Pith. sign in

REVIEW 3 major objections 7 minor 57 references

A template-morphing pipeline turns raw heart CT segmentations into watertight, point-matched meshes in minutes and builds a statistical model that can sample new anatomies.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

Template-based Chamfer-distance morphing converts defective cardiac CT segmentations into isotopological simulation-ready meshes and a PCA-GMM statistical shape model, validated on 58 healthy hearts.

T0 review reviewed 2026-07-12 challenge →

load-bearing objection Solid engineering pipeline that turns defective multi-chamber CT segmentations into isotopological, simulation-ready meshes in minutes, with open code and usable PCA-GMM cohorts on 58 healthy hearts. the 3 major comments →

arxiv 2607.02564 v1 pith:LESKAC3I submitted 2026-06-29 cs.CV cs.AIq-bio.TO

From Raw Segmentations to Simulation-Ready Cardiac Meshes: An Automated Framework for Anatomical Reconstruction and Virtual Cohort Generation

Francesco Fabbri , Martino Andrea Scarpolini , Paolo Ciancarella , Francesco Tudisco , Roberto Verzicco , Alessandro Ricci , Francesco Viola This is my paper
classification cs.CV cs.AIq-bio.TO
keywords statistical shape modelingcardiac mesh registrationChamfer distance morphingvirtual cohort generationwhole-heart reconstructionPCA-GMMsimulation-ready geometries
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The reading

Most computational heart studies still use one anatomy at a time because raw segmentations contain holes, broken junctions, and inconsistent topology that simulation solvers reject. This paper shows that a high-quality template can be non-rigidly deformed onto each patient's multi-chamber segmentation by minimizing chamber-wise Chamfer distance under Laplacian smoothing, producing watertight meshes that share identical connectivity and point correspondence. On 58 healthy CT hearts the morphogeometric descriptors (volumes, surfaces, axes, barycenter distances) are retained at reconstruction accuracies of 85–99 percent, and the entire cohort can be processed in roughly fifteen hours on a single GPU. Because the meshes are isotopological they live in a common shape space; a fifteen-mode PCA plus a three-component Gaussian mixture then captures most population variability and can generate synthetic yet anatomically plausible hearts. The open-source pipeline therefore supplies the missing link between clinical imaging and large-scale, uncertainty-aware multiphysics cohorts.

Core claim

A Chamfer-distance-driven, multi-resolution template registration converts imperfect multi-chamber CT segmentations into simulation-ready, topologically identical cardiac meshes within minutes per patient while preserving clinically relevant morphogeometric descriptors at high accuracy, thereby enabling a compact PCA–GMM statistical shape model of the whole heart and proximal vessels.

What carries the argument

Chamfer-distance morphing of a fixed high-quality template: each chamber of the template is driven toward the corresponding segmented surface by nearest-neighbor Chamfer loss, preconditioned by a combinatorial Laplacian that enforces smooth, topologically consistent deformations across chamber interfaces.

Load-bearing premise

A single healthy template deformed under Chamfer distance and Laplacian smoothing is assumed to capture the full range of anatomical variability without systematic bias; the work itself notes that rare or pathological shapes may need a library of templates, yet all validation uses only 58 healthy hearts.

What would settle it

Apply the identical pipeline to a set of hearts whose raw segmentations contain known topological anomalies or extreme morphologies (for example congenital defects); if the registered meshes either fail to remain watertight or systematically under-recover measured chamber volumes and axes relative to carefully curated ground truth, the single-template claim is falsified.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Any multiphysics heart solver that currently requires weeks of manual mesh repair can instead receive a new patient mesh in minutes.
  • Virtual cohorts drawn from the PCA–GMM can be used for population-level uncertainty quantification without additional imaging.
  • Because all meshes share point correspondence, regional statistics and simulation results can be averaged or compared across subjects without re-registration.
  • The same template-update loop can be re-run as larger CT or MRI cohorts become available, continuously refining the statistical shape model.

