pith. sign in

arxiv: 2404.06991 · v2 · submitted 2024-04-10 · 📡 eess.IV · cs.CV

Ray-driven Spectral CT Reconstruction Based on Neural Base-Material Fields

Ligen Shi , Ping Yang , Chang Liu , Wei Zhang , Xing Zhao , Jun Qiu This is my paper

Pith reviewed 2026-05-24 02:12 UTC · model grok-4.3

classification 📡 eess.IV cs.CV
keywords spectral CTneural fieldsbasis material decompositionray-driven discretizationimplicit functionsauto-differentiationtomographic reconstruction
0
0 comments X

The pith

Neural fields represent basis materials as continuous functions to enable resolution-independent spectral CT reconstruction.

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

The paper proposes a parameterization of object attenuation coefficients via neural fields to sidestep the computation of pixel-driven projection matrices when discretizing line integrals in spectral CT. Basis materials appear as continuous vector-valued implicit functions whose values are recovered through auto-differentiation inside a ray-driven integral scheme. This formulation is asserted to remain accurate regardless of the output image grid size and to produce compact, regular networks. Experiments are presented as evidence that the resulting reconstructions handle the ill-posed nonlinear system effectively and support high-resolution output.

Core claim

The paper claims that representing the basis materials as continuous vector-valued implicit functions inside a neural field model, paired with a lightweight ray-driven discretization of the line integrals, permits direct solution of the large-scale nonlinear integral equations of spectral CT via auto-differentiation, yielding reconstructions that are not constrained by the spatial resolution of the image grid.

What carries the argument

Neural base-material fields: continuous vector-valued implicit functions that stand for the attenuation coefficients and supply the line integrals through automatic differentiation without forming explicit projection matrices.

If this is right

  • Line-integral discretization gains accuracy while avoiding explicit pixel-driven matrices.
  • Reconstruction quality remains independent of the spatial resolution chosen for the output image.
  • The trained network stays compact and exhibits regular properties.
  • High-resolution images can be produced on demand without additional discretization steps.

Where Pith is reading between the lines

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

  • The same continuous representation could be tested on other line-integral inverse problems such as limited-angle or cone-beam CT.
  • Resolution independence may permit on-the-fly detail adjustment in clinical workflows without retraining the model.
  • Coupling the field with differentiable forward models from other modalities could extend the approach to joint multi-energy and structural reconstruction.

Load-bearing premise

The neural field plus ray-driven discretization can approximate the line integrals and invert the ill-posed nonlinear system without introducing instabilities or needing adjustments that tie accuracy to a particular image resolution.

What would settle it

A test that reconstructs the same phantom at successively higher resolutions and shows either growing artifacts or the need for network retraining scaled to the new resolution would falsify the resolution-independence claim.

Figures

Figures reproduced from arXiv: 2404.06991 by Chang Liu, Jun Qiu, Ligen Shi, Ping Yang, Wei Zhang, Xing Zhao.

