Pith. sign in

REVIEW 4 major objections 4 minor 102 references

A physics-informed neural network for improving surface reconstruction of intracranial saccular aneurysms via variational membrane equilibrium

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

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

Pith's one-line read The paper claims a physics-informed neural network that reconstructs intracranial aneurysm surfaces by enforcing Laplace membrane equilibrium, selectively removing non-physical concave imaging artifacts while preserving genuine high-curvatu

desk verdict A genuinely new combination for aneurysm surface reconstruction, but the physics content is oversold and the key convexity premise is asserted rather than proven. read the letter →

arxiv 2607.22055 v1 pith:JBQJAWXW submitted 2026-07-24 math.NA cs.NA

Hyomin Ryu , Seung Hwan Kim , Jaemin Kim This is my paper
classification math.NAcs.NA MSC 65D1768T0792C10
keywords physics-informedneuralnetworksLaplacemembraneequilibriumintracranialsaccularaneurysmsurfacereconstructionB-splinesurfacesruptureriskassessmentimagingartifactremovalstressratio
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to establish that replacing the purely mathematical L-curve smoothing criterion with a biomechanical one—Laplace membrane equilibrium—gives more trustworthy aneurysm geometries from routine CTA/MRA images. It claims that a neural network trained with a variational equilibrium loss can tell apart concave imaging artifacts, which cannot sustain internal pressure under a membrane model, from genuine pathological features like rupture-prone blebs. On a patient case, the reconstructed surface's rupture-risk map concentrates at the dome apex, which a neurosurgeon judged consistent with intraoperative experience, unlike the scattered hotspots from conventional L-curve smoothing. If true, patient-specific rupture-risk assessment could be done directly from standard imaging without invasive measurement.

What carries the argument

The load-bearing object is the variational membrane equilibrium condition δΠ = 0, implemented as the squared gradient norm of the total potential energy with respect to the control-point displacement field. The stretching energy is approximated by the integral of the squared displacement gradient, the pressure work by the integral of the sum of principal curvatures, and the relative stiffness λ ≈ 500 couples them. Around this sit a Manifold-Consistent CNN whose padding replicates the closed-surface topology (periodic seam, pole convergence, open neck) and a purely geometric stress ratio σR = 2 − κ1/κ2 that makes the risk score independent of wall thickness and pressure.

What would settle it

Take a series of ruptured bifurcation aneurysms where the exact rupture point was recorded during surgical clipping, run the PINN reconstruction on the pre-rupture imaging, and check whether the predicted highest-risk hotspot matches the documented rupture site; if it does not in a meaningful fraction of cases, the claim that the equilibrium constraint preserves rupture-prone features would be refuted.

Watch

Extended reading notes

Core claim

The central claim is that enforcing δΠ = 0 (the first variation of total potential energy, balancing membrane stretching against pressure work) as a physics-informed loss in a B-spline-based PINN drives reconstructed intracranial saccular aneurysm surfaces toward a Laplace membrane equilibrium state. In that state, concave regions (negative mean curvature) are mechanically inadmissible, so the optimizer suppresses them as imaging artifacts; genuine convex high-curvature features that satisfy the equilibrium are preserved. The authors demonstrate on a clinical MRA case that this yields a rupture-risk map with the highest scores localized at the dome apex, consistent with the surgeon's intraop

Load-bearing premise

The central premise is that a pressurized thin-walled membrane cannot have concave regions at all, so every concave surface patch in the imaging data is a non-physical artifact that can be safely smoothed away; this ignores that real aneurysm walls have variable thickness and tension, which can allow genuine concavities at lobe junctions or the neck without violating Laplace equilibrium.

Editorial extensions

If this is right

  • The physics-informed loss replaces L-curve hyperparameter tuning with a biomechanical equilibrium criterion, removing the need to choose a smoothing factor that inherently flattens high-curvature features.
  • Rupture-risk maps computed from the PINN geometry concentrate at the dome apex, matching the established clinical pattern, whereas L-curve smoothing produces scattered hotspots that would mislead surgical planning.
  • Because the stress ratio is purely geometric, patient-specific risk maps can be derived from routine CTA/MRA without invasive pressure or thickness measurements.
  • The framework filters noise at the control-point level while preserving localized blebs, so downstream stress and hemodynamic simulations start from a more physically valid geometry.
  • The equilibrium-driven reconstruction is a step toward a computational biomarker for longitudinal surveillance of unruptured aneurysms.

