Pith. sign in

REVIEW 4 major objections 6 minor 34 references

MitraClip Device Automated Localization in 3D Transesophageal Echocardiography via Deep Learning

T0 review · 4 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read An automated pipeline detects, localizes, and characterizes the MitraClip device in 3D transesophageal echocardiography in about 2.6 seconds per volume.

desk verdict First automated MitraClip detection in 3D TEE, with a sensible pipeline and honest limitations, but the headline accuracy numbers are weakened because the ground truth and the final refinement step share the same CAD template library. read the letter →

arxiv 2412.15013 v2 pith:YO4EJOSN submitted 2024-12-19 q-bio.QM

Riccardo Munafò , Simone Saitta , Luca Vicentini , Davide Tondi , Veronica Ruozzi , Francesco Sturla , Giacomo Ingallina , Andrea Guidotti
show 2 more authors
Eustachio Agricola Emiliano Votta
This is my paper · ORCID
classification q-bio.QM
keywords MitraCliptransesophagealechocardiographydeeplearningsegmentationAttentionUNetconfigurationclassificationCADtemplateregistrationtranscatheteredge-to-edgerepair
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

The paper claims that a three-stage automated pipeline can detect, localize, and characterize the MitraClip device in 3D transesophageal echocardiography (TEE), a task currently done by manual slice selection and expert interpretation. The pipeline segments the clip with an Attention UNet, classifies its opening configuration with a DenseNet into ten states from fully closed to fully open, and then registers a CAD-derived template of the clip to the segmentation to refine the geometry. On an in-vitro test set of 40 images, the authors report an average surface distance of 0.76 mm and a 95% Hausdorff distance of 2.44 mm before refinement, improving to 0.75 mm and 2.05 mm afterward, with a weighted F1-score near 0.74 for configuration classification. If these results hold in patients, the method could standardize intraoperative visualization and give quantitative, operator-independent feedback during mitral-valve repair.

What carries the argument

The load-bearing mechanism is a three-stage cascade. First, a 3D Attention UNet—a U-Net with attention gates that emphasize salient regions—segments the clip and its delivery catheter as a single label, and the largest connected component of the prediction is cropped. Second, a DenseNet classifier reads the cropped volume and assigns one of ten configurations, corresponding to opening angles from fully closed to fully open in 20-degree steps. Third, a library of CAD template models, built from a 3D scan of a real XTW clip and differing only in arm angle, is brought in: the template matching the predicted configuration is rigidly aligned to the segmentation surface by the iterative closest point (ICP) algorithm, refining the geometry and restoring details such as clip arms that the raw segmentation may miss. The same template library is used to build ground truth from manual masks by choosing the best-overlapping template via Dice score, so the template shapes carry both the annotation method and the refinement method.

What would settle it

Take 3D TEE volumes of a clip with independently measured geometry and opening angle (for example, from micro-CT or electromagnetic tracking in a phantom), run the pipeline, and compare its segmentations against that independent reference instead of against the template-refined masks; if the average surface distance and 95% Hausdorff distance rise well above the reported 0.75 mm and 2.05 mm, the claimed accuracy is largely template-to-template agreement rather than image-derived localization.

Watch

Extended reading notes

Core claim

The paper's central claim is that the MitraClip—a steerable device with two moving arms—can be automatically segmented and pose-classified from volumetric TEE images using a 3D Attention UNet for segmentation, a DenseNet for configuration classification into ten 20-degree bins, and rigid ICP registration of a matching CAD template to refine the segmentation. The authors report that the full pipeline processes a volume in about 2.55 seconds, with the Attention UNet reaching an average surface distance of 0.76 mm and a 95% Hausdorff distance of 2.44 mm, and with template refinement reducing the 95% Hausdorff distance to 2.05 mm. Configuration classification reaches a weighted F1-score of about 0.74 when the classifier receives segmentation-cropped input, and most misclassifications are one 20-degree step away from the true configuration. The paper argues this is the first automated localization of a device with this morphological complexity in intraoperative 3D TEE, and that the extracted position and pose information is exactly what procedural guidance would need.

