CoChat
Sign Up
Cited by 8Medical Image Segmentation TechniquesAnatomy and Medical TechnologyMedical Imaging and Analysis

Towards 2D/3D Registration of the Preoperative MRI to Intraoperative Fluoroscopic Images for Visualisation of Bone Defects

Open access
Ku, Ping-Cheng, Martin-Gomez, Alejandro, Gao, Cong, Grupp, Robert, Mears, Simon C., Armand, Mehran·2023·Computer Methods in Biomechanics and Biomedical Engineering: Imaging & Visualization
DOI 10.1080/21681163.2022.2152375PMC PMC10406464
Open this paper in CoChat — free accountOpen PDF

Abstract

Magnetic Resonance Imaging (MRI) is a medical imaging modality that allows for the evaluation of soft-tissue diseases and the assessment of bone quality. Preoperative MRI volumes are used by surgeons to identify defected bones, perform the segmentation of lesions, and generate surgical plans before the surgery. Nevertheless, conventional intraoperative imaging modalities such as fluoroscopy are less sensitive in detecting potential lesions. In this work, we propose a 2D/3D registration pipeline that aims to register preoperative MRI with intraoperative 2D fluoroscopic images. To showcase the feasibility of our approach, we use the core decompression procedure as a surgical example to perform 2D/3D femur registration. The proposed registration pipeline is evaluated using digitally reconstructed radiographs (DRRs) to simulate the intraoperative fluoroscopic images. The resulting transformation from the registration is later used to create overlays of preoperative MRI annotations and planning data to provide intraoperative visual guidance to surgeons. Our results suggest that the proposed registration pipeline is capable of achieving reasonable transformation between MRI and digitally reconstructed fluoroscopic images for intraoperative visualization applications.

AI snapshot

Written from the paper’s full text.

TL;DR

The paper proposes a 2D/3D registration pipeline that synthesizes CT from preoperative MRI with a gradient-consistency CycleGAN and registers it to simulated intraoperative fluoroscopic images for overlaying bone-defect annotations. In a core-decompression feasibility study using digitally reconstructed radiographs, the pipeline achieved mean femur registration error of 12.95 ± 4.90 mm and rotation error of 4.51 ± 2.67 degrees.

Key findings

  • Across 60 simulated DRR projections, the mean femur registration error was 12.95 ± 4.90 mm, with a mean rotation error of 4.51 ± 2.67 degrees.
  • The in-plane translation error, which the authors identify as critical for overlay accuracy, was 4.55 ± 1.92 mm; the total translation error was dominated by out-of-plane error.
  • The CycleGAN-based MR-to-CT synthesis produced synthetic CT with SSIM 0.3556 ± 0.1238 and MSE 0.0981 ± 0.0216 compared with ground-truth CT.
  • The resulting transformation allowed preoperative MRI annotations, including AVN lesion segmentation and planned drilling paths, to be overlaid on DRRs; the authors state this in-plane accuracy may provide useful visual guidance for a 6–10 mm drill, though it does not meet the 2-mm clinical requirement cited for core decompression.

Methods

  • The pipeline includes CycleGAN-based MR-to-synthetic-CT generation with gradient consistency loss, automatic bone segmentation with occasional manual touch-ups, ICP initialization using 200 points around the femoral head, and intensity-based 2D/3D registration using Grad-NCC similarity and CMA-ES optimization.
  • Data included three preoperative MRI volumes and three paired CT volumes from patients, plus 12 coronal torso CT volumes from the New Mexico Decedent Image Database for CycleGAN training; 60 DRRs were simulated using C-arm geometry approximating a Siemens CIOS Fusion C-Arm.
  • Registration accuracy was evaluated against known DRR camera poses, with translation error separated into in-plane and out-of-plane components; synthetic CT quality was assessed with SSIM and MSE.