Reading between the lines

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

  • If the concave premise holds, the same equilibrium filter might transfer to other pressurized membrane organs (aortic aneurysms, cardiac chambers) where artifact-versus-bleb ambiguity appears; this is an extrapolation beyond the paper's single-organ evidence.
  • A direct test of the discriminator would be to apply the framework to a cohort of ruptured aneurysms with surgically documented rupture points; a mismatch of hotspot and rupture site would indicate the equilibrium constraint is too strict.
  • The fixed λ ≈ 500 and the inconsistency between the two gradient expressions in Appendix B suggest the equilibrium loss's strength may be case-sensitive; the paper does not explore sensitivity to λ, and a sensitivity study would be needed before clinical adoption.
  • Single-patch parameterization already limits multilobed shapes; extending to multi-patch NURBS would reveal whether the equilibrium criterion can still discriminate artifacts at lobes and necks, where genuine concavities may exist.
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 / 4 minor

Summary. The paper proposes a physics-informed neural network (PINN) pipeline for reconstructing intracranial saccular aneurysm (ISA) surfaces from CTA/MRA point clouds. A B-spline surface is fitted to a structured point cloud, a Manifold-Consistent CNN predicts control-point displacements, and the total loss combines data fidelity, a variational-equilibrium term, and tangential-displacement suppression. The authors claim that the equilibrium loss enforces Laplace membrane equilibrium and thereby removes concave imaging artifacts while preserving genuine high-curvature features such as blebs. The reconstructed surface is then used to compute a geometry-derived risk score (σR = 2 − κ1/κ2), and the reported risk map concentrates at the dome apex, which the authors state is consistent with intraoperative observations. The method is demonstrated on a single side-wall unruptured ISA and compared against conventional L-curve-based smoothing.

Significance. If the central claims were established, the work would address a genuine clinical and computational need: incorporating biomechanical equilibrium into image-based reconstruction so that artifact removal does not also remove rupture-prone surface features. The manuscript contains some useful components, notably the manifold-consistent padding for the closed vascular surface and the use of C2-continuous B-splines to stabilize curvature-based analyses. However, the load-bearing physical premise is not demonstrated, the variational derivation is internally inconsistent, key hyperparameters are unreported, and the validation is largely circular relative to the prior that the method is designed to enforce. The claimed significance—reliable patient-specific rupture-risk maps from physics-informed reconstruction—is therefore not supported by the manuscript in its current form.

major comments (4)
  1. [Appendix B, Eqs. (B7)–(B8)] The first variation of the stretching energy is not the expression given in Eq. (B8). Eq. (B7) defines Ustretch as a quadratic functional in the gradient of the displacement field (with a 1/2 prefactor and squared gradient norms); its variational derivative is proportional to −Δd after integration by parts, not to λ∫∫(|∂u d| + |∂v d|) dudv. Eq. (B8) drops the squares, the 1/2 factor, and the modulus C. Consequently, the equilibrium loss in Eq. (12) is not the squared gradient norm of the stated potential Π, and the 'variational equilibrium' interpretation of the implemented loss is not established.
  2. [Section 3.2, Figs. 8–9] The assertion that a thin-walled pressurized membrane cannot have H < 0 is derived only under the uniform-isotropic-tension assumption T1 ≈ T2 = T_h used in Appendix B.2. In the general Laplace equilibrium κ1T1 + κ2T2 = P, localized concave patches (H < 0) are admissible when wall tensions vary spatially, as they do in real aneurysms with heterogeneous thickness, collagen anisotropy, and external tissue support. The paper's own Section 4.2 concedes that wall thickness variation and anisotropic collagen are neglected. Therefore Lequil is not a validated physics-based discriminator of artifacts versus genuine blebs; it is effectively a convexity prior. This premise is load-bearing for the artifact-removal and dome-apex risk-concentration claims.
  3. [Section 2.4.2, Appendix B.3, Eq. (B9)] The relative stiffness λ_rel ≈ 500 is introduced without reporting the constituent values of C, t, and T_h, and no sensitivity analysis is given. In addition, the loss weights λdata, λphys, and λtang in Eq. (10) are never reported. Because the final reconstructed geometry is the solution of a weighted optimization whose balance between data fidelity and the physics term directly controls the amount of artifact suppression, the unreported and apparently hand-chosen weights make the results non-reproducible and leave open the possibility that the reported outcome is largely an artifact of the weighting.
  4. [Sections 3.1, 3.3, 4.2] The validation is circular relative to the model's prior, and the diagnostic metric is not independently established. The success criterion—risk concentration at the dome apex—is the pattern that the convexity-regularized reconstruction is designed to produce. Section 4.2 admits that the validation is qualitative and that side-wall aneurysm ruptures rarely provide intraoperative confirmation. Moreover, Eq. (14) postulates T1 = P/(2κ2) without derivation; Laplace's equation alone does not determine the two principal tensions from geometry and pressure, so σR = 2 − κ1/κ2 is an ad hoc geometric index rather than a demonstrated biomechanical rupture-risk measure. The comparison with L-curve smoothing shows that the method produces the expected pattern, but does not establish preservation of genuine blebs or accuracy of the risk map.
minor comments (4)
  1. [Figure 10(b), Eq. (19)] Eq. (19) defines CRMS as an RMS difference without any additive constant, but Figure 10(b) displays a '+3.1377e5' offset on the right axis. Please clarify whether this is an axis-shift artifact and report CRMS directly; as written, the offset is inconsistent with the definition and confuses the claim that the curvature error is 'drastically altered.'
  2. [Nomenclature] The symbol P is used both for transmural pressure and for control-point arrays (P, P_macro, P_base). These are conceptually different objects; please use distinct notation to avoid ambiguity in Eqs. (6)–(8) and (14).
  3. [Section 3.1, Eq. (16)] The risk score uses a logarithm of σR = 2 − κ1/κ2, but σR can be non-positive for some admissible curvature combinations even when H > 0. The domain of the logarithmic normalization is not discussed.
  4. [Figure C1] The convergence curves are described only qualitatively. The authors should report the actual loss weights, the final loss values, and the converged λ_rel used in the computation.