Load-bearing premise

The accuracy figures depend on the CAD template library faithfully representing the real clip in every configuration, because the same templates are used both to build the ground-truth segmentations and to refine the predicted segmentations.

Editorial extensions

If this is right

  • If the reported accuracy holds, the pipeline can automatically extract 2D views centered on the clip, removing operator-dependent manual slice selection during TEER.
  • Combined with automated mitral-valve analysis, the clip's position and opening angle could be quantified relative to the valve, giving operators standardized intraprocedural feedback.
  • The near-real-time runtime of about 2.55 seconds per volume makes intraprocedural use plausible, although the ICP refinement consumes most of that time.
  • Because most classification errors are one 20-degree step off, the wrong template is still close in geometry, so the final surface-distance error remains small.
  • The same pipeline architecture could be extended to other edge-to-edge repair devices, such as the Pascal system or Triclip, with new template libraries.

Reading between the lines

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

  • The reported millimeter accuracy is measured against ground truth built from the same CAD template library used in the final refinement step, so an independent dataset with non-template annotations would be needed to know how much of the accuracy reflects true image information rather than template-to-template agreement.
  • Because configuration is predicted as one of ten discrete states, the finest angle resolution the pipeline can report is 20 degrees; regressing a continuous opening angle or interpolating between templates would give finer quantitative feedback.
  • Replacing the ICP refinement with an end-to-end pose regression network could cut the 2.55-second runtime to well under a second, since ICP alone accounts for about 2.44 seconds of the total.
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 / 6 minor

Summary. This paper proposes a fully automated three-stage pipeline for MitraClip localization and configuration assessment in 3D transesophageal echocardiography (TEE): a 3D Attention U-Net segments the clip, a DenseNet classifies its opening angle into ten states from 0° to 180° in 20° increments, and a CAD template selected by the classifier is rigidly registered via ICP to refine the segmentation. The pipeline is trained and evaluated on 196 in-vitro 3D TEE images acquired on a heart simulator, with a separate 40-volume test set. The authors report a post-refinement average surface distance of 0.75 mm, a 95% Hausdorff distance of 2.05 mm, and a weighted F1-score of 0.74 for configuration classification, together with a total inference time of about 2.55 s. Four segmentation architectures and two classifiers are compared.

Significance. If independently validated, the work would fill a clinical gap by automating a difficult step in TEER guidance and providing quantitative device characterization from 3D TEE. The simulator-based data collection with controlled poses is a strength, as is the systematic architectural comparison and the explicit reporting of runtime. The main limitation is that the evaluation is partly circular—ground-truth and final refinement are built from the same CAD templates—and the test set is too small and imbalanced to support the classification claims. A decisive validation using the simulator-known clip pose is feasible and should be required.

