Pith. sign in

REVIEW 4 major objections 3 minor 1 cited by

Variational volume reconstruction with the Deep Ritz Method

T0 review · 4 major / 3 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read The paper claims that sparse, noisy slice data can be turned into a high-quality 3D volume in seconds by minimizing a neural-network phase-field energy, with no segmentation step.

desk verdict A plausible Deep Ritz + Cahn-Hilliard pipeline for segmentation-free slice-to-volume reconstruction, but the central empirical claim is unverifiable in the corrupted text supplied. read the letter →

arxiv 2508.08309 v1 pith:H2HOA53Q submitted 2025-08-08 eess.IV cs.CVcs.LG

Conor Rowan , Sumedh Soman , John A. Evans This is my paper
classification eess.IVcs.CVcs.LG
keywords DeepRitzmethodslice-to-volumereconstructionCahn-HilliardenergyphasefieldMonteCarlointegrationneuralnetworkimagesegmentationbiomedicalimaging
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 tries to show that volume reconstruction from sparse, noisy slice data can be recast as a variational problem and solved directly by a neural network, without first segmenting the slices into object boundaries. The authors propose a Deep Ritz-style approach in which the reconstructed volume is represented as a phase field parameterized by a neural network, and the objective combines a regression term that fits raw grayscale slice intensities with a modified Cahn-Hilliard energy that uses anisotropic diffusion to regularize the shape. They solve the resulting objective by Monte Carlo integration and ADAM. They do not claim this stochastic procedure converges to the true minimizer; instead, they argue empirically that it reliably yields high-quality volumes in seconds even when slices are sparse and noisy. If true, this would make MRI-based slice-to-volume reconstruction practical and remove the segmentation step that many current pipelines depend on.

What carries the argument

The central object is a neural-network-parameterized phase field $\varphi$ that represents the reconstructed volume: regions where $\varphi$ is near one constant are object, near the other are background, and the transition layer marks the boundary. The Deep Ritz construction encodes the variational problem directly in the network loss: the regression term forces the phase field to match the observed slice intensities, and a modified Cahn-Hilliard energy with anisotropic diffusion penalizes interface area in an orientation-dependent way, so it prefers smooth, anatomically plausible boundaries. Monte Carlo integration makes each gradient step cheap without a mesh, and ADAM drives the network

What would settle it

Take a digitally sampled phantom with a known ground-truth volume, generate sparse noisy slices across a range of spacings and noise levels, reconstruct, and measure Dice score or surface distance against the truth. If quality degrades sharply as slice count drops below a modest threshold—or if rerunning with different Monte Carlo seeds gives materially different volumes—the claimed reliability is bounded, and the central practical claim is falsified outside that regime.

Watch

Extended reading notes

Core claim

The central claim is that the standard slice-to-volume reconstruction pipeline—extract boundaries from each slice by segmentation, then fit a surface or mesh—can be bypassed entirely. Instead, one minimizes a single variational objective over a phase field $u$ that represents the volume, with a data-fidelity term that compares the network's prediction to the noisy grayscale slice intensities directly and a regularizer given by a modified Cahn-Hilliard energy with anisotropic diffusion. The phase field is discretized by a neural network, so no mesh is needed, and the integrals in the objective are approximated at every optimization step by Monte Carlo sampling; ADAM then updates the network w

Load-bearing premise

The whole method rests on the empirical assumption that the minimum ADAM finds for the Monte-Carlo-approximated objective is close enough to the true variational solution to be anatomically accurate, even though the paper concedes this need not hold.

Editorial extensions

If this is right

  • Slice-to-volume reconstruction no longer needs a segmentation step: raw grayscale slice intensities enter the objective directly, so boundary extraction errors cannot propagate.
  • Because the volume is a neural network rather than a mesh or voxel grid, reconstruction cost is dominated by the optimizer, and the authors report runtimes on the order of seconds.
  • The method is designed to cope with both few slice planes and noisy acquisitions, so it targets low-signal, sparse clinical protocols.
  • The modified Cahn-Hilliard regularizer with anisotropic diffusion gives control over the smoothness and orientation bias of the reconstructed geometry, which can be tuned per anatomy.