Limitations

  • The paper states that translation error, especially out-of-plane error, is significant for robot-guided surgery and exceeds the 2-mm clinical requirement for core decompression; it also notes that synthetic CT segmentation can be inaccurate because of limited CycleGAN training data.
  • The pipeline is semi-automatic, requiring manual clinician annotation of landmarks and occasional manual segmentation touch-ups, and the evaluation uses DRRs rather than real intraoperative fluoroscopy.
  • Evident from the text: the clinical MRI sample is very small (three patients), no real intraoperative C-arm data are tested, and the reported SSIM of 0.3556 indicates limited synthetic CT image similarity.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1080/21681163.2022.2152375

Cites

40 of 40 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

11 of 29 cited sources checked · 4 supported · 0 flagged · accuracy not yet measured

  • Findingown data

    The geometry of the synthetic CT femur projections mostly aligns with the DRRs except for the greater trochanter of the femur.

    “An example is shown in Figure 8, where the geometry of the sCT femur projections mostly aligns with the DRRs except for the greater trochanter of the femur.”
    Quote in the paperSubstantial claim
  • Findingown data

    The 2D annotations of the preoperatively planned drilling trajectory and AVN lesion annotations are not visually affected by the out-of-plane translation errors.

    “An example of the overlay is shown in Figure 9, where the 2D annotations of the preoperatively planned drilling trajectory and AVN lesions annotations are not visually affected by the out-of-plane translation errors.”
    Quote in the paperSubstantial claim
  • Findingown data

    The in-plane translation error has a mean of 4.55 ± 1.92 mm.

    “The accuracy of the preoperative annotations on the intraoperative 2D radiographs is the same as the in-plane translation error acquired in Table 1, which has a mean of 4.55 ± 1.92 mm translation error and the same rotation registration error.”
    Quote in the paperNot a substantial claim
  • Interpretationown data

    The 4.55 ± 1.92 mm in-plane translation error and the 4.51 ± 2.67 degree rotation error acquired in this study ensure that the projection of the synthetic CT is positionally close to the target on the 2D imaging plane.

    “Although the translation registration error in out-of-plane direction needs to be improved for clinical applications, the 4.55 ± 1.92 mm in-plane translation error and the 4.51 ± 2.67 degree rotation error acquired in this study ensure that the projection of the sCT is positionally close to the target on the 2D imaging plane.”
    Quote in the paperSubstantial claim
  • Interpretationreasoning

    Because the synthetic CT is generated from the CycleGAN-based network with limited training size, the border of the proximal femur in the synthetic CT may be affected by soft tissues in the MRI and become less accurate.

    “Since the sCT is generated from the CycleGAN-based network with limited training size, the border of the proximal femur in the sCT may be affected by the soft tissues in the MRI and become less accurate.”
    Quote in the paperSubstantial claim
  • May combine two claims. 17
  • Classified as probably not a substantial claim. 13

  • May combine two claims.

    Premisecitation

    Intraoperative imaging techniques such as C-arm fluoroscopy or X-ray are important for localizing the greater trochanter, tracking and navigating the drilling and injection device to the femoral head, and positioning the trephine and scraping spoon while removing dead bone.

    “Intraoperative imaging techniques such as C-arm fluoroscopy or X-ray have become important for the localization of the greater trochanter, tracking and navigation of the drilling and injection device to the femoral head, and positioning of the trephine and scraping spoon while removing the dead bones (Nolte et al. 2000; Geisert et al. 2017; Tian et al. 2021).”
    Quote in the paperSubstantial claim3 of 3 sources checked
    • cites Nolte et al. 2000 source

      The Introduction and Surgical procedure sections describe C-arm fluoroscopy for navigating instruments during spinal pedicle-screw placement. They do not address the greater trochanter, femoral head, drilling or injection devices, or instruments for removing dead bone. Page numbers were not provided for these passages. (LLM, conf=0.88)