Figure 1
Figure 1. Figure 1: Framework of polychromatic projection for dual-material decomposition using NeMFs. Given sampling point coordinates x = (x, y, z) along a ray L(t), the network predicts material densities ˆf = (fˆ1, fˆ2). These predictions are integrated to compute the polychromatic projection pˆk,L under the kth spectrum via (4). The loss between pˆk,L and the measured projection pk is used for backpropagation. provides a… view at source ↗
Figure 3
Figure 3. Figure 3: (a) Slice of the thorax phantom used in the numerical simulation; (b) X-ray spectra used in the numerical simulation. Scanning configuration. The distance between the X-ray source and the rotation center (Turntable axis) is SOD = 1000 mm, and the distance between the source and detector is SDD = 1536 mm. The linear detector consists of 512 ele￾ments of size 0.8 mm. Polychromatic projections are generated u… view at source ↗
Figure 2
Figure 2. Figure 2: The architecture diagram illustrates the fan-beam NeMFs. Hid￾den layers, depicted in blue, indicate the vector dimensions within each block; blue arrows denote connections to ReLU activation functions. A ReLU activation is applied at the output layer to enforce non-negativity of material densities. A. Dual-Spectral Dual-Material Experiments This section uses dual-spectrum X-ray scanning to recon￾struct dua… view at source ↗
Figure 5
Figure 5. Figure 5: Profiles at line 290: Water material density image on noise-free data. TABLE I QUANTITATIVE COMPARISON ON NOISE-FREE DATA. Material IFBP E-ART dNCPD Ours PSNR SSIM PSNR SSIM PSNR SSIM PSNR SSIM Bone 43.99 0.856 35.40 0.999 52.97 0.999 45.43 0.999 Water 38.93 0.800 31.97 0.975 38.68 0.992 39.47 0.998 algorithms using sparse angle data. Observation of [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 4
Figure 4. Figure 4: Comparison of noise-free basis-material reconstructions ob￾tained using IFBP, E-ART, dNCPD, and the proposed method. For each basis material, the reconstructed image and two local magnified views are shown. For each image in the first row, the lower-left inset (yellow dashed box) presents the residual image with a display window of [0,0.2]. 2) Comparison Experiment with Sparse Angle Data: In this experimen… view at source ↗
Figure 7
Figure 7. Figure 7: Profiles at line 290: Water material density image on sparse angle data. TABLE II QUANTITATIVE COMPARISON ON SPARSE-ANGLE DATA. Material IFBP E-ART dNCPD Ours PSNR SSIM PSNR SSIM PSNR SSIM PSNR SSIM Bone 32.79 0.462 40.55 0.998 44.27 0.998 40.22 0.995 Water 28.50 0.398 28.24 0.916 32.27 0.939 34.63 0.991 3) Comparison Experiment with Noise and Geometric Incon￾sistency data: This subsection evaluates the ro… view at source ↗
Figure 9
Figure 9. Figure 9: Profiles at line 290: Water material density image with noise and geometric inconsistencies. TABLE III QUANTITATIVE COMPARISON ON NOISY AND GEOMETRICALLY INCONSISTENT DATA. Material IFBP E-ART dNCPD Ours PSNR SSIM PSNR SSIM PSNR SSIM PSNR SSIM Bone 35.72 0.782 35.26 0.998 43.82 0.998 38.13 0.998 Water 27.51 0.418 30.87 0.936 30.60 0.954 33.87 0.993 the simulated data [PITH_FULL_IMAGE:figures/full_fig_p008… view at source ↗
Figure 10
Figure 10. Figure 10: Comparison of basis material density image reconstructed by different methods. TABLE V ABLATION STUDY: QUANTITATIVE COMPARISON WITH AND WITHOUT MER. Setting Material w/o MER w/ MER PSNR SSIM PSNR SSIM Noise-free Bone 44.99 0.999 45.43 0.999 Water 39.11 0.998 39.47 0.998 Sparse-angle Bone 39.44 0.991 40.22 0.995 Water 33.04 0.987 34.63 0.991 Noise + Geo. Inc. Bone 37.41 0.997 38.13 0.998 Water 33.14 0.991 … view at source ↗
read the original abstract

In spectral CT reconstruction, the basis materials decomposition involves solving a large-scale nonlinear system of integral equations, which is highly ill-posed mathematically. This paper proposes a model that parameterizes the attenuation coefficients of the object using a neural field representation, thereby avoiding the complex calculations of pixel-driven projection coefficient matrices during the discretization process of line integrals. It introduces a lightweight discretization method for line integrals based on a ray-driven neural field, enhancing the accuracy of the integral approximation during the discretization process. The basis materials are represented as continuous vector-valued implicit functions to establish a neural field parameterization model for the basis materials. The auto-differentiation framework of deep learning is then used to solve the implicit continuous function of the neural base-material fields. This method is not limited by the spatial resolution of reconstructed images, and the network has compact and regular properties. Experimental validation shows that our method performs exceptionally well in addressing the spectral CT reconstruction. Additionally, it fulfils the requirements for the generation of high-resolution reconstruction images.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The manuscript proposes a neural field parameterization of basis-material attenuation coefficients for spectral CT reconstruction. Basis materials are modeled as continuous vector-valued implicit functions; line integrals are discretized via a ray-driven approach that avoids explicit pixel-driven projection matrices, and the resulting nonlinear system is solved using auto-differentiation. The authors claim the method is independent of reconstructed-image spatial resolution, possesses compact and regular network properties, and yields superior experimental performance for high-resolution spectral CT.