major comments (4)
  1. [2.2.3, 2.3.2, Tables 2 and 4] Section 2.2.3 creates the ground-truth segmentation by fitting each CAD template to the manual mask and retaining the template with maximal Dice score (steps 5–7). Section 2.3.2 then refines the predicted mask by registering a template from the same CAD library to it via ICP. As a result, the post-refinement ASD and 95% HD in Table 4 primarily measure the agreement between two template fits (one in the GT, one in the prediction) rather than the accuracy of image-derived localization. The simulator in Section 2.1 provides known clip poses and opening angles, which could serve as an independent reference; the authors should evaluate against these parameters to discount the circularity.
  2. [2.2.3, Table 3] The classification labels are defined in Section 2.2.3, step 7 as the template with maximal Dice against the manual segmentation, not from an independent measurement of the clip opening angle. The F1-scores in Table 3 thus measure agreement with the template-fitting labels, which can be inflated if the template library does not faithfully represent real clip morphology. Since the acquisition protocol controlled the opening angle, the authors should report classification accuracy with respect to the simulator-known angles.
  3. [Table 1] Table 1 shows a highly imbalanced test set: 19 of 40 volumes are at 0°, the 20° class is absent, and classes 40°, 60°, 140°, and 160° contain at most two instances. The weighted average F1 of 0.74 is therefore dominated by the 0° and 180° classes, and per-class F1 values for several intermediate angles are zero (e.g., 60° and 160° in Table 3). This does not support the claim of reliable classification across all ten configurations; the absence of 20° means that the model's behavior on that configuration is completely untested.
  4. [2.2.2, 2.3.2] Section 2.2.2 describes nine CAD templates with opening angles from 10° to 90° (plus a 120° template used for shape adjustment), but the classification task in Table 1 includes 0°, 100°, 140°, 160°, and 180°. The paper does not specify how these states map to the template library in the refinement step (Section 2.3.2), nor whether dedicated closed and fully-open templates exist. This is a load-bearing detail for the template-matching stage and must be clarified.
minor comments (6)
  1. [Abstract] Abstract: 'transesophagel' should be 'transesophageal'.
  2. [Figure 5] Figure 5 caption: 'Normilized' should be 'Normalized'.
  3. [Section 3.2] Section 3.2 states DenseNet achieved a weighted average F1-score of 0.75 vs. 0.63, but Table 3 reports 0.74 vs. 0.66; the abstract also states 0.75. Please unify the numbers.
  4. [Sections 2.2.1–2.2.3] Section 2.2.1: the manual annotation includes both the delivery catheter and the clip with the same label, while the CAD templates (Section 2.2.2) represent only the clip. Section 2.2.3 computes the tip and axes from the full catheter+clip surface; please clarify why this does not bias the refined GT, or separate the catheter and clip annotations.
  5. [Section 2.2.2] Section 2.2.2: please specify the total number of templates and the exact mapping between the ten classification states and the template angles, since the current description is ambiguous.
  6. [Table 3] Table 3 includes a row for the 20° class with zero support; consider removing it or adding a comment that this class is absent from the test set.

Circularity Check

2 steps flagged · score 6.0 of 10

Post-refinement accuracy partly reduces to template-to-template agreement: the same CAD template library defines the ground truth (Sec. 2.2.3) and the final ICP refinement (Sec. 2.3.2); classification labels are also max-Dice template identities rather than simulator-known angles.

  1. fitted input called prediction [Section 2.2.3 (Clip Annotation CAD-based Refinement, steps 5-7), Section 2.3.2 (Template Matching Step), Section 3.3/Table 4]
    "To obtain accurate GT data, manual segmentations were automatically refined using the clip templates generated through CAD modeling. ... the co-registered template with maximal Dice score was used to refine the raw segmentation mask. ... the predicted segmentations were compared with the refined clip GTs described in Section 2.2.3. ... the surface of the clip template, selected based on the classification performed in the second step of the pipeline, was aligned to the surface of the predicted segmentation mask through the iterative closest point (ICP) algorithm."

    The GT is built by rigidly co-registering every CAD template to the manual mask and keeping the max-Dice template (Sec. 2.2.3). The final pipeline stage then fits a template from the same CAD library to the CNN mask via ICP (Sec. 2.3.2), and the evaluation compares the result against the template-derived GT (Sec. 2.4). Hence the post-refinement ASD of 0.75 mm and 95% HD of 2.05 mm in Table 4 compare one template fit against another template fit. Whenever the manual and predicted masks suggest the same configuration, the shared template prior guarantees small surface distances even if neither fit reflects the true image content. The refined 'prediction' is therefore not an independent localization estimate.

  2. fitted input called prediction [Section 2.1 (known simulator pose/opening angle), Section 2.2.3 (max-Dice template labels), Section 3.2/Table 3 (classification F1)]
    "TEE imaging was recorded at every stage of the procedure following the access of the delivery catheter into the LA. When the clip was close to the MV, imaging was acquired for each combination of clip pose and opening angle. ... the co-registered template with maximal Dice score was used to refine the raw segmentation mask."

