray-transform invariant anatomical landmark detection for pelvic traumasurgery.In:MedicalImageComputingandComputerAssistedIntervention (MICCAI). Springer (2018), arXiv:1803.08608

  3. [3]

    International Journal of Sustainable Construction and Design 7(1) (2016)

    Daems, R., Victor, J., De Baets, P., Van Onsem, S., Verstraete, M.: Validation of three-dimensional total knee replacement kinematics measurement using single- plane fluoroscopy. International Journal of Sustainable Construction and Design 7(1) (2016)

  4. [4]

    Machine Vision and Applica- tions37(1), 2 (2025).https://doi.org/10.1007/s00138-025-01763-z 16 R

    De Roovere, P., Daems, R., Croenen, J., Wyffels, F.: CCPose: high-precision six- dimensional pose estimation for industrial objects. Machine Vision and Applica- tions37(1), 2 (2025).https://doi.org/10.1007/s00138-025-01763-z 16 R. Daems et al

  5. [5]

    Physics in Medicine & Biology61(8), 3009–3025 (2016).https://doi.org/10.1088/0031-9155/61/8/3009

    De Silva, T., Uneri, A., Ketcha, M.D., Reaungamornrat, S., Kleinszig, G., Vogt, S., Aygun, N., Lo, S.F., Wolinsky, J.P., Siewerdsen, J.H.: 3d–2d image registration for target localization in spine surgery: investigation of similarity metrics providing robustness to content mismatch. Physics in Medicine & Biology61(8), 3009–3025 (2016).https://doi.org/10.1...

  6. [6]

    Journal of Biomechanical Engineering127(4), 692–699 (2005).https: //doi.org/10.1115/1.1933949

    Fregly, B.J., Rahman, H.A., Banks, S.A.: Theoretical accuracy of model-based shape matching for measuring natural knee kinematics with single-plane fluo- roscopy. Journal of Biomechanical Engineering127(4), 692–699 (2005).https: //doi.org/10.1115/1.1933949

  7. [7]

    In: Medical Image Computing and Computer Assisted Intervention (MICCAI)

    Gao, C., Liu, X., Gu, W., Killeen, B., Armand, M., Taylor, R., Unberath, M.: Gen- eralizing spatial transformers to projective geometry with applications to 2d/3d registration. In: Medical Image Computing and Computer Assisted Intervention (MICCAI). pp. 329–339. Springer (2020).https://doi.org/10.1007/978-3-030- 59716-0_32, arXiv:2003.10987

  8. [8]

    arXiv preprint arXiv:2503.16309 (2025)

    Gopalakrishnan, V., Chlorogiannis, D.D., Abumoussa, A., Larson, A.M., Haou- chine, N., Orbach, D.B., Frisken, S., Dey, N., Golland, P.: Rapid patient-specific neural networks for intraoperative x-ray to volume registration. arXiv preprint arXiv:2503.16309 (2025)

Show all 33 references
  1. [9]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024), arXiv:2312.06358

    Gopalakrishnan, V., Dey, N., Golland, P.: Intraoperative 2d/3d image registration via differentiable x-ray rendering. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024), arXiv:2312.06358

  2. [10]

    In: Workshop on Clinical Image-Based Procedures (MICCAI)

    Gopalakrishnan, V., Golland, P.: Fast auto-differentiable digitally reconstructed radiographs for solving inverse problems in intraoperative imaging. In: Workshop on Clinical Image-Based Procedures (MICCAI). pp. 1–11. Springer (2022).https: //doi.org/10.1007/978-3-031-23179-7_...

  3. [11]

    International Journal of Computer Assisted Radiology and Surgery15(5), 759–769 (2020).https://doi.org/10.1007/s11548-020-02162-7

    Grupp, R.B., Unberath, M., Gao, C., Hegeman, R.A., Murphy, R.J., Alexander, C.P., Otake, Y., McArthur, B.A., Armand, M., Taylor, R.H.: Automatic anno- tation of hip anatomy in fluoroscopy for robust and efficient 2d/3d registration. International Journal of Computer Assisted R...