Reading between the lines

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

  • An untested extension suggested by the construction is to make the diffusion tensor depend on local image structure, which could help reconstruct layered or fibrous tissues such as white-matter tracts or articular cartilage.
  • Because the reliability claim is empirical, a stress test varying slice orientation, slice thickness, and noise level against a known phantom would reveal where the method breaks down and how much of the success depends on the energy prior.
  • The same objective could be applied to time-series slice data, effectively reconstructing volumes frame-by-frame, since the neural representation is differentiable and optimizable at any set of measurement planes.
  • If the Monte Carlo approximation error is the main limitation, variance-reduction strategies or quasi-Monte Carlo sampling could shrink the gap between the approximated and true variational solutions without changing the method's structure.
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 / 3 minor

Summary. The paper proposes a neural-network-based variational method for slice-to-volume reconstruction (SVR). It combines a regression loss acting directly on noisy slice data with a modified Cahn-Hilliard energy containing anisotropic diffusion as a regularizer. The phase field is discretized by a neural network, the objective is approximated at each optimization step by Monte Carlo integration, and ADAM is used to minimize this approximation. The abstract claims that, despite the stochastic integration not necessarily yielding the true variational solution, the method reliably produces high-quality reconstructed volumes in seconds even under sparse and noisy slice data. The supplied full text is corrupted mojibake; no equations, experiments, implementation details, or quantitative results are readable.

Significance. If the empirical claim were supported, the method would be a practically relevant, segmentation-free alternative to conventional SVR pipelines, potentially removing the need for boundary extraction and mesh generation and offering rapid reconstructions. The application of the Deep Ritz method to volumetric phase-field reconstruction is conceptually interesting. However, because the demonstration is not accessible in the current manuscript, no conclusion about the validity or significance of the contribution can be drawn from the version under review.

major comments (4)
  1. [Full text (as provided)] The central claim is explicitly empirical: 'we demonstrate that our method reliably produces high-quality reconstructed volumes' (Abstract). No quantitative evaluation — datasets, baselines, metrics, error bars, runtime measurements, or visual comparisons — is readable in the full text, which consists entirely of corrupted mojibake characters. The claimed demonstration therefore cannot be inspected or reproduced. This is load-bearing: the abstract itself concedes the approximation may not give the true variational solution, so the paper's value rests entirely on the empirical demonstration.
  2. [Abstract] The concession that 'the stochastic integration may not yield the true solution to the variational problem' frames the method as a heuristic. The manuscript provides no analysis quantifying the gap between the Monte Carlo/ADAM minimum and the exact variational solution, nor any surrogate empirical evidence that the approximation error is acceptable. Without either a theoretical bound or a bracketing of the approximation error, or experimental validation, the reliability claim is unsupported.
  3. [Full text (as provided)] The variational formulation and its discretization cannot be checked: no readable equations for the regression loss, the modified Cahn-Hilliard energy, the anisotropic diffusion tensor, the Monte Carlo sample-count selection, or the network architecture are present. Free hyperparameters such as the loss weighting, interface width, and anisotropic diffusion coefficients are not stated. This prevents reproducibility and makes it impossible to assess whether the objective is well-posed (e.g., coercivity, regularity) for the proposed phase-field reconstruction.
  4. [Cahn-Hilliard prior (as described in Abstract)] The modified Cahn-Hilliard energy imposes a smoothness and geometry prior on the reconstructed volume. The abstract offers no evidence that this prior is appropriate for anatomical shapes. If the anisotropic diffusion or interface parameters are poorly matched to the tissue geometry, reconstructions will be systematically biased. This requires empirical validation against realistic biological volumes, ideally including comparison with segmentation-based SVR methods.
minor comments (3)
  1. [Full text header] The full text contains a different arXiv identifier ('arXiv:2508.08313v1 [cs.CY]') than the paper's stated ID (arXiv:2508.08309). This needs correction.
  2. [Full text (as provided)] The running title and page headers are garbled. The submitted PDF should be re-encoded so that equations, figures, and references are legible.
  3. [Abstract] The term 'reliably' should be concretely defined: against which ground-truth volumes, which error metric, and under which levels of sparsity and noise? The abstract should indicate where the experimental section will support this claim.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation found; central claim is empirical and unverifiable from the corrupted text, but not circular.