Circularity Check

2 steps flagged · score 6.0 of 10

The artifact discriminator is self-definitional: artifacts are defined as H<0 and the uniform-tension equilibrium loss is constructed so that H>0, so the reported removal of concavities and the dome-apex risk concentration are partly built into the loss rather than independently predicted.

  1. self definitional [Section 3.2 (artifact criterion); Eq. (12) and Appendix B.2 (loss construction)]
    "a thin-walled membrane under positive internal pressure—analogous to a rubber balloon (Boo et al., 2026)—cannot sustain concave surface profiles under Laplace membrane equilibrium. Because ISA pathogenesis shares this loading condition, the mean curvature must remain strictly positive (H >0) on the aneurysm wall."

    The paper defines the artifacts to be removed as exactly the H<0 regions, then builds Lequil from a potential whose pressure work is Wpress ∝ ∫(κ1+κ2)dA under the Appendix B assumption T1≈T2=Th=const. With that assumption Laplace's equation reduces to H = P/(2Th) > 0. Thus H>0 is an input of the loss, not a finding. The reported elimination of negative-curvature regions is the optimizer minimizing the very objective designed to enforce convexity; it cannot discriminate artifact from bleb except by the pre-asserted sign of H. The paper's own Limitations concede that wall thickness variation and anisotropy—mechanisms that allow concave Laplace equilibria with nonuniform tension—are neglected, so the premise is an assumption rather than a consequence of the physics.

  2. other [Section 3.3 (validation); Eqs. (15)-(16)]
    "The PINN-based risk map (Figure 11b) concentrates the highest risk scores at the dome apex, consistent with this well-established finding."

    Risk Score is defined purely from the principal curvatures (Eq. 15: σR = 2 − κ1/κ2; Eq. 16: min-max log normalization). The same optimization, by construction, drives the geometry toward a convex, smooth Laplace-equilibrium shape, so the curvature field entering the risk score is the regularized field produced by the loss. The dome-apex concentration is therefore largely a restatement of the imposed convex equilibrium plus the normalization, used as confirmation of the method; it is not an independent check that the physics loss identified true rupture sites rather than merely enforcing H>0. As the paper acknowledges, validation is qualitative, without direct comparison to confirmed rupture sites.

full rationale

The paper is not wholly circular because the data-fidelity term anchors the surface to the measured point cloud and the final risk map is checked against a neurosurgeon's qualitative intraoperative experience, which is external to the loss. However, the central artifact-discrimination claim is circular in a narrow sense: the 'non-physical' class is defined by H<0, and the equilibrium loss is constructed from a uniform-tension potential that entails H>0, so removing H<0 is the loss's designed effect rather than a physics-informed discovery. The Appendix B inconsistency (Eq. B7 has squared gradient norms and a 1/2 prefactor; Eq. B8 drops both) means the implemented Lequil is not demonstrably the variation of the stated potential, further weakening the 'physics' label. Self-citations to Kim et al. 2023a,b for Laplace's equation and Boo et al. 2026 for the balloon analogy are not the load-bearing circularity because the Laplace relation is independently standard. Overall, one central construction reduces a headline prediction to its input, so score 6.