  4. [12]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2022)

    Haugaard, R.L., Buch, A.G.: SurfEmb: Dense and continuous correspondence distributions for object pose estimation with learnt surface embeddings. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2022)

  5. [13]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2020)

    Hodaň, T., Baráth, D., Matas, J.: EPOS: Estimating 6d pose of objects with sym- metries. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2020)

  6. [14]

    In: International Conference on Computer Vision (ICCV)

    Iwase, S., Liu, X., Khirodkar, R., Yokota, R., Kitani, K.M.: RePOSE: Fast 6d object pose refinement via deep texture rendering. In: International Conference on Computer Vision (ICCV). pp. 3283–3292 (2021)

  7. [15]

    The Journal of Arthroplasty38(10), 2068–2074 (2023).https://doi.org/10

    Jensen, A.J., Flood, P.D., Palm-Vlasak, L.S., Burton, W.S., Chevalier, A., Rul- lkoetter, P.J., Banks, S.A.: Joint track machine learning: An autonomous method of measuring total knee arthroplasty kinematics from single-plane x-ray images. The Journal of Arthroplasty38(10), 20...

  8. [16]

    Journal of Biomechanics36(6), 873–882 (2003).https://doi.org/10

    Kaptein, B.L., Valstar, E.R., Stoel, B.C., Rozing, P.M., Reiber, J.H.: A new model- based RSA method validated using CAD models and models from reversed engi- neering. Journal of Biomechanics36(6), 873–882 (2003).https://doi.org/10. 1016/S0021-9290(03)00002-2 Dense Corresponde...

  9. [17]

    IEEE Transactions on Medical Imaging24(9), 1177–1189 (2005).https://doi.org/10.1109/TMI

    van de Kraats, E.B., Penney, G.P., Tomaževič, D., van Walsum, T., Niessen, W.J.: Standardized evaluation methodology for 2-d–3-d registration. IEEE Transactions on Medical Imaging24(9), 1177–1189 (2005).https://doi.org/10.1109/TMI. 2005.853240

  10. [18]

    In: European Conference on Computer Vision (ECCV)

    Labbé, Y., Carpentier, J., Aubry, M., Sivic, J.: CosyPose: Consistent multi-view multi-object 6d pose estimation. In: European Conference on Computer Vision (ECCV). pp. 574–591. Springer (2020)

  11. [19]

    In: Conference on Robot Learning (CoRL) (2022)

    Labbé, Y., Manuelli, L., Mousavian, A., Tyree, S., Birchfield, S., Tremblay, J., Carpentier, J., Aubry, M., Fox, D., Sivic, J.: MegaPose: 6d pose estimation of novel objects via render & compare. In: Conference on Robot Learning (CoRL) (2022)

  12. [20]

    In: European Conference on Computer Vision (ECCV)

    Li, Y., Wang, G., Ji, X., Xiang, Y., Fox, D.: DeepIM: Deep iterative matching for 6d pose estimation. In: European Conference on Computer Vision (ECCV). pp. 683–698 (2018)

  13. [21]

    IEEE Transactions on Medical Imaging22(12), 1561–1574 (2003)

    Mahfouz, M.R., Hoff, W.A., Komistek, R.D., Dennis, D.A.: A robust method for registration of three-dimensional knee implant models to two-dimensional fluo- roscopy images. IEEE Transactions on Medical Imaging22(12), 1561–1574 (2003). https://doi.org/10.1109/TMI.2003.820027

  14. [22]

    Medical Image Analysis16(3), 642–661 (2012).https://doi.org/10.1016/j.media.2010.03.005

    Markelj, P., Tomaževič, D., Likar, B., Pernuš, F.: A review of 3d/2d registration methods for image-guided interventions. Medical Image Analysis16(3), 642–661 (2012).https://doi.org/10.1016/j.media.2010.03.005

  15. [23]

    IEEE Transactions on Medical Imaging31(4), 948–962 (2012)

    Otake, Y., Armand, M., Armiger, R.S., Kutzer, M.D., Basafa, E., Kazanzides, P., Taylor, R.H.: Intraoperative image-based multiview 2d/3d registration for image- guided orthopaedic surgery: incorporation of fiducial-based C-arm tracking and GPU-acceleration. IEEE Transactions o...