    The simulator provided the true clip pose and opening angle at acquisition, but the paper does not use these as the classification target. The configuration label is instead the identity of the CAD template with maximal Dice against the manual segmentation. The DenseNet weighted F1 of 0.74 in Section 3.2 is therefore an agreement rate with an internally generated template-selection target, not with a physically measured configuration. Since the same template library is used in the final refinement, the classification and refinement stages are evaluated within a closed template-fitting loop.

full rationale

The paper's raw segmentation comparison retains some independent content: the CNN outputs a free-form mask and the raw Dice of 0.62 shows it does not trivially coincide with the template-refined GT. However, the GT itself is not an independent measurement; it is a CAD template chosen by max-Dice fit to the manual mask (Sec. 2.2.3). The post-refinement claim is substantially weaker because the final stage registers the same small CAD template library to the predicted mask (Sec. 2.3.2), so the reported ASD/HD largely measure template-to-template consistency rather than image-derived device localization. The classification evaluation is likewise anchored to template-selection labels even though the simulator's known opening angles could have provided an external reference. These are construction-level circularities, not mere missing ablations, and the manuscript does not discuss them as limitations. No load-bearing self-citation issue is present: the authors' prior works are cited for context (e.g., MV segmentation in [6,30]) but do not carry the clip-localization argument. Overall, central claims are partially circular, so the appropriate score is 6, not 0-2.

Assumptions & free parameters 5 free parameters · 5 assumptions · 1 invented entities

The central performance numbers rest primarily on the CAD templates: they define the ground truth, the classification labels, and the final refinement. Learned network weights are trained from data, but the template geometry, hyperparameters, and the simulator-to-clinic assumption are inputs the paper does not pay for with external evidence. The interpolation of clip dimensions across angles and the choice of 20° bins are hand-set. This makes the benchmark partly self-referential, though the physical scan of the clip provides some external grounding.

free parameters (5)
  • Dropout rate = 0.1
    Set via hyper-parameter search on early experiments (Section 2.3.1b); affects regularization of segmentation networks.
  • Loss weights = 0.6 Dice, 0.4 Focal
    Hand-chosen weighting of the segmentation loss (Section 2.3.1b); changes what the network optimizes.
  • CAD template arm dimensions and interpolation = Dimensions matched to scans at 60° and 120°; linearly interpolated for other angles
    The template geometries used for GT refinement and final registration are fitted to three scanned poses and interpolated (Section 2.2.2).
  • Configuration class binning = 10 bins, 0° to 180° in 20° steps
    Defines the classification task and template selection; assumes 20° resolution is clinically meaningful (Section 2.4).
  • ICP maximum iterations = 1000
    Set in Section 2.3.2 to balance accuracy and runtime; no convergence analysis.
assumptions (5)
  • domain assumption The heart simulator produces 3D TEE images whose appearance and artifacts are representative of in-vivo transesophageal imaging.
    All 196 images were acquired in a water-filled phantom (Section 2.1); the authors acknowledge in Limitations (Section 4.1) that in-vivo validation is still needed.
  • ad hoc to paper The template with maximal Dice score against a manual mask identifies the true clip configuration and geometry.
    Section 2.2.3 uses this rule to create ground truth; the rule is not validated against an independent measurement.
  • ad hoc to paper The CAD template library spans all clip configurations with sufficient geometric fidelity, including angles not directly scanned.
    Section 2.2.2 describes fitting scans at closed, 60°, and 120°, then interpolating dimensions; no independent check of intermediate geometries.
  • domain assumption Manual segmentations produced by two users with Otsu thresholding and manual cutting are accurate enough to seed template fitting.
    Section 2.2.1; no inter-annotator variability or comparison against an independent gold standard is reported.
  • domain assumption A random 70:10:20 split of the 196 images yields independent training and test sets.
    Section 2.3.1b; one frame per recording mitigates temporal leakage, but recordings share the same simulator setup, so full independence is not guaranteed.
invented entities (1)
  • CAD template models of the MitraClip independent evidence
    purpose: Shape priors used to refine manual ground truth and to register to predicted segmentations in the final pipeline step.
    The templates are derived from a 3D scan of a physical XTW clip (Section 2.2.2), so they have an external anchor, but the extension to all opening angles relies on the paper's interpolation and is not independently validated.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MitraClip Device Automated Localization in 3D Transesophageal Echocardiography via Deep Learning." pith.science (2026). https://pith.science/paper/YO4EJOSN