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

The central equilibrium loss rests on a hand-set stiffness (λ≈500) and the assumption that a pressurized membrane is necessarily convex; the stress-ratio risk score is computed under a particular tension split. The claim of artifact/bleb discrimination therefore inherits all these assumptions. No new physical entities are introduced.

free parameters (3)
  • λ_rel (relative stiffness) = ≈500
    Balances stretching energy vs pressure work in the equilibrium loss; estimated in Appendix B (Eq. B9) under the assumption T1≈T2 and uniform thickness, with no sensitivity analysis.
  • λdata, λphys, λtang (loss weights) = not reported
    Weights in Eq. (10); the paper never gives their values, though the behavior of the reconstruction depends on them.
  • Control-point grid size (nθ, nϕ) = 16×8 (adjusted from 18×9)
    The Nyquist-style derivation in Appendix A yields 18×9, but the number is reduced to 16×8 to fit the CNN architecture; this under-samples relative to the paper's own resolution bound.
assumptions (5)
  • domain assumption A pressurized thin-walled membrane cannot have concave surface regions (H > 0 is required)
    Section 3.2 uses this to classify all concave regions as non-physical artifacts and to drive the filtering; it is asserted, not derived, and ignores heterogeneous wall properties and external loads.
  • domain assumption Isotropic linear elastic membrane, uniform tension T1≈T2=Th, steady-state, negligible bending
    Appendix B reduces the aneurysm wall to a membrane with uniform tension; thickness variation, collagen anisotropy, and hemodynamic loading are neglected (acknowledged in Section 4.2).
  • ad hoc to paper σR = 2 − κ1/κ2 is a valid rupture-risk indicator
    Eq. (15) is derived under the particular tension split T1=P/(2κ2), T2=P/κ2(1−κ1/(2κ2)); there is no independent clinical validation of σR as a risk marker.
  • domain assumption The segmented point cloud is an unbiased sample of the true surface
    The data fidelity loss compares against the segmented CTA/MRA points; segmentation errors and partial-volume effects are not modeled.
  • standard math Hadamard–Zolésio theorem justifies suppressing all tangential control-point displacement
    Section 2.4.2, Eq. (13) uses the theorem to argue only normal displacements matter; applying it to a discrete control-point displacement field is an extrapolation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A physics-informed neural network for improving surface reconstruction of intracranial saccular aneurysms via variational membrane equilibrium." pith.science (2026). https://pith.science/paper/JBQJAWXW

@misc{pith2026260722055,
  author       = {Pith},
  title        = {Pith review of: A physics-informed neural network for improving surface reconstruction of intracranial saccular aneurysms via variational membrane equilibrium},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JBQJAWXW}},
  note         = {Machine review of arXiv:2607.22055}
}
abstract

Intracranial saccular aneurysms (ISAs) pose severe health risks, yet conventional population-based risk stratification scores (PHASES, UIATS, and ELAPSS) offer limited capacity for patient-specific rupture risk assessment. Image-based computational approaches have gained prominence, but traditional surface reconstruction relies on mathematical smoothing (e.g., L-curve criteria) that indiscriminately suppresses both imaging artifacts and genuine pathological features such as rupture-prone blebs. Although Laplace's membrane equilibrium ($\kappa_1 T_1 + \kappa_2 T_2 = P$) has long governed aneurysm wall mechanics (Humphrey and Kyriacou [Neurol. Res., 18 (1996)]), its integration into geometric reconstruction pipelines remains unexplored. This work introduces a physics-informed neural network (PINN) framework with B-spline representations, whose key contributions are: (i) embedding the variational equilibrium condition ($\delta \Pi = 0$) as a physics-informed loss that replaces the mathematical L-curve criterion with a biomechanically grounded artifact discrimination, (ii) developing a Manifold-Consistent CNN ansatz that preserves the closed-surface topology of vascular geometries, and (iii) establishing a Laplace equilibrium-driven reconstruction that filters imaging noise while preserving diagnostically critical high-curvature features. Application to patient-specific clinical datasets demonstrates that the framework eliminates non-physical concave artifacts without compromising genuine geometric anomalies. Clinical evaluation by a practicing neurosurgeon confirms that the resulting risk map---with rupture risk concentrated at the dome apex---is consistent with intraoperative observations, establishing a computational biomarker foundation for patient-specific rupture risk assessment of intracranial saccular aneurysms.

Figures

Figures reproduced from arXiv: 2607.22055 by the authors.