Where Pith is reading between the lines

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

  • If the method scales to pathological cohorts via a template library, it could supply the anatomical prior needed for in-silico device trials that must span both healthy and diseased anatomy.
  • The observed sex-linked size mode in the first PCA coordinate suggests that even modest latent-space sampling can already generate gender-stratified synthetic cohorts for sex-specific hemodynamic studies.
  • Because the registration is differentiable, gradients of any downstream simulation loss with respect to anatomy could in principle be back-propagated through the morphing step, opening a route to anatomy-aware optimization of devices or therapies.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 7 minor

Summary. The manuscript presents a semi-automatic pipeline that converts multi-chamber CT segmentations into watertight, isotopological, point-correspondence cardiac meshes by multi-scale Chamfer-distance morphing of a high-quality template under Laplacian-preconditioned gradient descent, with iterative template averaging. Applied to 58 healthy ECG-gated CT scans (all chambers plus proximal AO, PA, PV, VC), registration takes ~4 minutes per case on one GPU and retains morphogeometric descriptors (volumes, surfaces, axes, barycenter distances) at reconstruction accuracies of roughly 85–99% (Table 1, Fig. 4). Shared topology then enables a PCA shape model (d=15 explaining ~90% CEV; LOO reconstruction ~95%) and a GMM (K=3) for synthetic cohort sampling and sensitivity analysis. Code and the fitted SSM are released open-source.

Significance. The work targets a genuine bottleneck between modern DL segmentations and multiphysics cardiac simulation: topological defects, holes, and inconsistent openings that prevent direct use of clinical segmentations. If the reported fidelity and throughput hold, the pipeline substantially lowers the cost of building anatomically consistent virtual cohorts for uncertainty quantification and in-silico trials. Concrete strengths include the open-source release (pipeline + PCA/GMM), explicit multi-chamber Chamfer formulation with multi-resolution optimization, quantitative morphogeometric validation against the source segmentations, and LOO reconstruction tests of the SSM. The healthy-only, single-template scope is stated by the authors and bounds rather than negates the contribution inside that domain.

major comments (3)
  1. §2.2, Table 1 and §4.3: All morphogeometric validation compares registered meshes only to the homogenized segmentations that the method is designed to regularize. Those targets contain the holes, closed valve planes, and local defects illustrated in Fig. 1, so high Ra can partly reflect smoothing of segmentation artifacts rather than recovery of true anatomy. The Discussion correctly notes that 85% is a conservative lower bound, but an independent reference (expert-corrected meshes or a held-out manually curated subset) is still needed for a subset of cases to separate registration error from target noise; without it, the central fidelity claim rests on imperfect ground truth.
  2. Abstract / §1 / §3: The meshes are repeatedly described as “simulation-ready” for multiphysics (electromechanics, hemodynamics), yet the manuscript reports no solver run, mesh-quality metrics beyond Chamfer/morphogeometry (e.g., element aspect ratios, Jacobian positivity after volume meshing), or even a simple CFD/FSI smoke test on a registered geometry. Watertightness and isotopology are necessary but not sufficient. A short demonstration that at least one registered mesh (and one GMM sample) can be volume-meshed and advanced in a representative solver would make the central claim load-bearing rather than inferred.
  3. §2.1–2.3 and Methods: There is no quantitative baseline against alternative non-rigid registration or mesh-repair strategies (e.g., diffeomorphic surface registration, commercial mesh-fix workflows, or prior cardiac SSMs). Runtime and Ra are reported only for the proposed method. Even a small head-to-head on a subset of the 58 cases (or against the authors’ own prior single-structure morphing [7]) is needed to substantiate the claimed acceleration from “one week” of manual work to minutes and to show that Chamfer+Laplacian multi-scale morphing is competitive rather than merely workable.
minor comments (7)
  1. §4.3: The reconstruction-accuracy formula is written with ambiguous notation (P_i and mixed absolute/ℓ2 norms on scalar descriptors). Please rewrite Ra explicitly as an average over subjects of (1 − |q_H − q_R| / |q_H|) × 100% for scalar metrics.
  2. Fig. 2: Axis labels and legend text appear corrupted by glyph encoding in the manuscript PDF; ensure loss curves and epoch annotations render cleanly.
  3. §4.1: Vessel trimming ratios (0.7ℓ PA, 1.5ℓ AO) and the 10 mm PV/VC cut are free parameters that affect the SSM domain; state them as fixed protocol choices and note sensitivity if available.
  4. §2.3.1 / Fig. 5(d): AIC/BIC are reported “for completeness” while K is chosen from inertia/silhouette; make the decision rule primary in the figure caption so readers do not over-weight AIC/BIC on n=58, d=15.
  5. §4.2: Learning rate η=0.01 and γ∈{120,80} are taken from prior work; a brief ablation or justification for multi-chamber cardiac surfaces (vs. the original supra-aortic setting) would help reproducibility.
  6. Discussion: The suggestion of a template library for pathological topologies is important; a short statement on how point correspondence and the released SSM would be maintained across multiple templates would strengthen the path to broader use.
  7. Typos / style: “homogenized” vs “homogeneous” in Table 1 header; “tipically” → “typically” (§2.2); arXiv date line shows “29 Jun 2026” which appears inconsistent with the preprint stamp.