@misc{pith2026241215013,
  author       = {Pith},
  title        = {Pith review of: MitraClip Device Automated Localization in 3D Transesophageal Echocardiography via Deep Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YO4EJOSN}},
  note         = {Machine review of arXiv:2412.15013}
}
read the original abstract

The MitraClip is the most widely percutaneous treatment for mitral regurgitation, typically performed under the real-time guidance of 3D transesophagel echocardiography (TEE). However, artifacts and low image contrast in echocardiography hinder accurate clip visualization. This study presents an automated pipeline for clip detection from 3D TEE images. An Attention UNet was employed to segment the device, while a DenseNet classifier predicted its configuration among ten possible states, ranging from fully closed to fully open. Based on the predicted configuration, a template model derived from computer-aided design (CAD) was automatically registered to refine the segmentation and enable quantitative characterization of the device. The pipeline was trained and validated on 196 3D TEE images acquired using a heart simulator, with ground-truth annotations refined through CAD-based templates. The Attention UNet achieved an average surface distance of 0.76 mm and 95% Hausdorff distance of 2.44 mm for segmentation, while the DenseNet achieved an average weighted F1-score of 0.75 for classification. Post-refinement, segmentation accuracy improved, with average surface distance and 95% Hausdorff distance reduced to 0.75 mm and 2.05 mm, respectively. This pipeline enhanced clip visualization, providing fast and accurate detection with quantitative feedback, potentially improving procedural efficiency and reducing adverse outcomes.

Figures

Figures reproduced from arXiv: 2412.15013 by the authors.

Figure 1
Figure 1. Left: in-vitro setup for TEE acquisitions in TEER procedures. Black arrows indicate the TEE probe, the MitraClip and the MV. Right: example of In-vitro 3D TEE acquisition with MitraClip catheter. 2.2. Dataset Annotation 2.2.1. Clip Manual Annotation Two trained users segmented the MitraClip delivery catheter and clip, both assigned with the same label, using 3D Slicer [14]. Voxel thresholding was applied using the O… view at source ↗
Figure 2
Figure 2. d and Figure 2.e, show two examples of GT segmentation before [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 2
Figure 2. a)-c): CAD models of the clip in three different configurations. Only one [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figures from the paper (5 more)
Figure 3
Figure 3. Figure 3: Schematic representation of the proposed pipeline. Orange box: [PITH_FULL_IMAGE:figures/full_fig_p010_3.png]
Figure 4
Figure 4. Figure 4: Segmentation examples provided by the Attention UNet for closed (a-f) and [PITH_FULL_IMAGE:figures/full_fig_p014_4.png]
Figure 5
Figure 5. Figure 5: Confusion matrices for DenseNet and ResNet-50 classification performance [PITH_FULL_IMAGE:figures/full_fig_p016_5.png]
Figure 6
Figure 6. Figure 6: Boxplots for segmentation performance achieved by the Attention UNet with [PITH_FULL_IMAGE:figures/full_fig_p018_6.png]
Figure 7
Figure 7. Figure 7: Segmentation examples provided by Attention UNet for closed configurations [PITH_FULL_IMAGE:figures/full_fig_p019_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 31 canonical work pages

  1. [1]

    Vahanian, F

    A. Vahanian, F. Beyersdorf, F. Praz, M. Milojevic, S. Baldus, J. Bauer- sachs, D. Capodanno, L. Conradi, M. De Bonis, R. De Paulis, et al. , “2021 esc/eacts guidelines for the management of valvular heart dis- ease: developed by the task force for the management of valvular heart disease of the european society of cardiology (esc) and the european as- soc...

    work page 2021

  2. [2]

    Randomized comparison of percutaneous repair and surgery for mitral regurgitation: 5-year results of everest ii,

    T. Feldman, S. Kar, S. Elmariah, S. C. Smart, A. Trento, R. J. Siegel, P. Apruzzese, P. Fail, M. J. Rinaldi, R. W. Smalling,et al., “Randomized comparison of percutaneous repair and surgery for mitral regurgitation: 5-year results of everest ii,” Journal of the American College of Cardi- ology, vol. 66, no. 25, pp. 2844–2854, 2015

    work page 2015

  3. [3]