Figure 1
Figure 1. Schematic workflow of the proposed analysis-led modeling framework. The pipeline integrates clinical MRA data [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Representative workflow of 3D geometric reconstruction in 3D Slicer. (a) Coronal slice of a 3D TOF Brain MRA with [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Schematic workflow of the unstructured point cloud extraction. ⃝1 Voxelization of the ISA and parent vessel from MRA or CTA imaging data. ⃝2 Surface mesh generation in 3D Slicer. ⃝3 Manual separation of the ISA dome from the parent vessel. ⃝4 Identification of the bottom (neck) plane via normal vector analysis on the mesh surface. ⃝5 Equispaced slicing along the meridional direction (v = 0 to v = 1) and extraction o… view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: NURBS curve components. (a) Periodic curve with knot vector U ( [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Components for surface reconstruction. (a) The raw ISA point cloud extracted from the mesh as described in Section [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: Overview of the convolutional layers that extract spatial features (illustrated for the first channel) and predict the [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 7
Figure 7. Figure 7: Schematic of the Manifold-Consistent Padding. Flattening the closed 3D ISA manifold into a 2D patch disrupts [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]
Figure 8
Figure 8. Figure 8: Evolution of signed log mean curvature profiles over the parametric domain [PITH_FULL_IMAGE:figures/full_fig_p014_8.png]
Figure 9
Figure 9. Figure 9: Evolution of the log risk score maps over the parametric domain [PITH_FULL_IMAGE:figures/full_fig_p014_9.png]
Figure 10
Figure 10. Figure 10: L-curve analysis for the smoothing factor [PITH_FULL_IMAGE:figures/full_fig_p015_10.png]
Figure 11
Figure 11. Figure 11: Comparison of rupture risk maps between the conventional L-curve method and the proposed PINN framework [PITH_FULL_IMAGE:figures/full_fig_p016_11.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

102 extracted references · 2 linked inside Pith

  1. [1]

    Medical & Biological Engineering & Computing , volume =

    An image-based modeling framework for patient-specific computational hemodynamics , author =. Medical & Biological Engineering & Computing , volume =

  2. [2]

    and Farley, C

    Banatwala, M. and Farley, C. and Feinberg, D. and Humphrey, J. , journal =. Parameterization of the shape of intracranial saccular aneurysms using

  3. [3]

    Computer-Aided Design , year =

    Surface fitting of structured point clouds , author =. Computer-Aided Design , year =

  4. [4]

    Scientific Reports , volume =

    Span of regularization for solution of inverse problems with application to magnetic resonance relaxometry of the brain , author =. Scientific Reports , volume =

  5. [5]

    Geometric Modeling and Characterization of the Circle of

    Bogunovi\'. Geometric Modeling and Characterization of the Circle of

  6. [6]

    Numerische Mathematik , volume =

    Smoothing noisy data with spline functions , author =. Numerische Mathematik , volume =

  7. [7]

    Proceedings of the 26th Annual Conference on Computer Graphics and Interactive Techniques , pages =

    Implicit fairing of irregular meshes using diffusion and curvature flow , author =. Proceedings of the 26th Annual Conference on Computer Graphics and Interactive Techniques , pages =

  8. [8]

    Curve and Surface Fitting with Splines , author =

Show all 102 references
  1. [9]

    Farley, C. W. , school =. Parameterization of the Size and Shape of Intracranial Saccular Aneurysms Using

  2. [10]

    and Beichel, R

    Fedorov, A. and Beichel, R. and Kalpathy-Cramer, J. and Finet, J. and Fillion-Robin, J.-C. and Pujol, S. and others and Kikinis, R. , journal =

  3. [11]

    The effect of noise and

    Fern. The effect of noise and. Journal of Applied Geophysics , volume =

  4. [12]

    Fessler, J. A. , journal =. Model-based image reconstruction for

  5. [13]

    and others , journal =

    Geng, X. and others , journal =. Quantitative assessment of the effects of spatial smoothing on cortical surface-based

  6. [14]

    and others , journal =

    Gerig, G. and others , journal =. Nonlinear anisotropic filtering of

  7. [15]

    Hansen, P. C. , journal =. Analysis of discrete ill-posed problems by means of the

  8. [16]

    arXiv preprint arXiv:2001.08248 , year =

    How much position information do convolutional neural networks encode? , author =. arXiv preprint arXiv:2001.08248 , year =

    2001 arXiv

  9. [17]

    Advances in Neural Information Processing Systems , volume =

    Spatial transformer networks , author =. Advances in Neural Information Processing Systems , volume =

  10. [18]

    The International Journal of Advanced Manufacturing Technology , volume =

    Unconstrained and constrained curve fitting for reverse engineering , author =. The International Journal of Advanced Manufacturing Technology , volume =

  11. [19]

    Journal of Computational Physics , volume =

    Model reduction of dynamical systems on nonlinear manifolds using deep convolutional autoencoders , author =. Journal of Computational Physics , volume =

  12. [20]

    Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages =

    Fully convolutional networks for semantic segmentation , author =. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages =

  13. [21]

    Lorensen, W. E. and Cline, H. E. , journal =. Marching cubes: A high resolution

  14. [22]

    Advances in Neural Information Processing Systems , volume =

    Understanding the effective receptive field in deep convolutional neural networks , author =. Advances in Neural Information Processing Systems , volume =

  15. [23]

    Annals of Biomedical Engineering , volume =

    Three-dimensional geometrical characterization of cerebral aneurysms , author =. Annals of Biomedical Engineering , volume =

  16. [24]

    and Corpuz, A

    Moola, A. and Corpuz, A. M. and Burkhart, M. J. and Ross, C. J. and Mir, A. and Burkhart, H. M. and others and Pawar, A. , journal =

  17. [25]

    IEEE Transactions on Medical Imaging , volume =

    A framework for geometric analysis of vascular structures: application to cerebral aneurysms , author =. IEEE Transactions on Medical Imaging , volume =

  18. [26]

    and Tiller, W

    Piegl, L. and Tiller, W. , publisher =. The

  19. [27]

    and Fischer, P

    Ronneberger, O. and Fischer, P. and Brox, T. , booktitle =

  20. [28]

    and others , journal =

    Styner, M. and others , journal =. Framework for the statistical shape analysis of brain structures using

  21. [29]

    2024 , url =

    Simpleware. 2024 , url =

    2024

  22. [30]

    and others , journal =

    Tang, W. and others , journal =. Neural co-optimization of knots and parameterization for

  23. [31]

    Proceedings of the 22nd Annual Conference on Computer Graphics and Interactive Techniques , pages =

    A signal processing approach to fair surface design , author =. Proceedings of the 22nd Annual Conference on Computer Graphics and Interactive Techniques , pages =

  24. [32]

    , journal =

    Unser, M. , journal =. Sampling---50 years after

  25. [33]

    and Bazilevs, Y

    Zhang, Y. and Bazilevs, Y. and Goswami, S. and Bajaj, C. L. and Hughes, T. J. , journal =. Patient-specific vascular

  26. [34]

    Journal of NeuroInterventional Surgery , volume =

    Blebs in intracranial aneurysms: prevalence and general characteristics , author =. Journal of NeuroInterventional Surgery , volume =

  27. [35]

    Neurosurgical Review , volume =

    Factors affecting formation and rupture of intracranial saccular aneurysms , author =. Neurosurgical Review , volume =

  28. [36]

    European Radiology , volume =

    Compressed sensing time-of-flight magnetic resonance angiography with high spatial resolution for evaluating intracranial aneurysms: comparison with digital subtraction angiography , author =. European Radiology , volume =

  29. [37]

    and others , journal =

    Berg, P. and others , journal =. Does the

  30. [38]

    New England Journal of Medicine , volume =

    Cerebral aneurysms , author =. New England Journal of Medicine , volume =

  31. [39]

    Boo, S. H. and Kim, J. , journal =. Multiphysics modeling of surface diffusion coupled with large deformation in

  32. [40]

    Annals of Biomedical Engineering , year =

    Statistical wall shear stress maps of ruptured and unruptured middle cerebral artery aneurysms , author =. Annals of Biomedical Engineering , year =

  33. [41]

    and Zhao, Y

    Cui, X. and Zhao, Y. and Zhang, Y. and Yao, Y. and Yang, P. and Liu, D. and others and Yang, X. , journal =. Prevalence, geometry, and hemodynamics of small and medium-sized intracranial aneurysms with and without blebs in the. 2025 , note =

    2025

  34. [42]

    Neurosurgery , volume =

    Morphology parameters for intracranial aneurysm rupture risk assessment , author =. Neurosurgery , volume =

  35. [43]

    Hollingsworth, K. G. , journal =. Reducing acquisition time in clinical

  36. [44]

    and others , journal =

    Huston, J. and others , journal =. Unruptured intracranial aneurysms: evaluation with

  37. [45]

    , journal =

    Inagawa, T. , journal =. Risk factors for the formation and rupture of intracranial saccular aneurysms in

  38. [46]

    Annals of Biomedical Engineering , volume =

    In vivo deformation of the human basilar artery , author =. Annals of Biomedical Engineering , volume =

  39. [47]

    Extreme Mechanics Letters , volume =

    A model for mechanosensitive cell migration in dynamically morphing soft tissues , author =. Extreme Mechanics Letters , volume =