full rationale

The readable portion of the manuscript (the abstract) describes a variational objective combining a regression loss on noisy slice data with a modified Cahn-Hilliard energy, discretized by a neural network and optimized with Monte Carlo integration and ADAM. The reconstruction is produced by minimizing this independently defined objective; nothing in the abstract indicates that a parameter is fit to the target volume and then renamed a prediction, nor that a self-citation is load-bearing, nor that the objective is defined in terms of the desired output. The paper explicitly concedes that the stochastic minimization 'may not yield the true solution to the variational problem,' so the practical reliability claim rests on empirical demonstration rather than on a derived guarantee. This is a limitation and a correctness/verifiability risk—especially because the full-text body is corrupted mojibake in the provided input, making the experiments unreadable—but it is not circularity. Without readable equations, no specific reduction (Eq. X = Eq. Y by construction, or fitted input called prediction) can be identified. The concern that the Cahn-Hilliard prior may impose inappropriate smoothness is likewise an empirical modeling question, not a circular step. Accordingly, the manuscript exhibits no detectable circular derivation, and the score is 0.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central method rests on several unstated choices: the energy/balance hyperparameters, the suitability of Cahn-Hilliard as a geometric prior, and the assumption that approximate stochastic minimization yields useful reconstructions. None are independently verified in the abstract.

free parameters (3)
  • loss weighting hyperparameters (regression vs. Cahn-Hilliard energy)
    Abstract does not specify how the regression loss and regularization energy are balanced; these weights strongly affect reconstruction quality and are presumably tuned.
  • Cahn-Hilliard energy parameters (interface width, anisotropic diffusion coefficients)
    The modified Cahn-Hilliard energy with anisotropic diffusion requires coefficients that set interface thickness and directional smoothing; not reported in abstract.
  • Monte Carlo sample count per step
    Number of samples used to approximate the variational integral is unstated and affects stochastic approximation error.
assumptions (4)
  • domain assumption A minimizer of the Monte Carlo-approximated objective, even when not the true variational solution, still yields an accurate reconstruction.
    The abstract's last sentence acknowledges stochastic integration may not yield the true solution; the reliability claim depends on the approximate minimizer being good enough.
  • domain assumption The modified Cahn-Hilliard energy with anisotropic diffusion is a suitable regularizer for anatomical volumes from slice data.
    The choice of regularization is motivated by geometry but not independently justified in the abstract.
  • domain assumption The Deep Ritz neural-network parameterization can represent the relevant phase-field solutions.
    The method assumes the neural network has sufficient expressivity to approximate the true phase field; no universal approximation argument for this specific energy is given in the abstract.
  • domain assumption Monte Carlo integration and ADAM converge in practice for this high-dimensional objective.
    The abstract does not report convergence guarantees or failure cases, only that the method 'reliably produces' good volumes.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Variational volume reconstruction with the Deep Ritz Method." pith.science (2026). https://pith.science/paper/H2HOA53Q

@misc{pith2026250808309,
  author       = {Pith},
  title        = {Pith review of: Variational volume reconstruction with the Deep Ritz Method},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/H2HOA53Q}},
  note         = {Machine review of arXiv:2508.08309}
}
read the original abstract

We present a novel approach to variational volume reconstruction from sparse, noisy slice data using the Deep Ritz method. Motivated by biomedical imaging applications such as MRI-based slice-to-volume reconstruction (SVR), our approach addresses three key challenges: (i) the reliance on image segmentation to extract boundaries from noisy grayscale slice images, (ii) the need to reconstruct volumes from a limited number of slice planes, and (iii) the computational expense of traditional mesh-based methods. We formulate a variational objective that combines a regression loss designed to avoid image segmentation by operating on noisy slice data directly with a modified Cahn-Hilliard energy incorporating anisotropic diffusion to regularize the reconstructed geometry. We discretize the phase field with a neural network, approximate the objective at each optimization step with Monte Carlo integration, and use ADAM to find the minimum of the approximated variational objective. While the stochastic integration may not yield the true solution to the variational problem, we demonstrate that our method reliably produces high-quality reconstructed volumes in a matter of seconds, even when the slice data is sparse and noisy.