  16. [24]

    IEEE Transactions on Medical Imaging17(4), 586–595 (1998).https://doi.org/ 10.1109/42.730403

    Penney, G.P., Weese, J., Little, J.A., Desmedt, P., Hill, D.L.G., Hawkes, D.J.: A comparison of similarity measures for use in 2-d–3-d medical image registration. IEEE Transactions on Medical Imaging17(4), 586–595 (1998).https://doi.org/ 10.1109/42.730403

  17. [25]

    Os- teoarthritis and Cartilage16(12), 1433–1441 (2008).https://doi.org/10.1016/ j.joca.2008.06.016

    Peterfy, C.G., Schneider, E., Nevitt, M.: The osteoarthritis initiative: report on the design rationale for the magnetic resonance imaging protocol for the knee. Os- teoarthritis and Cartilage16(12), 1433–1441 (2008).https://doi.org/10.1016/ j.joca.2008.06.016

  18. [26]

    Medical Engineering & Physics77, 107–113 (2020).https: //doi.org/10.1016/j.medengphy.2020.01.002

    Postolka, B., List, R., Thelen, B., Schütz, P., Taylor, W.R., Zheng, G.: Evalua- tion of an intensity-based algorithm for 2d/3d registration of natural knee vide- ofluoroscopy data. Medical Engineering & Physics77, 107–113 (2020).https: //doi.org/10.1016/j.medengphy.2020.01.002

  19. [27]

    Proceedings of the Institution of Mechanical Engineers, Part H225(8), 753–761 (2011).https://doi.org/10.1177/0954411911407669

    Prins, A., Kaptein, B., Stoel, B., Nelissen, R., Reiber, J., Valstar, E.: Integrated contour detection and pose estimation for fluoroscopic analysis of knee implants. Proceedings of the Institution of Mechanical Engineers, Part H225(8), 753–761 (2011).https://doi.org/10.1177/0...

  20. [28]

    In: Medical Image Computing and Computer Assisted Intervention (MICCAI) (2023), arXiv:2311.15087

    Shrestha, P., Xie, C., Shishido, H., Yoshii, Y., Kitahara, I.: X-ray to CT rigid registration using scene coordinate regression. In: Medical Image Computing and Computer Assisted Intervention (MICCAI) (2023), arXiv:2311.15087

  21. [29]

    In: Asian Conference on Computer Vision (ACCV) (2024), arXiv:2410.08152

    Shrestha, P., Xie, C., Yoshii, Y., Kitahara, I.: RayEmb: Arbitrary landmark de- tection in x-ray images using ray embedding subspace. In: Asian Conference on Computer Vision (ACCV) (2024), arXiv:2410.08152

  22. [30]

    Daems et al

    Unberath, M., Gao, C., Hu, Y., Judish, M., Taylor, R.H., Armand, M., Grupp, R.: The impact of machine learning on 2d/3d registration for image-guided interven- 18 R. Daems et al. tions: A systematic review and perspective. Frontiers in Robotics and AI8, 716007 (2021).https://d...

  23. [31]

    PLOS ONE17(6), e0270596 (2022).https://doi.org/10.1371/ journal.pone.0270596

    Vogl, F., Schütz, P., Postolka, B., List, R., Taylor, W.R.: Personalised pose es- timation from single-plane moving fluoroscope images using deep convolutional neural networks. PLOS ONE17(6), e0270596 (2022).https://doi.org/10.1371/ journal.pone.0270596

  24. [32]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2019)

    Wang, H., Sridhar, S., Huang, J., Valentin, J., Song, S., Guibas, L.J.: Normalized object coordinate space for category-level 6d object pose and size estimation. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2019)

  25. [33]

    Radiology: Ar- tificial Intelligence5(5), e230024 (2023).https://doi.org/10.1148/ryai.230024

    Wasserthal, J., Breit, H.C., Meyer, M.T., Pradella, M., Hinck, D., Sauter, A.W., Heye, T., Boll, D.T., Cyriac, J., Yang, S., Bach, M., Segeroth, M.: TotalSegmenta- tor: Robust segmentation of 104 anatomic structures in CT images. Radiology: Ar- tificial Intelligence5(5), e2300...

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.