    Tran- scatheter or surgical repair for degenerative mitral regurgitation in el- derly patients: a propensity-weighted analysis,

    N. Buzzatti, M. Van Hemelrijck, P. Denti, S. Ruggeri, D. Schiavi, I. S. Scarf` o, D. Reser, M. Taramasso, A. Weber, G. La Canna,et al., “Tran- scatheter or surgical repair for degenerative mitral regurgitation in el- derly patients: a propensity-weighted analysis,” The Journal of thoracic and cardiovascular surgery, vol. 158, no. 1, pp. 86–94, 2019

    work page 2019

  4. [4]

    Mitraclip step by step; how to simplify the procedure,

    M. Sherif, L. Paranskaya, S. Yuecel, S. Kische, O. Thiele, G. D’Ancona, A. Neuhausen-Abramkina, J. Ortak, H. Ince, and A. ¨Oner, “Mitraclip step by step; how to simplify the procedure,” Netherlands heart journal, vol. 25, pp. 125–130, 2017

    work page 2017

  5. [5]

    Transcatheter mitral valve repair using the edge-to- edge clip,

    C. B. Nyman, G. B. Mackensen, S. Jelacic, S. H. Little, T. W. Smith, and F. Mahmood, “Transcatheter mitral valve repair using the edge-to- edge clip,” Journal of the American Society of Echocardiography, vol. 31, no. 4, pp. 434–453, 2018

    work page 2018

  6. [6]

    A deep learning-based fully automated pipeline for regurgitant mitral valve anatomy analysis from 3d echocar- diography,

    R. Munaf` o, S. Saitta, G. Ingallina, P. Denti, F. Maisano, E. Agri- cola, A. Redaelli, and E. Votta, “A deep learning-based fully automated pipeline for regurgitant mitral valve anatomy analysis from 3d echocar- diography,” IEEE Access, 2024

    work page 2024

  7. [7]

    Medical instrument detection in ultrasound: a review,

    H. Yang, C. Shan, A. F. Kolen, and P. H. de With, “Medical instrument detection in ultrasound: a review,” Artificial Intelligence Review, vol. 56, no. 5, pp. 4363–4402, 2023. 25

    work page 2023

  8. [8]

    Efficient catheter segmentation in 3d cardiac ultrasound using slice-based fcn with deep supervision and f-score loss,

    H. Yang, C. Shan, A. F. Kolen, and P. H. de With, “Efficient catheter segmentation in 3d cardiac ultrasound using slice-based fcn with deep supervision and f-score loss,” in 2019 IEEE International Conference on Image Processing (ICIP), pp. 260–264, IEEE, 2019

    work page 2019

Show all 34 references
  1. [9]

    Improving catheter segmentation & localization in 3d cardiac ultrasound using direction- fused fcn,

    H. Yang, C. Shan, A. F. Kolen, and P. H. de With, “Improving catheter segmentation & localization in 3d cardiac ultrasound using direction- fused fcn,” in 2019 IEEE 16th International Symposium on Biomedical Imaging (ISBI 2019) , pp. 1122–1126, IEEE, 2019

    2019

  2. [10]

    Transfer- ring from ex-vivo to in-vivo: Instrument localization in 3d cardiac ultra- sound using pyramid-unet with hybrid loss,

    H. Yang, C. Shan, T. Tan, A. F. Kolen, and P. H. de With, “Transfer- ring from ex-vivo to in-vivo: Instrument localization in 3d cardiac ultra- sound using pyramid-unet with hybrid loss,” in Medical Image Comput- ing and Computer Assisted Intervention–MICCAI 2019: 22nd Interna...