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Learning Gradient Flow: Using Equation Discovery to Accelerate Engineering Optimization

    math.OC 2026-02 conditional novelty 6.0 of 10

    An optimizer that fits a SINDy polynomial model to recent optimization-variable trajectories and then integrates that surrogate flow instead of evaluating the true objective/gradient can cut gradient-evaluation counts...

Reference graph

Works this paper leans on

48 extracted references · 43 canonical work pages · cited by 1 Pith paper

  1. [1]

    A deep learning energy method for hyperelasticity and viscoelasticity

    Diab W. Abueidda, Seid Koric, Rashid Abu Al-Rub, Corey M. Parrott, Kai A. James, and Nahil A. Sobh. A deep learning energy method for hyperelasticity and viscoelasticity. European Journal of Mechanics - A/Solids , 95:104639, September 2022. arXiv:2201.08690 [cs]

    work page Pith review arXiv 2022

  2. [2]

    Arigovindan, M

    M. Arigovindan, M. Suhling, P. Hunziker, and M. Unser. Variational image reconstruction from arbitrarily spaced samples: a fast multiresolution spline solution. IEEE Transactions on Image Processing , 14(4):450--460, April 2005

    work page 2005

  3. [3]

    Three-dimensional reconstruction of complex shapes based on the Delaunay triangulation

    Jean-Daniel Boissonnat and Bernhard Geiger. Three-dimensional reconstruction of complex shapes based on the Delaunay triangulation. pages 964--975, San Jose, CA, 1993

    work page 1993

  4. [4]

    Volume Reconstruction from Slices

    Elie Bretin, François Dayrens, and Simon Masnou. Volume Reconstruction from Slices . SIAM Journal on Imaging Sciences , 10(4):2326--2358, January 2017

    work page 2017

  5. [5]

    Learning phase field mean curvature flows with neural networks

    Elie Bretin, Roland Denis, Simon Masnou, and Garry Terii. Learning phase field mean curvature flows with neural networks. Journal of Computational Physics , 470:111579, December 2022. arXiv:2112.07343 [math]

    work page Pith review arXiv 2022

  6. [6]

    A penalized Allen - Cahn equation for the mean curvature flow of thin structures, May 2024

    Elie Bretin, Chih-Kang Huang, and Simon Masnou. A penalized Allen - Cahn equation for the mean curvature flow of thin structures, May 2024. arXiv:2310.10272 [math]

    arXiv 2024

  7. [7]

    Variational image segmentation model coupled with image restoration achievements

    Xiaohao Cai. Variational image segmentation model coupled with image restoration achievements. Pattern Recognition , 48(6):2029--2042, June 2015

    work page

  8. [8]

    J. C. Carr, R. K. Beatson, J. B. Cherrie, T. J. Mitchell, W. R. Fright, B. C. McCallum, and T. R. Evans. Reconstruction and representation of 3D objects with radial basis functions. In Proceedings of the 28th annual conference on Computer graphics and interactive techniques , pages 67--76. ACM, August 2001

    work page 2001

Show all 48 references
  1. [9]

    Introduction to variational image-processing models and applications

    Ke Chen. Introduction to variational image-processing models and applications. International Journal of Computer Mathematics , 90(1):1--8, January 2013. Publisher: Taylor & Francis \_eprint: https://doi.org/10.1080/00207160.2012.757073

    2013

  2. [10]

    Existence of equilibria for the cahn-hilliard equation via local minimizers of the perimeter

    Xinfu Chen, , and Michal Kowalczyk. Existence of equilibria for the cahn-hilliard equation via local minimizers of the perimeter. Communications in Partial Differential Equations , 21(7-8):1207--1233, January 1996. Publisher: Taylor & Francis \_eprint: https://doi.org/10.1080/...

    1996 doi

  3. [11]

    Surface Reconstruction from Scattered Point via RBF Interpolation on GPU , May 2013

    Salvatore Cuomo, Ardelio Gallettiy, Giulio Giuntay, and Alfredo Staracey. Surface Reconstruction from Scattered Point via RBF Interpolation on GPU , May 2013. arXiv:1305.5179 [cs]

    2013 arXiv

  4. [12]

    Surface Reconstruction from Sparse and Mutually Intersected Contours for Freehand 3D Ultrasound Using Variational Method

    Shuangcheng Deng, Yunhua Li, Lipei Jiang, Yingyu Cao, and Junwen Zhang. Surface Reconstruction from Sparse and Mutually Intersected Contours for Freehand 3D Ultrasound Using Variational Method . In 2012 Second International Conference on Intelligent System Design and Engineeri...