Circularity Check

1 steps flagged

Minor self-citation of the authors' prior Chamfer-registration formulation; all reported reconstruction accuracies, LOO PCA fidelity, and GMM sampling results are independently computed on the new multi-chamber cardiac cohort.

specific steps
  1. self citation load bearing [Section 4.2 Registration / Section 4.5 Implementation]
    "The registration algorithm follows the formulation introduced in [7], and is extended to support multi-chamber cardiac anatomies. ... A learning rate η=0.01 is used, following previous analysis in [7]. ... To capture geometric features at different spatial scales, the deformation approach in Equation 1 is applied sequentially at multiple mesh resolutions, following the strategy proposed in [7]."

    The core non-rigid morphing engine (Chamfer distance + Laplacian preconditioning + multi-resolution schedule) is taken from the authors' own prior aorta paper [7]. This is ordinary method reuse rather than a load-bearing uniqueness claim; the present multi-chamber cardiac results, iterative template averaging, and all quantitative validation remain independent of that citation and do not reduce to it by construction.

full rationale

The paper's central empirical claims (watertight isotopological meshes retaining morphogeometric descriptors at 85–99 % Ra on 58 hearts, compact PCA latent space, and GMM-generated virtual cohorts) rest on direct numerical evaluation of independent geometric descriptors (volumes, surfaces, axes, barycenter distances) and leave-one-out reconstruction metrics. These quantities are not algebraically identical to the Chamfer objective being minimized, nor are they fitted parameters re-labeled as predictions. The only self-citation of note is the reuse of the multi-scale Chamfer-plus-Laplacian registration scheme from Scarpolini et al. (2023) [7], which shares an author; that citation supplies the algorithmic substrate but is not invoked as a uniqueness theorem or as the sole justification of the present results. Template iteration, multi-chamber extension, GPA alignment, PCA/GMM construction, and all validation tables/figures are new and self-contained against the reported CT cohort. No self-definitional loop, fitted-input-as-prediction, or ansatz-smuggling reduction is present. Score 2 reflects the single non-load-bearing self-citation of method; the derivation chain itself is non-circular.

Axiom & Free-Parameter Ledger

5 free parameters · 4 axioms · 0 invented entities

The central claim rests on standard geometric optimization and statistical tools plus a handful of hand-chosen regularization and trimming parameters; no new physical entities are postulated. The main domain assumptions are that a single template topology is adequate for healthy hearts and that Chamfer-plus-Laplacian deformation preserves clinically relevant morphology.