      Experimental — accuracy not yet measured
    • cites Geisert et al. 2017 source

      The Core Decompression section (page not provided) describes C-arm fluoroscopy guiding a pin toward the femoral-head lesion and live fluoroscopy guiding reaming. The Approach to Hip section describes an incision centered over the greater trochanter, but not imaging to localize it. The Osseoscopy section describes camera-guided curette debridement, not fluoroscopic positioning of a trephine or scraping spoon; the paper does not describe an injection device. (LLM, conf=0.88)

      Experimental — accuracy not yet measured
    • cites Tian et al. 2021 source

      The Surgical Procedure section says C-arm fluoroscopy located the greater trochanter and confirmed the drilling direction, while intraoperative X-rays helped plan the robot-assisted path. It also reports fluoroscopic checks during bone grafting, but does not say imaging positioned the trephine or scraping spoon while dead bone was removed. (LLM, conf=0.86)

      Experimental — accuracy not yet measured

  • May combine two claims.

    Premisecitation

    The translation component of registration error is significant for purposes such as robot-guided surgery, while a 2-mm error is considered to match the clinical requirement of core decompression.

    “In this experiment, the translation component of the registration error is a significant error for purposes such as robot-guided surgery, while a 2-mm error is considered to match the clinical requirement of core decompression (Wang et al. 2017).”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites Wang et al. 2017 source

      The Discussion (page not provided) says a 1.92-mm error between the needle tip and target was considered to meet the clinical requirement for core decompression, supporting the approximate 2-mm claim. It does not assess the translation component of registration error or its significance in robot-guided surgery; the measured error was drilling accuracy using a navigation template. (LLM, conf=0.88)

      Experimental — accuracy not yet measured

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Registration between breast MRI and X-ray mammography is feasible through a volume-preserving affine transformation model or gradient-based registration methods, reaching accuracy on the order of 10 mm.

    “The registration between breast MRI and X-ray mammography has shown to be feasible through a properly defined volume-preserving affine transformation model (Mertzanidou et al. 2012) or through gradient-based registration methods (García et al. 2019). Both algorithms are able to reach registration accuracy in the order of 10mm.”
    Quote in the paperNot a substantial claim2 of 2 sources checked
    • cites Mertzanidou et al. 2012 source
    • cites García et al. 2019 source

  • May combine two claims.

    Premisecitation

    Core decompression requires acquisition of multiple intraoperative radiographs, which increases the risk of surgical complications and increases radiation exposure for both the patient and the surgical team.

    “Core decompression procedure requires the acquisition of multiple intraoperative radiographs, which increases the risk of surgical complications and increases the radiation exposure for both the patient and the surgical team (Beckmann et al. 2006; Wang et al. 2017).”
    Quote in the paperSubstantial claim1 of 2 sources checked
    • cites Beckmann et al. 2006 source
    • cites Wang et al. 2017 source

      The Introduction says conventional core decompression is guided by C-arm X-ray, and the Discussion links multiple drillings to greater radiation exposure for patients and the surgical team and to surgical complications. But the paper does not say every core decompression procedure requires multiple intraoperative radiographs; it describes a template-guided procedure performed without fluoroscopy or C-arm X-ray. (LLM, conf=0.84)

      Experimental — accuracy not yet measured

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Due to the lack of distinct features of the proximal femur in DRRs, Gao et al. proposed a two-step registration where the pelvis is first registered for better femoral pose initialization because the pelvis has a rich collection of distinct features leading to accurate registration.

    “Due to the lack of distinct features of the proximal femur in the DRRs, Gao et al. have proposed a two-step registration where the pelvis is first registered for a better femoral pose initialization since the pelvis has a rich collection of distinct features that leads to accurate registration accuracy (Gao et al. 2020).”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites Gao et al. 2020 source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Core decompression has become a preferred procedure to treat early-stage AVN due to its simplicity, low complication rate, and promising efficacy.

    “Core decompression has become a preferred procedure to treat early-stage AVN due to its simplicity, low complication rate, and promising efficacy (Lieberman 2004; Veillette et al. 2006; Talmaç 2018; Nori 2015).”
    Quote in the paperNot a substantial claim1 of 3 sources checked
    • cites Lieberman 2004 source
    • cites Veillette et al. 2006 source
    • cites Talmaç 2018 source

  • Classified as probably not a substantial claim. May combine two claims.