    2012

  5. [13]

    Image Super - Resolution Using Deep Convolutional Networks , July 2015

    Chao Dong, Chen Change Loy, Kaiming He, and Xiaoou Tang. Image Super - Resolution Using Deep Convolutional Networks , July 2015. arXiv:1501.00092 [cs]

    2015 arXiv

  6. [14]

    A level set formulation for Willmore flow

    Marc Droske and Martin Rumpf. A level set formulation for Willmore flow. Interfaces and Free Boundaries, Mathematical Analysis, Computation and Applications , 6(3):361--378, September 2004

    2004

  7. [15]

    Super-resolution reconstruction of single anisotropic 3D MR images using residual convolutional neural network

    Jinglong Du, Zhongshi He, Lulu Wang, Ali Gholipour, Zexun Zhou, Dingding Chen, and Yuanyuan Jia. Super-resolution reconstruction of single anisotropic 3D MR images using residual convolutional neural network. Neurocomputing , 392:209--220, June 2020

    2020

  8. [16]

    The Deep Ritz method: A deep learning-based numerical algorithm for solving variational problems, September 2017

    Weinan E and Bing Yu. The Deep Ritz method: A deep learning-based numerical algorithm for solving variational problems, September 2017. arXiv:1710.00211 [cs]

    2017 arXiv

  9. [17]

    Miller, Harini Veeraraghavan, Bernd Freisleben, Alexandra J

    Jan Egger, Tina Kapur, Andriy Fedorov, Steve Pieper, James V. Miller, Harini Veeraraghavan, Bernd Freisleben, Alexandra J. Golby, Christopher Nimsky, and Ron Kikinis. GBM Volumetry using the 3D Slicer Medical Image Computing Platform . Scientific Reports , 3(1):1364, March 201...

    2013

  10. [18]

    Colliding Interfaces in Old and New Diffuse -interface Approximations of Willmore -flow, February 2013

    Selim Esedoglu, Andreas Rätz, and Matthias Röger. Colliding Interfaces in Old and New Diffuse -interface Approximations of Willmore -flow, February 2013. arXiv:1209.6531 [math]

    2013 arXiv

  11. [19]

    Daniela Famiani, Carlo Alberto Brunori, Luca Pizzimenti, Fabrizio Cara, Marco Caciagli, Laura Melelli, Francesco Mirabella, and Massimiliano R. Barchi. Geophysical reconstruction of buried geological features and site effects estimation of the Middle Valle Umbra basin (central...

    2020

  12. [20]

    Jimack, and René de Borst

    Yousef Ghaffari Motlagh, Peter K. Jimack, and René de Borst. Deep learning phase-field model for brittle fractures. International Journal for Numerical Methods in Engineering , 124(3):620--638, 2023

    2023

  13. [21]

    Giannakopoulos, Matthew J

    Ilias I. Giannakopoulos, Matthew J. Muckley, Jesi Kim, Matthew Breen, Patricia M. Johnson, Yvonne W. Lui, and Riccardo Lattanzi. Accelerated MRI reconstructions via variational network and feature domain learning. Scientific Reports , 14(1):10991, May 2024. Publisher: Nature P...

    2024

  14. [22]

    A simple shape transformation method based on phase-field model

    Ziwei Han, Heming Xu, and Jian Wang. A simple shape transformation method based on phase-field model. Computers & Mathematics with Applications , 147:121--129, October 2023

    2023

  15. [23]

    A Variational Approach for Volume -to- Slice Registration

    Stefan Heldmann and Nils Papenberg. A Variational Approach for Volume -to- Slice Registration . In Xue-Cheng Tai, Knut Mørken, Marius Lysaker, and Knut-Andreas Lie, editors, Scale Space and Variational Methods in Computer Vision , pages 624--635, Berlin, Heidelberg, 2009. Springer

    2009

  16. [24]