  40. [48]

    arXiv preprint arXiv:2308.02979 , year =

    Modeling the mechanosensitive collective migration of cells on the surface and the interior of morphing soft tissues , author =. arXiv preprint arXiv:2308.02979 , year =

    arXiv

  41. [49]

    and Kim, J

    Kim, S. and Kim, J. , journal =. A phase-field fracture model for

  42. [50]

    and Donoho, D

    Lustig, M. and Donoho, D. L. and Santos, J. M. and Pauly, J. M. , journal =. Compressed sensing

  43. [51]

    Journal of Neurosurgery , volume =

    Morphological and clinical risk factors for the rupture of anterior communicating artery aneurysms , author =. Journal of Neurosurgery , volume =

  44. [52]

    and Mihalea, C

    Pagiola, I. and Mihalea, C. and Caroff, J. and others , journal =. The

  45. [53]

    Neurosurgical Review , volume =

    Retrospective application of risk scores to ruptured intracranial aneurysms: would they have predicted the risk of bleeding? , author =. Neurosurgical Review , volume =

  46. [54]

    StatPearls , publisher =

    Pediatric Cerebral Aneurysm , author =. StatPearls , publisher =

  47. [55]

    and Mao, Z

    Cai, S. and Mao, Z. and Wang, Z. and Yin, M. and Karniadakis, G. E. , journal =. Physics-informed neural networks (

  48. [56]

    Journal of Heat Transfer , volume =

    Physics-informed neural networks for heat transfer problems , author =. Journal of Heat Transfer , volume =

  49. [57]

    and Qi, L

    Hu, H. and Qi, L. and Chao, X. , journal =. Physics-informed neural networks (

  50. [58]

    Nature Reviews Physics , volume =

    Physics-informed machine learning , author =. Nature Reviews Physics , volume =

  51. [59]

    Multiscale Science and Engineering , volume =

    A review of physics informed neural networks for multiscale analysis and inverse problems , author =. Multiscale Science and Engineering , volume =

  52. [60]

    and Hutter, F

    Loshchilov, I. and Hutter, F. , journal =

  53. [61]

    ICLR 2019 , year =

    Decoupled weight decay regularization , author =. ICLR 2019 , year =

    2019

  54. [62]

    2020 IEEE Power & Energy Society General Meeting (PESGM) , pages =

    Physics-informed neural networks for power systems , author =. 2020 IEEE Power & Energy Society General Meeting (PESGM) , pages =

    2020

  55. [63]

    Journal of Computational Physics , volume =

    Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations , author =. Journal of Computational Physics , volume =

  56. [64]

    SIAM Journal on Scientific Computing , volume =

    Understanding and mitigating gradient flow pathologies in physics-informed neural networks , author =. SIAM Journal on Scientific Computing , volume =

  57. [65]

    and Meng, X

    Yang, L. and Meng, X. and Karniadakis, G. E. , journal =

  58. [66]

    and Raissi, M

    Perdikaris, P. and Raissi, M. and Karniadakis, G. E. , journal =. Nonlinear information fusion in

  59. [67]

    and Sun, L

    Gao, H. and Sun, L. and Wang, J. X. , journal =

  60. [68]

    Computer Methods in Applied Mechanics and Engineering , volume =

    A physics-informed deep learning framework for inversion and surrogate modeling in solid mechanics , author =. Computer Methods in Applied Mechanics and Engineering , volume =

  61. [69]

    and Espath, L

    Saidaoui, H. and Espath, L. and Tempone, R. , journal =. Deep

  62. [70]

    Computer Methods in Applied Mechanics and Engineering , volume =

    Physics informed surface autoencoders for thin shell analysis , author =. Computer Methods in Applied Mechanics and Engineering , volume =

  63. [71]

    International Journal of Mechanical Sciences , volume =

    Graph convolutional networks-driven multiscale topology optimization based on polygonal coarse-grid elements , author =. International Journal of Mechanical Sciences , volume =

  64. [72]

    Computer Methods in Biomechanics and Biomedical Engineering , volume =

    Simulation of transcatheter aortic valve implantation: a patient-specific finite element approach , author =. Computer Methods in Biomechanics and Biomedical Engineering , volume =

  65. [73]

    Journal of Elasticity , volume =

    Competition between radial expansion and thickening in the enlargement of an intracranial saccular aneurysm , author =. Journal of Elasticity , volume =

  66. [74]

    Journal of the Royal Society Interface , volume =

    Flow instability and wall shear stress variation in intracranial aneurysms , author =. Journal of the Royal Society Interface , volume =

  67. [75]

    Computational Mechanics , volume =

    Isogeometric fluid--structure interaction analysis with applications to arterial blood flow , author =. Computational Mechanics , volume =

  68. [76]

    American Journal of Neuroradiology , volume =

    Characterization of cerebral aneurysms for assessing risk of rupture by using patient-specific computational hemodynamics models , author =. American Journal of Neuroradiology , volume =