    Premisecitation

    MRI is a medical imaging modality that allows for the evaluation of soft-tissue diseases and the assessment of bone quality.

    “Magnetic Resonance Imaging (MRI) is a medical imaging modality that allows for the evaluation of soft-tissue diseases and the assessment of bone quality.”
    • Also stated in: § Introduction · plain
      “This property has made MRI a good candidate for the evaluation of soft-tissue diseases, the assessment of bone quality, and the detection of bone tumors and tumor-like lesions (Chung et al. 2012; Chang et al. 2017; Nascimento et al. 2014).”
    Quote in the paperSubstantial claim2 of 3 sources checked
    • cites Chung et al. 2012 source
    • cites Chang et al. 2017 source

      The paper’s Abstract (page not provided) describes MRI as a proposed method for assessing bone quality in vivo. The passages address bone structure and physiology, but do not substantiate the separate claim that MRI evaluates soft-tissue diseases (searched: soft-tissue disease and musculoskeletal disease imaging). (LLM, conf=0.85)

      Experimental — accuracy not yet measured
    • cites Nascimento et al. 2014 source

      In “Applying MRI in tumour detection and diagnosis” (page unavailable), the paper says MRI can demonstrate soft-tissue pathology associated with osteochondromas. The Introduction and tumour-detection section also describe MRI’s ability to characterise bone and bone-marrow lesions, but they do not establish that it assesses bone quality. (LLM, conf=0.85)

      Experimental — accuracy not yet measured

  • May combine two claims.

    Findingown data

    The final SSIM result has a mean of 0.3556 and a standard deviation of 0.1238, while the MSE result has a mean of 0.0981 and a standard deviation of 0.0216.

    “The final SSIM result has a mean of 0.3556 and a standard deviation of 0.1238; while the MSE result has a mean of 0.0981 and a standard deviation of 0.0216.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The generated synthetic CT volumes preserve compact bone structures from the input MRI volumes and map them to high CT intensity, while visible structures in MRI such as muscles, urinary bladder, and sciatic nerves are mapped to low intensities in the synthetic CT.

    “The generated sCT volumes preserve compact bone structures from the input MRI volumes and map them to a high CT intensity; while visible structures in MRI such as muscles, urinary bladder, and sciatic nerves are mapped to low intensities in the sCT.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    The left or right femur does not demonstrate a significant impact on the registration result.

    “The left or right femur does not demonstrate a significant impact on the registration result.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingown data

    The in-plane translation error in the first subject is slightly smaller than in the other two subjects, while the rotation error remains similar among all three subjects.

    “As shown in Figure 10, the in-plane translation error in the first subject is slightly smaller than the other two subjects, while the rotation error remains similar among all three subjects.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    The mean femur registration error is 12.95 ± 4.90 mm, and the rotation error is 4.51 ± 2.67 degrees.

    “The mean femur registration error is 12.95 ± 4.90 mm, and the rotation error is 4.51 ± 2.67 degrees.”
    • Also stated in: § Discussion · plain
      “The registration result of the femur bone between the sCT and the DRR has a mean translation error of 12.95 ± 4.90 mm with a mean rotation error of 4.51 ± 2.67 degrees.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Interpretationnone given

    The proposed 2D/3D registration pipeline registers preoperative MRI and intraoperative fluoroscopic images without requiring preoperative CT scans from the patient.

    “In this paper, we propose a novel 2D/3D registration pipeline that aims to register preoperative MRI and intraoperative fluoroscopic images without the requirement of preoperative CT scans from the patient.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationreasoning

    Because slight movement along the z-axis maintains the projected shape of the femur in synthetic CT, it has a smaller impact on similarity scores during registration, leading to low rotation registration errors but less ideal translation registration results.