    B. E. Hirsch, K. D. Thoben, X. Sheng, and M. Joppe. Reconstruction of volume boundary from layers by triangulation. Journal of Manufacturing Systems , 16(4):249--259, January 1997

    1997

  17. [25]

    Osher, and R

    Hong-Kai Zhao , S. Osher, and R. Fedkiw. Fast surface reconstruction using the level set method. In Proceedings IEEE Workshop on Variational and Level Set Methods in Computer Vision , pages 194--201, Vancouver, BC, Canada, 2001. IEEE Computer Soc

    2001

  18. [26]

    C1-continuous Terrain Reconstruction from Sparse Contours

    Kai Hormann, Salvatore Spinello, and Peter Schroder. C1-continuous Terrain Reconstruction from Sparse Contours . Proceedings of the Vision, Modeling, and Visualization Conference , 2003

    2003

  19. [27]

    Three-dimensional volume reconstruction from multi-slice data using a shape transformation

    Hyundong Kim, Chaeyoung Lee, Soobin Kwak, Youngjin Hwang, Sangkwon Kim, Yongho Choi, and Junseok Kim. Three-dimensional volume reconstruction from multi-slice data using a shape transformation. Computers & Mathematics with Applications , 113:52--58, May 2022

    2022

  20. [28]

    Accurate Surface Reconstruction in 3D Using Two -dimensional Parallel Cross Sections

    Sang-Un Kim and Chang-Ock Lee. Accurate Surface Reconstruction in 3D Using Two -dimensional Parallel Cross Sections . Journal of Mathematical Imaging and Vision , 53:182--195, October 2015. ADS Bibcode: 2015JMIV...53..182K

    2015

  21. [29]

    An efficient volume repairing method by using a modified Allen - Cahn equation

    Yibao Li, Shouren Lan, Xin Liu, Bingheng Lu, and Lisheng Wang. An efficient volume repairing method by using a modified Allen - Cahn equation. Pattern Recognition , 107:107478, November 2020

    2020

  22. [30]

    Three-dimensional volume reconstruction from slice data using phase-field models

    Yibao Li, Jaemin Shin, Yongho Choi, and Junseok Kim. Three-dimensional volume reconstruction from slice data using phase-field models. Computer Vision and Image Understanding , 137:115--124, August 2015

    2015

  23. [31]

    Weighted 3D volume reconstruction from series of slice data using a modified Allen – Cahn equation

    Yibao Li, Xin Song, Soobin Kwak, and Junseok Kim. Weighted 3D volume reconstruction from series of slice data using a modified Allen – Cahn equation. Pattern Recognition , 132:108914, December 2022

    2022

  24. [32]

    Multicomponent volume reconstruction from slice data using a modified multicomponent Cahn – Hilliard system

    Yibao Li, Jing Wang, Bingheng Lu, Darae Jeong, and Junseok Kim. Multicomponent volume reconstruction from slice data using a modified multicomponent Cahn – Hilliard system. Pattern Recognition , 93:124--133, September 2019

    2019

  25. [33]

    Measuring three-dimensional tibiofemoral kinematics using dual-slice real-time magnetic resonance imaging

    Cheng-Chung Lin, Shuo Zhang, Chao-Yu Hsu, Jens Frahm, Tung-Wu Lu, and Ting-Fang Shih. Measuring three-dimensional tibiofemoral kinematics using dual-slice real-time magnetic resonance imaging. Medical Physics , 46(10):4588--4599, October 2019

    2019

  26. [34]

    Higher-order multi-scale deep Ritz method for multi-scale problems of authentic composite materials, August 2023

    Jiale Linghu, Hao Dong, Junzhi Cui, and Yufeng Nie. Higher-order multi-scale deep Ritz method for multi-scale problems of authentic composite materials, August 2023

    2023

  27. [35]

    Deep Ritz method with adaptive quadrature for linear elasticity

    Min Liu, Zhiqiang Cai, and Karthik Ramani. Deep Ritz method with adaptive quadrature for linear elasticity. Computer Methods in Applied Mechanics and Engineering , 415:116229, October 2023

    2023

  28. [36]

    Manav, R

    M. Manav, R. Molinaro, S. Mishra, and L. De Lorenzis. Phase-field modeling of fracture with physics-informed deep learning. Computer Methods in Applied Mechanics and Engineering , 429:117104, September 2024