free parameters (5)
  • Laplacian regularization strengths γ = 120 / 80
    Set to 120 (coarse) and 80 (fine) by heuristic to balance fidelity versus smoothness given segmentation noise; directly controls how closely the registered mesh follows local defects.
  • Learning rate η = 0.01
    Fixed at 0.01 following prior work; affects convergence of the preconditioned gradient descent.
  • PCA latent dimension d = 15
    Chosen as 15 to reach ~90 % CEV and ~95 % LOO reconstruction accuracy; used for all subsequent GMM and sampling.
  • GMM number of clusters K = 3
    Selected as 3 from inertia, silhouette, AIC/BIC curves on the limited cohort of 58 subjects.
  • Vessel trimming ratios = 0.7l / 1.5l
    0.7l for pulmonary artery and 1.5l for aorta chosen to homogenize field-of-view truncation; defines the anatomical domain of every mesh.
axioms (4)
  • domain assumption Chamfer distance between corresponding chamber point sets is a sufficient geometric fidelity measure for multi-chamber cardiac registration.
    Invoked as the sole data term in Eq. (1)–(2); no alternative metrics (Hausdorff, landmark, etc.) are compared.
  • domain assumption A single high-quality template topology is adequate for all healthy adult hearts in the cohort.
    All 58 registrations start from one manually curated mesh; pathological or topologically variant hearts are deferred to future multi-template work.
  • domain assumption Laplacian preconditioning produces anatomically realistic deformations without over-smoothing clinically relevant features.
    Used in the update rule x ← x − η(I+γL)−1 ∂D/∂x; γ values are chosen heuristically rather than cross-validated against independent clinical landmarks.
  • standard math Standard linear algebra (SVD for PCA, EM for GMM) correctly captures population shape variability once point correspondence is established.
    Applied after GPA alignment; no novel statistical theory is claimed.

reviewed 2026-07-12 · how reviews work

0 comments
Cite this review

Pith. "Pith review of From Raw Segmentations to Simulation-Ready Cardiac Meshes: An Automated Framework for Anatomical Reconstruction and Virtual Cohort Generation." pith.science (2026). https://pith.science/paper/LESKAC3I

@misc{pith2026260702564,
  author       = {Pith},
  title        = {Pith review of: From Raw Segmentations to Simulation-Ready Cardiac Meshes: An Automated Framework for Anatomical Reconstruction and Virtual Cohort Generation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LESKAC3I}},
  note         = {Machine review of arXiv:2607.02564}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Computational models of the human heart are widely used to study electromechanical and fluid-dynamical cardiac function and to support applications such as in silico clinical trials. However, most studies remain limited to single or patient-specific anatomies, restricting the inclusion of population-level variability required for uncertainty quantification. A key challenge is translating medical-image segmentations, which may contain artifacts, mesh defects or disjoint domains, into topologically coherent geometries suitable for multiphysics simulations. In this work, we present a semi-automatic pipeline that converts CT-based segmentations into simulation-ready cardiac meshes within a few minutes while preserving anatomical and topological consistency. Building on modern deep learning segmentation methods, the framework incorporates a template-based registration stage to regularize artifacts and enforce mesh-quality constraints. A Chamfer-distance morphing strategy deforms a high-quality template toward each segmented heart, matching individual chambers while preserving topology. The resulting meshes are watertight, isotopological, and endowed with consistent point-to-point correspondence. The pipeline is validated on 58 healthy cardiac CT scans, including all cardiac chambers and proximal vessel segments. The resulting meshes can be represented in a unified shape space, enabling the construction of a statistical shape model of the heart and major vessels. Principal Component Analysis shows that a low-dimensional latent space efficiently captures population variability, while Gaussian Mixture Modeling enables synthetic anatomy generation. Overall, the proposed framework (released open-source) provides a pathway from raw segmentations to simulation-ready cardiac geometries, enabling anatomically consistent virtual cohorts for large-scale in silico studies.

discussion (0)

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

Reference graph

Works this paper leans on

57 extracted references · 1 canonical work pages

  1. [1]

    McQueen, D. M. & Peskin, C. S. A three-dimensional computer model of the human heart for studying cardiac fluid dynamics.ACM Siggraph Computer Graphics34, 56–60 (2000)

    2000

  2. [2]

    Takizawa, K., Takagi, H., Tezduyar, T. E. & Torii, R. Estimation of element- based zero-stress state for arterial fsi computations.Computational Mechanics 54, 895–910 (2014)

    2014

  3. [3]