  69. [77]

    American Journal of Neuroradiology , volume =

    Hemodynamics and bleb formation in intracranial aneurysms , author =. American Journal of Neuroradiology , volume =

  70. [78]

    American Journal of Neuroradiology , volume =

    Association of hemodynamic characteristics and cerebral aneurysm rupture , author =. American Journal of Neuroradiology , volume =

  71. [79]

    Journal of Vascular Surgery , volume =

    Prediction of rupture risk in abdominal aortic aneurysm during observation: Wall stress versus diameter , author =. Journal of Vascular Surgery , volume =

  72. [80]

    Computational Mechanics , volume =

    Fluid--structure interaction analysis of bioprosthetic heart valves: significance of arterial wall deformation , author =. Computational Mechanics , volume =

  73. [81]

    and Woodward, S

    Hoi, Y. and Woodward, S. H. and Kim, M. and Taulbee, D. B. and Meng, H. , journal =. Validation of

  74. [82]

    Humphrey, J. D. and Kyriacou, S. K. , journal =. The use of

  75. [83]

    Journal of Elasticity , volume =

    Structure, mechanical properties, and mechanics of intracranial saccular aneurysms , author =. Journal of Elasticity , volume =

  76. [84]

    Journal of Biomechanics , volume =

    Finite strain elastodynamics of intracranial saccular aneurysms , author =. Journal of Biomechanics , volume =

  77. [85]

    Computer Methods in Biomechanics and Biomedical Engineering , volume =

    Further roles of geometry and properties in the mechanics of saccular aneurysms , author =. Computer Methods in Biomechanics and Biomedical Engineering , volume =

  78. [86]

    Annual Review of Biomedical Engineering , volume =

    Intracranial and abdominal aortic aneurysms: similarities, differences, and need for a new class of computational models , author =. Annual Review of Biomedical Engineering , volume =

  79. [87]

    and others , journal =

    Kamensky, D. and others , journal =. Immersogeometric cardiovascular fluid--structure interaction analysis with divergence-conforming

  80. [88]

    Computer Methods in Applied Mechanics and Engineering , volume =

    Patient-specific isogeometric structural analysis of aortic valve closure , author =. Computer Methods in Applied Mechanics and Engineering , volume =

  81. [89]

    Annals of Biomedical Engineering , volume =

    Physical factors effecting cerebral aneurysm pathophysiology , author =. Annals of Biomedical Engineering , volume =

  82. [90]

    Advanced Materials , volume =

    Tissue engineering with mechanically induced solid-fluid transitions , author =. Advanced Materials , volume =

  83. [91]

    World Neurosurgery , volume =

    Hemodynamic features of rupture-prone blebs in ruptured intracranial aneurysms: A computational fluid dynamics study , author =. World Neurosurgery , volume =

  84. [92]

    Computational Mechanics , volume =

    Fluid--structure interaction modeling of a patient-specific cerebral aneurysm: influence of structural modeling , author =. Computational Mechanics , volume =

  85. [93]

    and others , journal =

    Valen-Sendstad, K. and others , journal =. Real-world variability in the prediction of intracranial aneurysm wall shear stress: The 2015 international aneurysm

    2015

  86. [94]

    Arteriosclerosis, Thrombosis, and Vascular Biology , volume =

    Biomechanical determinants of abdominal aortic aneurysm rupture , author =. Arteriosclerosis, Thrombosis, and Vascular Biology , volume =

  87. [95]

    Biomechanics and Mechanobiology of Aneurysms , year =

    Modelling cerebral aneurysm evolution , author =. Biomechanics and Mechanobiology of Aneurysms , year =

  88. [96]

    Three Dimensional Finite-Element Modeling of Blood Flow in Elastic Vessels: Effects of Arterial Geometry and Elasticity on Aneurysm Growth and Rupture , author =

  89. [97]

    Topology Optimization: Theory, Methods, and Applications , author =

  90. [98]

    Benson, D. J. and others , journal =. Isogeometric shell analysis: The

  91. [99]

    Boo, S. H. and Ryu, H. and Kim, S. H. and Kim, J. , journal =. An isogeometric finite element implementation of visco-hyperelastic

  92. [100]

    Computer Methods in Applied Mechanics and Engineering , volume =

    The variational multiscale method---a paradigm for computational mechanics , author =. Computer Methods in Applied Mechanics and Engineering , volume =

  93. [101]

    Hughes, T. J. and Cottrell, J. A. and Bazilevs, Y. , journal =. Isogeometric analysis:

  94. [102]

    Introduction to Shape Optimization , author =

Pith tools

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