    “Since slight movement along this axis maintains the projected shape of the femur in sCT, it has a smaller impact on the calculation of similarity scores during registration, which leads to low rotation registration errors but less ideal translation registration results.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

    Interpretationreasoning

    With improved intraoperative visualization of bone defects, the proposed registration pipeline could improve the performance of surgical procedures and potentially reduce the number of fluoroscopic images required during surgery.

    “With the improved intraoperative visualization of bone defects, the proposed registration pipeline could improve the performance of surgical procedures and potentially reduce the number of fluoroscopic images required during surgery.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: Generation of DRRs creates a realistic simulation of intraoperative fluoroscopy.
    • Rests on an assumption: The performance of the pipeline on DRRs will translate to real intraoperative fluoroscopic images.

  • May combine two claims.

    Interpretationreasoning

    In this study, the translation error mostly originates from the out-of-plane translation error and the less accurate synthetic CT model used for registration.

    “In this study, the translation error mostly originates from the out-of-plane translation error and the less accurate sCT model used for registration.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    During core decompression surgery, the precision of drilling is essential to accurately decompress targeted lesions while avoiding femoral cartilage damage.

    “During the surgery, the precision of the drilling is essential to accurately decompress the targeted lesions while avoiding any femoral cartilage damage (Beckmann et al. 2006).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Beckmann et al. 2006 source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Additional loss functions such as gradient consistency loss and structure-consistency loss have shown promising results for avoiding geometric distortions in MR-to-CT synthesis.

    “To avoid geometric distortions, additional loss functions such as gradient consistency loss (Hiasa et al. 2018) and structure-consistency loss (Yang et al. 2020) have shown promising results.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites Hiasa et al. 2018 source
    • cites Yang et al. 2020 source

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Core decompression creates a drilling hole from the distal end of the trochanter to stimulate blood supply to the necrotic area, diminish femoral head pressure, and delay progression of osteonecrosis.

    “The core decompression method is a surgical procedure that creates a drilling hole made from the distal end of the trochanter to stimulate the blood supply to the necrotic area, diminish the femoral head pressure, and delay the progression of osteonecrosis (Lieberman 2004; Yu et al. 2018).”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites Lieberman 2004 source
    • cites Yu et al. 2018

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Interpretationnone given

    With a trained network model, the generated synthetic CT will detect and preserve bone structures with high image intensities and reduce the intensity impact of soft tissues from the input MRI.

    “With a trained network model, the generated sCT will detect and preserve the bone structures with high image intensities and reduce the intensity impact of the soft tissues from the input MRI.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

    Interpretationnone given

    Although the experiment is restricted to core decompression procedures, no assumptions were made that would preclude application of the proposed registration pipeline to other surgical procedures.

    “Although the experiment carried out in this paper is restricted to core decompression procedures, no assumptions were made that would preclude the application of the proposed registration pipeline to other surgical procedures.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: The core decompression procedure is a suitable surgical example for evaluating the proposed pipeline.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Performing MR-to-CT synthesis through the naive CycleGAN often results in inconsistent structures between input and synthetic images.

    “However, performing MR-to-CT synthesis through the naive CycleGAN often results in inconsistent structures between input and synthetic images.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites Hiasa et al. 2018 source
    • cites Yang et al. 2020 source

  • May combine two claims.

    Premisecitation

    Avascular necrosis of the femoral head is visible in preoperative MRI but hardly detectable intraoperatively; radiographs have low sensitivity for early-stage AVN, while MRI is sensitive to necrotic tissues but hard to acquire intraoperatively.

    “A clear example of a lesion that is visible in preoperative MRI but hardly detectable intraoperatively is avascular necrosis (AVN).”
    • Also stated in: § Introduction · boosted
      “However, studies have shown that radiographs exhibit low sensitivity for the early stages of AVN; while MRI scans are particularly sensitive to the detection of necrotic tissues, MRI scans are hard to acquire intraoperatively (Malizos et al. 2007).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Malizos et al. 2007 source
23 of 23

  • The paper relies on this without support.