    URL https://doi.org/10.1115/ 1.2746378

    Dumont, K.et al.Comparison of the hemodynamic and thrombogenic perfor- mance of two bileaflet mechanical heart valves using a cfd/fsi model.Journal of Biomechanical Engineering129, 558–565 (2007). URL https://doi.org/10.1115/ 1.2746378

    2007

  4. [4]

    S., Nasif, M

    Abbas, S. S., Nasif, M. S. & Al-Waked, R. State-of-the-art numerical fluid– structure interaction methods for aortic and mitral heart valves simulations: A review.Simulation98, 3–34 (2022)

    2022

  5. [5]

    & Verzicco, R

    Viola, F., Meschini, V. & Verzicco, R. Fluid–structure-electrophysiology interac- tion (fsei) in the left-heart: A multi-way coupled computational model.European Journal of Mechanics - B/Fluids79, 212–232 (2020). URL https://www. sciencedirect.com/science/article/pii/S0997754619301505

    2020

  6. [6]

    URL https://www.sciencedirect.com/ science/article/pii/S001046552100360X

    Viola, F.et al.Fsei-gpu: Gpu accelerated simulations of the fluid–structure–electrophysiology interaction in the left heart.Computer Physics Communications273, 108248 (2022). URL https://www.sciencedirect.com/ science/article/pii/S001046552100360X

    2022

  7. [7]

    A., Mazzoli, M

    Scarpolini, M. A., Mazzoli, M. & Celi, S. Enabling supra-aortic vessels inclusion in statistical shape models of the aorta: a novel non-rigid registration method.Fron- tiers in PhysiologyVolume 14 - 2023(2023). URL https://www.frontiersin. org/journals/physiology/articles/10.3389/fphys.2023.1211461

    work page doi:10.3389/fphys.2023.1211461 2023

  8. [8]

    & Veneziani, A.Cardiovascular Mathematics: Modeling and simulation of the circulatory systemVol

    Formaggia, L., Quarteroni, A. & Veneziani, A.Cardiovascular Mathematics: Modeling and simulation of the circulatory systemVol. 1 (Springer Science & Business Media, 2010)

    2010

  9. [9]

    & Mittal, R

    Vedula, V., George, R., Younes, L. & Mittal, R. Hemodynamics in the left atrium and its effect on ventricular flow patterns.Journal of biomechanical engineering 137, 111003 (2015)

    2015

  10. [10]

    & Verzicco, R

    De Tullio, M., Cristallo, A., Balaras, E. & Verzicco, R. Direct numerical sim- ulation of the pulsatile flow through an aortic bileaflet mechanical heart valve. Journal of Fluid Mechanics622, 259–290 (2009). 28

    2009

  11. [11]

    A., Verzicco, R

    Lupi, V., Caruso Lombardi, F., Scarpolini, M. A., Verzicco, R. & Viola, F. Impact of blood rheology on left heart haemodynamics: Newtonian vs. non-newtonian modelling.European Journal of Mechanics - B/Fluids117, 204445 (2026). URL https://www.sciencedirect.com/science/article/pii/S0997754625002262

    2026

  12. [12]

    Boˇ snjak, D., Schussnig, R., Ranftl, S., Holzapfel, G. A. & Fries, T.-P. Geo- metric uncertainty of patient-specific blood vessels and its impact on aortic hemodynamics: A computational study.Computers in Biology and Medicine 190, 110017 (2025). URL https://www.sciencedirect.com/science/article/pii/ S0010482525003683

    2025

  13. [13]

    Cheng, R., Lai, Y. G. & Chandran, K. B. Three-dimensional fluid-structure interaction simulation of bileaflet mechanical heart valve flow dynamics.Annals of biomedical engineering32, 1471–1483 (2004)

    2004

  14. [14]

    URL https://doi.org/10.1093/pnasnexus/pgae392

    Davey, M.et al.Simulating cardiac fluid dynamics in the human heart.PNAS Nexus3, pgae392 (2024). URL https://doi.org/10.1093/pnasnexus/pgae392

    doi:10.1093/pnasnexus/pgae392 2024

  15. [15]

    Ataei, A.et al.Evaluation of inter-and intra-operator reliability of manual segmentation of femoral metastatic lesions.International journal of computer assisted radiology and surgery16, 1841–1849 (2021)