    2019

  3. [11]

    Effi- cient and robust instrument segmentation in 3d ultrasound using patch- of-interest-fusenet with hybrid loss,

    H. Yang, C. Shan, A. Bouwman, A. F. Kolen, and P. H. de With, “Effi- cient and robust instrument segmentation in 3d ultrasound using patch- of-interest-fusenet with hybrid loss,” Medical Image Analysis , vol. 67, p. 101842, 2021

    2021

  4. [12]

    Efficient medical instrument detection in 3d volumetric ultrasound data,

    H. Yang, C. Shan, A. F. Kolen, and P. H. de With, “Efficient medical instrument detection in 3d volumetric ultrasound data,” IEEE Trans- actions on Biomedical Engineering , vol. 68, no. 3, pp. 1034–1043, 2020

    2020

  5. [13]

    Accu- rate model-based segmentation of gynecologic brachytherapy catheter collections in mri-images,

    A. Mastmeyer, G. Pernelle, R. Ma, L. Barber, and T. Kapur, “Accu- rate model-based segmentation of gynecologic brachytherapy catheter collections in mri-images,” Medical image analysis, vol. 42, pp. 173–188, 2017

    2017

  6. [14]

    3d slicer: a platform for subject-specific image analysis, visualization, and clinical support,

    R. Kikinis, S. D. Pieper, and K. G. Vosburgh, “3d slicer: a platform for subject-specific image analysis, visualization, and clinical support,” in Intraoperative imaging and image-guided therapy, pp. 277–289, Springer, 2013

    2013

  7. [15]

    A threshold selection method from gray-level histograms.,

    N. Ostu, “A threshold selection method from gray-level histograms.,” IEEE Trans SMC, vol. 9, p. 62, 1979. 26

    1979

  8. [16]

    Laplace–beltrami spectra as ‘shape-dna’of surfaces and solids,

    M. Reuter, F.-E. Wolter, and N. Peinecke, “Laplace–beltrami spectra as ‘shape-dna’of surfaces and solids,” Computer-Aided Design, vol. 38, no. 4, pp. 342–366, 2006

    2006

  9. [17]

    3d u-net: learning dense volumetric segmentation from sparse annotation,

    ¨O. C ¸ i¸ cek, A. Abdulkadir, S. S. Lienkamp, T. Brox, and O. Ron- neberger, “3d u-net: learning dense volumetric segmentation from sparse annotation,” in Medical Image Computing and Computer-Assisted Intervention–MICCAI 2016: 19th International Conference, Athens, Greece, Oc...

    2016

  10. [18]

    3d mri brain tumor segmentation using autoencoder regularization,

    A. Myronenko, “3d mri brain tumor segmentation using autoencoder regularization,” 2019

    2019

  11. [19]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 770–778, 2016

    2016

  12. [20]

    Attention gated networks: Learning to leverage salient regions in medical images,

    J. Schlemper, O. Oktay, M. Schaap, M. Heinrich, B. Kainz, B. Glocker, and D. Rueckert, “Attention gated networks: Learning to leverage salient regions in medical images,” Medical image analysis , vol. 53, pp. 197–207, 2019

    2019

  13. [21]

    Mitral valve segmentation and tracking from transthoracic echocardiography using deep learning,

    S. V. Wifstad, H. A. Kildahl, B. Grenne, E. Holte, S. W. Hauge, S. Sæbø, D. Mekonnen, B. Nega, R. Haaverstad, M.-E. Estensen, et al. , “Mitral valve segmentation and tracking from transthoracic echocardiography using deep learning,” Ultrasound in Medicine & Biology , 2024

    2024

  14. [22]

    Unetr: Transformers for 3d medical image segmentation,

    A. Hatamizadeh, Y. Tang, V. Nath, D. Yang, A. Myronenko, B. Land- man, H. R. Roth, and D. Xu, “Unetr: Transformers for 3d medical image segmentation,” in Proceedings of the IEEE/CVF winter conference on applications of computer vision , pp. 574–584, 2022

    2022

  15. [23]

    Densely connected convolutional networks,

    G. Huang, Z. Liu, L. Van Der Maaten, and K. Q. Weinberger, “Densely connected convolutional networks,” in Proceedings of the IEEE confer- ence on computer vision and pattern recognition , pp. 4700–4708, 2017