    No research exists on MRI to 2D radiography registration as a tool for intraoperative visualization of bone defects.

    “To our knowledge, there is no research on MRI to 2D radiography registration as a tool for intraoperative visualization of bone defects.”

  • The paper relies on this without support.

    Generation of DRRs creates a realistic simulation of intraoperative fluoroscopy.

    “The generation of DRR creates a realistic simulation of the intraoperative fluoroscopy, and it also enables us to perform a more comprehensive 2D/3D registration analysis with the increased number of projections available in various projecting directions.”

  • The paper relies on this without support.

    The core decompression procedure is a suitable surgical example for evaluating the proposed pipeline.

    “In this paper, we use the core decompression procedure as a surgical example for the proposed pipeline.”

  • The paper relies on this without support.

    The performance of the pipeline on DRRs will translate to real intraoperative fluoroscopic images.

    “The generated DRRs (Im) are used to simulate intraoperative radiographs and are used in the 2D/3D registration step.”

  • The paper relies on this without support.

    Manual annotation of three anatomical landmarks reliably defines the femoral anatomical coordinate system.

    “To construct the femoral ACS, three anatomical landmarks are manually annotated in the CT volumes prior to the experiments: the center of the femoral head (FH), the tip of the supero-posterior facet of the greater trochanter (GTR), and the tip of the lesser trochanter (LTR) (Domb and Carreira 2013).”

  • The paper relies on this without support.

    The 12 coronal torso CT volumes from the New Mexico Decedent Image Database are representative for training the MR-to-CT synthesis network.

    “a total of 12 coronal torso CT volumes from the New Mexico Decedent Image Database (NMDID) are used as our CT dataset.”
6 of 6

Cites

Flagged first, then by citations

3 of the 43 references couldn’t be matched to a known paper.

Research works printed but not linked: 1

  1. Goodfellow I, 2014, Advances in Neural Information Processing Systems, 27, 2672

Cited by

Sorted by citations

Related

Authors
Ku, P.-C., Martin-Gomez, A., Gao, C., Grupp, R., Mears, S. C., & Armand, M.
Affiliations
  • Johns Hopkins University (US)
  • Biomechanical- and Image-Guided Surgical Systems (BIGSS), Laboratory for Computational Sensing and Robotics, Johns Hopkins University, Baltimore, MD, USA
  • University of Arkansas for Medical Sciences (US)
Corresponding author
Ku, Ping-Cheng
ORCID
Venue
Computer Methods in Biomechanics and Biomedical Engineering: Imaging & Visualization
Publisher
Informa UK Limited
Published
2022-12-15
Volume
11
Issue
4
Pages
1096-1105
ISSN
  • 2168-1163 (ISSN-L)
  • 2168-1171
DOI
10.1080/21681163.2022.2152375
PMC
PMC10406464
OpenAlex
W4311755298
Semantic Scholar
bb129dc415fbdde9f0413e393182d97deb026ac5
Cited by
8
References
43
Citations by year
2026: 1 · 2025: 1 · 2024: 5 · 2023: 1
FWCI
0.57
Type
journal article
Language
en
Open access
Locations
  • Computer Methods in Biomechanics and Biomedical Engineering Imaging & Visualization (publishedVersion)
  • PubMed (publishedVersion)
  • PubMed Central (submittedVersion)
Funders
National Institutes of Health
Topics
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Medical Image Segmentation Techniques
  • Physical Sciences › Engineering › Biomedical Engineering › Anatomy and Medical Technology
  • Physical Sciences › Engineering › Biomedical Engineering › Medical Imaging and Analysis
Keywords
2D-3D registration, preoperative MRI, intraoperative fluoroscopy, bone defects, core decompression, femur, digitally reconstructed radiographs, intraoperative visualization, surgical planning, bone quality assessment

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-14.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-14.

Cited sources are checked by gpt-6-sol.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review