    2021

  16. [16]

    C.et al.Intra-and inter-operator variability in mri-based manual segmentation of hcc lesions and its impact on dosimetry.EJNMMI physics9, 90 (2022)

    Covert, E. C.et al.Intra-and inter-operator variability in mri-based manual segmentation of hcc lesions and its impact on dosimetry.EJNMMI physics9, 90 (2022)

    2022

  17. [17]

    & Voiculescu, I

    Fischer, S. & Voiculescu, I. ? (ed.)Quantifying inter-operator variability and its causes for medical semantic segmentation. (ed.?)2025 IEEE/CVF International Conference on Computer Vision Workshops (ICCVW), 7381–7389 (2025)

    2025

  18. [18]

    Sudlow, C.et al.Uk biobank: an open access resource for identifying the causes of a wide range of complex diseases of middle and old age.PLoS Medicine12, e1001779 (2015)

    2015

  19. [19]

    F.et al.Hvsmr-2.0: A 3d cardiovascular mr dataset for whole-heart segmentation in congenital heart disease.Scientific Data11, 721 (2024)

    Pace, D. F.et al.Hvsmr-2.0: A 3d cardiovascular mr dataset for whole-heart segmentation in congenital heart disease.Scientific Data11, 721 (2024)

    2024

  20. [20]

    L.et al.(eds)Imagechd: A 3d computed tomography image dataset for classification of congenital heart disease

    Xu, X.et al.Martel, A. L.et al.(eds)Imagechd: A 3d computed tomography image dataset for classification of congenital heart disease. (eds Martel, A. L. et al.)Medical Image Computing and Computer Assisted Intervention – MICCAI 2020, 77–87 (Springer International Publishing, Cham, 2020)

    2020

  21. [21]

    & Comaniciu, D

    Zheng, Y., Barbu, A., Georgescu, B., Scheuering, M. & Comaniciu, D. Four- chamber heart modeling and automatic segmentation for 3-d cardiac ct volumes using marginal space learning and steerable features.IEEE Transactions on Medical Imaging27, 1668–1681 (2008). 29

    2008

  22. [22]

    & Viehmann, T.Deep Learning with PyTorch: Build, train, and tune neural networks using Python tools(Manning, 2020)

    Stevens, E., Antiga, L. & Viehmann, T.Deep Learning with PyTorch: Build, train, and tune neural networks using Python tools(Manning, 2020)

    2020

  23. [23]

    Deep learning models in medical image analysis.Journal of Oral Biosciences64, 312–320 (2022)

    Tsuneki, M. Deep learning models in medical image analysis.Journal of Oral Biosciences64, 312–320 (2022)

    2022

  24. [24]

    & Zhu, H

    Li, M., Jiang, Y., Zhang, Y. & Zhu, H. Medical image analysis using deep learning algorithms.Frontiers in public health11, 1273253 (2023)

    2023

  25. [25]

    & Suk, H.-I

    Shen, D., Wu, G. & Suk, H.-I. Deep learning in medical image analysis.Annual review of biomedical engineering19, 221–248 (2017)

    2017

  26. [26]

    Wasserthal, J.et al.Totalsegmentator: Robust segmentation of 104 anatomic structures in ct images.Radiology: Artificial Intelligence5, e230024 (2023)

    2023

  27. [27]

    URL https://doi.org/10.1148/radiol.241613

    Akinci D’Antonoli, T.et al.Totalsegmentator mri: Robust sequence-independent segmentation of multiple anatomic structures in mri.Radiology314, e241613 (2025). URL https://doi.org/10.1148/radiol.241613. PMID: 39964271

    doi:10.1148/radiol.241613 2025

  28. [28]

    URL https://www.sciencedirect.com/science/article/pii/S0730725X12001816

    Fedorov, A.et al.3d slicer as an image computing platform for the quantita- tive imaging network.Magnetic Resonance Imaging30, 1323–1341 (2012). URL https://www.sciencedirect.com/science/article/pii/S0730725X12001816. Quanti- tative Imaging in Cancer

    2012

  29. [29]