    2017

  16. [24]

    Unified focal loss: Generalising dice and cross entropy-based losses to handle class imbal- anced medical image segmentation,

    M. Yeung, E. Sala, C.-B. Sch¨ onlieb, and L. Rundo, “Unified focal loss: Generalising dice and cross entropy-based losses to handle class imbal- anced medical image segmentation,” Computerized Medical Imaging and Graphics, vol. 95, p. 102026, 2022. 27

    2022

  17. [25]

    Pytorch: An imperative style, high- performance deep learning library,

    A. Paszke, S. Gross, F. Massa, A. Lerer, J. Bradbury, G. Chanan, T. Killeen, Z. Lin, N. Gimelshein, L. Antiga, A. Desmaison, A. Kopf, E. Yang, Z. DeVito, M. Raison, A. Tejani, S. Chilamkurthy, B. Steiner, L. Fang, J. Bai, and S. Chintala, “Pytorch: An imperative style, high- p...

    2019

  18. [26]

    Monai: An open-source framework for deep learning in healthcare,

    M. J. Cardoso, W. Li, R. Brown, N. Ma, E. Kerfoot, Y. Wang, B. Murrey, A. Myronenko, C. Zhao, D. Yang, et al. , “Monai: An open-source framework for deep learning in healthcare,” arXiv preprint arXiv:2211.02701, 2022

    2022 arXiv

  19. [27]

    Iterative point matching for registration of free-form curves and surfaces,

    Z. Zhang, “Iterative point matching for registration of free-form curves and surfaces,” International journal of computer vision , vol. 13, no. 2, pp. 119–152, 1994

    1994

  20. [28]

    Schroeder, K

    W. Schroeder, K. Martin, and B. Lorensen, The Visualization Toolkit (4th ed.). Kitware, 2006

    2006

  21. [29]

    A hybrid catheter localisation frame- work in echocardiography based on electromagnetic tracking and deep learning segmentation,

    F. Jia, S. Wang, and V. Pham, “A hybrid catheter localisation frame- work in echocardiography based on electromagnetic tracking and deep learning segmentation,” Computational Intelligence and Neuroscience , vol. 2022, no. 1, p. 2119070, 2022

    2022

  22. [30]

    Automatic 4d mitral valve segmentation from trans- esophageal echocardiography: a semi-supervised learning approach,

    R. Munaf` o, S. Saitta, D. Tondi, G. Ingallina, P. Denti, F. Maisano, E. Votta, et al. , “Automatic 4d mitral valve segmentation from trans- esophageal echocardiography: a semi-supervised learning approach,” Medical & Biological Engineering & Computing , pp. 1–16, 2025

    2025

  23. [31]

    Attention is all you need,

    A. Vaswani, “Attention is all you need,”Advances in Neural Information Processing Systems, 2017

    2017

  24. [32]

    Mitral transcatheter edge-to-edge repair with the pascal precision system: Device knobology and review of advanced steering maneuvers,

    S. Garcia, S. Elmariah, R. J. Cubeddu, F. Zahr, M. F. Eleid, S. K. Kodali, P. Seshiah, R. Sharma, and D. S. Lim, “Mitral transcatheter edge-to-edge repair with the pascal precision system: Device knobology and review of advanced steering maneuvers,” Structural Heart, vol. 8, n...

    2024

  25. [33]

    Transcatheter re- pair for patients with tricuspid regurgitation,

    P. Sorajja, B. Whisenant, N. Hamid, H. Naik, R. Makkar, P. Tadros, M. J. Price, G. Singh, N. Fam, S. Kar, et al. , “Transcatheter re- pair for patients with tricuspid regurgitation,” New England Journal of Medicine, vol. 388, no. 20, pp. 1833–1842, 2023

    2023

  26. [34]

    Pointnet++: Deep hierarchical feature learning on point sets in a metric space,

    C. R. Qi, L. Yi, H. Su, and L. J. Guibas, “Pointnet++: Deep hierarchical feature learning on point sets in a metric space,” Advances in neural information processing systems, vol. 30, 2017. 29

    2017

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.