    2024

  29. [37]

    Srinivasan, Matthew Tancik, Jonathan T

    Ben Mildenhall, Pratul P. Srinivasan, Matthew Tancik, Jonathan T. Barron, Ravi Ramamoorthi, and Ren Ng. NeRF : Representing Scenes as Neural Radiance Fields for View Synthesis , August 2020. arXiv:2003.08934 [cs]

    2020 arXiv

  30. [38]

    Real time 3D reconstruction for enhanced cybersecurity of additive manufacturing processes

    Ankush Kumar Mishra, Shi Yong Goh, Baskar Ganapathysubramanian, and Adarsh Krishnamurthy. Real time 3D reconstruction for enhanced cybersecurity of additive manufacturing processes. Journal of Manufacturing Processes , 145:274--285, July 2025

    2025

  31. [39]

    Interpolating Implicit Surfaces From Scattered Surface Data Using Compactly Supported Radial Basis Functions

    Bryan Morse, David Chen, Penny Rheingans, Kalpathi Subramanian, and Terry Yoo. Interpolating Implicit Surfaces From Scattered Surface Data Using Compactly Supported Radial Basis Functions . Faculty Publications , May 2001

    2001

  32. [40]

    Brain MRI super-resolution using deep 3D convolutional networks

    Chi-Hieu Pham, Aurélien Ducournau, Ronan Fablet, and François Rousseau. Brain MRI super-resolution using deep 3D convolutional networks. In 2017 IEEE 14th International Symposium on Biomedical Imaging ( ISBI 2017) , pages 197--200, April 2017. ISSN: 1945-8452

    2017

  33. [41]

    Hajnal, Anthony Price, and Daniel Rueckert

    Jo Schlemper, Jose Caballero, Joseph V. Hajnal, Anthony Price, and Daniel Rueckert. A Deep Cascade of Convolutional Neural Networks for MR Image Reconstruction , March 2017. arXiv:1703.00555 [cs]

    2017 arXiv

  34. [42]

    Variational methods with application to medical image segmentation: A survey

    Xiu Shu, Zhihui Li, Xiaojun Chang, and Di Yuan. Variational methods with application to medical image segmentation: A survey. Neurocomputing , 639:130260, July 2025

    2025

  35. [43]

    Variational Methods for Image Segmentation

    Jack A Spencer and Ke Chen. Variational Methods for Image Segmentation . Thesis , 2016

    2016

  36. [44]

    Sukumar and Ankit Srivastava

    N. Sukumar and Ankit Srivastava. Exact imposition of boundary conditions with distance functions in physics-informed deep neural networks. Computer Methods in Applied Mechanics and Engineering , 389:114333, February 2022

    2022

  37. [45]

    Deep variational network for rapid 4D flow MRI reconstruction

    Valery Vishnevskiy, Jonas Walheim, and Sebastian Kozerke. Deep variational network for rapid 4D flow MRI reconstruction. Nature Machine Intelligence , 2(4):228--235, April 2020. arXiv:2004.09610 [eess]

    2020 arXiv

  38. [46]

    Young, Yaël Balbastre, Bruce Fischl, Polina Golland, and Juan Eugenio Iglesias

    Sean I. Young, Yaël Balbastre, Bruce Fischl, Polina Golland, and Juan Eugenio Iglesias. Fully Convolutional Slice -to- Volume Reconstruction for Single - Stack MRI . In 2024 IEEE / CVF Conference on Computer Vision and Pattern Recognition ( CVPR ) , pages 11535--11545, Seattle...

    2024

  39. [47]

    Super- Resolution Surface Reconstruction from Few Low - Resolution Slices

    Yiyao Zhang, Ke Chen, and Shang-Hua Yang. Super- Resolution Surface Reconstruction from Few Low - Resolution Slices . Inverse Problems and Imaging , 18(2):447--479, 2024. arXiv:2309.05071 [math]

    2024 arXiv

  40. [48]

    Image Segmentation Using Euler ’s Elastica as the Regularization

    Wei Zhu, Xue-Cheng Tai, and Tony Chan. Image Segmentation Using Euler ’s Elastica as the Regularization . Journal of Scientific Computing , 57(2):414--438, November 2013

    2013

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.