    F., Kohl, S

    Isensee, F., Jaeger, P. F., Kohl, S. A. A., Petersen, J. & Maier-Hein, K. H. nnu-net: a self-configuring method for deep learning-based biomedical image segmentation. Nature Methods18, 203–211 (2021)

    2021

  30. [30]

    (ed.?)proceedings of Medical Image Computing and Computer Assisted Intervention – MICCAI 2024, Vol

    Isensee, F.et al.? (ed.)nnU-Net Revisited: A Call for Rigorous Validation in 3D Medical Image Segmentation. (ed.?)proceedings of Medical Image Computing and Computer Assisted Intervention – MICCAI 2024, Vol. LNCS 15009 (Springer Nature Switzerland, 2024)

    2024

  31. [31]

    Iglesias, J. E. & Sabuncu, M. R. Multi-atlas segmentation of biomedical images: a survey.Medical image analysis24, 205–219 (2015)

    2015

  32. [32]

    S.et al.Ai in medical imaging informatics: current challenges and future directions.IEEE journal of biomedical and health informatics24, 1837–1857 (2020)

    Panayides, A. S.et al.Ai in medical imaging informatics: current challenges and future directions.IEEE journal of biomedical and health informatics24, 1837–1857 (2020)

    2020

  33. [33]

    & Buntine, W

    Zheng, L.et al.Yanıko˘ glu, B. & Buntine, W. (eds)Automatic segmentation of aortic and mitral valves for heart surgical planning of hypertrophic obstructive cardiomyopathy. (eds Yanıko˘ glu, B. & Buntine, W.)Proceedings of the 15th Asian Conference on Machine Learning, Vol. 222 ofProceedings of Machine Learning Research, 1715–1730 (PMLR, 2024). URL https:...

    2024

  34. [34]

    & Frangi, A

    Ordas, S., Oubel, E., Leta, R., Carreras, F. & Frangi, A. F. ? (ed.)A statistical shape model of the heart and its application to model-based segmentation. (ed.?) Medical Imaging 2007: Physiology, Function, and Structure from Medical Images, Vol. 6511, 490–500 (SPIE, 2007)

    2007

  35. [35]

    URL https://www.sciencedirect.com/ science/article/pii/S1361841515001346

    Bai, W.et al.A bi-ventricular cardiac atlas built from 1000+ high resolution mr images of healthy subjects and an analysis of shape and motion.Medi- cal Image Analysis26, 133–145 (2015). URL https://www.sciencedirect.com/ science/article/pii/S1361841515001346

    2015

  36. [36]

    Rodero, C.et al.Linking statistical shape models and simulated function in the healthy adult human heart.PLoS computational biology17, e1008851 (2021)

    2021

  37. [37]

    URL https://www

    Kong, F.et al.Sdf4chd: Generative modeling of cardiac anatomies with congenital heart defects.Medical Image Analysis97, 103293 (2024). URL https://www. sciencedirect.com/science/article/pii/S1361841524002184

    2024

  38. [38]

    26 (Asiatic Society., 1933)

    Asiatic Society (Kolkata, I.Journal and Proceedings of the Asiatic Society of Bengalv. 26 (Asiatic Society., 1933). URL https://books.google.it/books?id= gOYRAAAAMAAJ

    1933

  39. [39]

    Adams, J., Karanam, M. S. T. & Elhabian, S. Point2ssm++: Self-supervised learning of anatomical shape models from point clouds.Medical Image Analysis 104073 (2026)

    2026

  40. [40]

    Kingma, D. P. & Welling, M. Auto-encoding variational bayes.arXiv preprint arXiv:1312.6114(2013)

    Pith/arXiv arXiv 2013

  41. [41]

    Kobyzev, I., Prince, S. J. & Brubaker, M. A. Normalizing flows: An introduc- tion and review of current methods.IEEE transactions on pattern analysis and machine intelligence43, 3964–3979 (2020)

    2020

  42. [42]

    J.et al.Generative adversarial nets.Advances in neural information processing systems27(2014)

    Goodfellow, I. J.et al.Generative adversarial nets.Advances in neural information processing systems27(2014)

    2014

  43. [43]

    & Abbeel, P

    Ho, J., Jain, A. & Abbeel, P. Denoising diffusion probabilistic models.Advances in neural information processing systems33, 6840–6851 (2020)

    2020

  44. [44]

    A., Iollo, A., Viola, F

    Fabbri, F., Scarpolini, M. A., Iollo, A., Viola, F. & Tudisco, F. Graph- convolutional-beta-vae for synthetic abdominal aortic aneurysm generation. Medical & Biological Engineering & Computing64, 827–845 (2026)

    2026

  45. [45]

    & Geveci, B

    Schroeder, W., Maynard, R. & Geveci, B. ? (ed.)Flying edges: A high- performance scalable isocontouring algorithm. (ed.?)2015 IEEE 5th symposium on large data analysis and visualization (ldav), 33–40 (IEEE, 2015). 31

    2015

  46. [46]

    A solution for the best rotation to relate two sets of vectors

    Kabsch, W. A solution for the best rotation to relate two sets of vectors. Foundations of Crystallography32, 922–923 (1976)

    1976

  47. [47]

    Gower, J. C. Generalized procrustes analysis.Psychometrika40, 33–51 (1975)

    1975

  48. [48]

    & Seidel, R

    Edelsbrunner, H., Kirkpatrick, D. & Seidel, R. On the shape of a set of points in the plane.IEEE Transactions on information theory29, 551–559 (2003)

    2003

  49. [49]

    & Van Dyk, D

    Meng, X.-L. & Van Dyk, D. The em algorithm—an old folk-song sung to a fast new tune.Journal of the Royal Statistical Society Series B: Statistical Methodology59, 511–567 (1997)

    1997

  50. [50]

    J., Moulavi, D

    Campello, R. J., Moulavi, D. & Sander, J. ? (ed.)Density-based clustering based on hierarchical density estimates. (ed.?)Pacific-Asia conference on knowledge discovery and data mining, 160–172 (Springer, 2013)

    2013

  51. [51]

    MacQueen, J. B. ? (ed.)Some methods for classification and analysis of mul- tivariate observations. (ed.?)Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability, Vol. 1, 281–297 (University of Califor- nia Press, Berkeley, CA, 1967). URL http://projecteuclid.org/euclid.bsmsp/ 1200512992

    1967

  52. [52]

    Rousseeuw, P. J. Silhouettes: A graphical aid to the interpretation and validation of cluster analysis.Journal of Computational and Applied Mathemat- ics20, 53–65 (1987). URL https://www.sciencedirect.com/science/article/pii/ 0377042787901257

    1987

  53. [53]

    & Viola, F

    Del Corso, G., Verzicco, R. & Viola, F. Sensitivity analysis of an electrophysiology model for the left ventricle.Journal of the Royal Society Interface17, 20200532 (2020)

    2020

  54. [54]

    & Kaszynski, A

    Sullivan, B. & Kaszynski, A. PyVista: 3D plotting and mesh analysis through a streamlined interface for the Visualization Toolkit (VTK).Journal of Open Source Software4, 1450 (2019). URL https://doi.org/10.21105/joss.01450

    doi:10.21105/joss.01450 2019

  55. [55]

    & Lorensen, B.The Visualization Toolkit (4th ed.) (Kitware, 2006)

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

    2006

  56. [56]

    Paszke, A.et al.Pytorch: An imperative style, high-performance deep learning library.Advances in neural information processing systems32(2019)

    2019

  57. [57]

    32 Extended Data Fig

    Ravi, N.et al.Accelerating 3d deep learning with pytorch3d.arXiv:2007.08501 (2020). 32 Extended Data Fig. 1: Sensitivity analysis Long axis Short lateral axis Short sagittal axisBarycenters distance/gid00023/gid00042/gid00039/gid00048/gid00040/gid00032 Fig. 9Sensitivity analysis of morphogeometric descriptors with respect to PCA modesz i. Each sen- sitivi...

    Pith/arXiv arXiv 2007

This paper was first reviewed by grok-4.5 on July 12, 2026.