Significance. If the stability and accuracy claims hold, the work would supply a resolution-independent parameterization for the ill-posed material-decomposition problem in spectral CT, replacing discrete matrix constructions with a differentiable implicit representation. The combination of neural fields with ray-driven integration is a potentially useful technical direction for continuous-domain inverse problems.

major comments (2)
  1. [Abstract / model introduction paragraph] Abstract, paragraph on model introduction: the central claim that the method 'is not limited by the spatial resolution of reconstructed images' rests on the unexamined assumption that the ray-driven neural integral approximation remains stable and accurate for the nonlinear decomposition; no analysis is provided of conditioning, network-capacity limits, or whether auto-differentiation introduces new instabilities that would re-introduce effective resolution dependence.
  2. [Abstract] Abstract, experimental validation sentence: the statement that the method 'performs exceptionally well' is unsupported by any reported quantitative metrics, error analysis, or comparison against established baselines in the provided text; without these, the performance claim cannot be assessed.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thoughtful and constructive comments. We address each major comment below and outline planned revisions to strengthen the manuscript.

read point-by-point responses

  1. Referee: [Abstract / model introduction paragraph] Abstract, paragraph on model introduction: the central claim that the method 'is not limited by the spatial resolution of reconstructed images' rests on the unexamined assumption that the ray-driven neural integral approximation remains stable and accurate for the nonlinear decomposition; no analysis is provided of conditioning, network-capacity limits, or whether auto-differentiation introduces new instabilities that would re-introduce effective resolution dependence.

    Authors: The resolution-independence claim follows directly from the continuous implicit representation: the neural base-material field defines attenuation as a vector-valued function over continuous space rather than a discrete voxel grid, and the ray-driven integrator evaluates line integrals via quadrature without constructing a resolution-dependent projection matrix. This formulation is in principle independent of any chosen reconstruction grid. We acknowledge, however, that the manuscript provides no explicit analysis of numerical conditioning of the resulting nonlinear system, bounds on network capacity needed for accurate high-frequency representation, or potential instabilities introduced by automatic differentiation through the ray integral. In the revision we will add a dedicated subsection discussing these aspects, including a brief conditioning analysis and empirical checks of gradient stability across increasing spatial frequencies. revision: yes

  2. Referee: [Abstract] Abstract, experimental validation sentence: the statement that the method 'performs exceptionally well' is unsupported by any reported quantitative metrics, error analysis, or comparison against established baselines in the provided text; without these, the performance claim cannot be assessed.

    Authors: The abstract is a concise summary; the full manuscript contains quantitative results (RMSE, SSIM, material-decomposition error tables) together with comparisons against pixel-driven matrix methods and other neural baselines. Nevertheless, the referee is correct that the abstract itself offers no numerical support. We will revise the abstract to include one or two key quantitative metrics and a brief statement of the baselines used, while preserving brevity. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation relies on independent neural field parameterization

full rationale

The paper introduces a neural field parameterization for basis materials in spectral CT, using continuous implicit functions and ray-driven discretization solved via auto-differentiation. No load-bearing steps reduce claimed results (e.g., resolution independence or reconstruction accuracy) to quantities defined by the method itself. The approach is presented as an external parameterization with experimental validation, without self-definitional equations, fitted inputs renamed as predictions, or self-citation chains that force the central claims. The derivation chain remains self-contained against the stated assumptions.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

Only the abstract is available, so the ledger reflects components explicitly named there. The central claim rests on the expressivity of neural fields and the accuracy of the ray-driven integral approximation.

axioms (1)
  • domain assumption Neural networks can represent continuous functions sufficiently accurately for attenuation coefficient modeling
    Implicit in the choice of neural field parameterization for basis materials.
invented entities (1)
  • neural base-material fields no independent evidence
    purpose: Continuous vector-valued implicit functions that parameterize basis material attenuation coefficients
    New representation introduced to avoid pixel-driven matrices.

pith-pipeline@v0.9.0 · 5708 in / 1260 out tokens · 24482 ms · 2026-05-24T02:12:52.797262+00:00 · methodology

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

50 extracted references · 50 canonical work pages

  1. [1]

    Evaluation of a prototype dual-energy computed tomographic apparatus. i. phantom studies

    W. A. Kalender, W. H. Perman, J. R. Vetter, and E. Klotz, “Evaluation of a prototype dual-energy computed tomographic apparatus. i. phantom studies.”Medical Physics, vol. 13, pp. 334–339, 1986

    work page 1986

  2. [2]

    Iterative image-domain decomposition for dual-energy ct,

    T. Niu, X. Dong, M. Petrongolo, and L. Zhu, “Iterative image-domain decomposition for dual-energy ct,”Medical physics, vol. 41, no. 4, p. 041901, 2014

    work page 2014

  3. [3]

    Nerf: Representing scenes as neural radiance fields for view synthesis,

    B. Mildenhall, P. P. Srinivasan, M. Tancik, J. T. Barron, R. Ramamoorthi, and R. Ng, “Nerf: Representing scenes as neural radiance fields for view synthesis,”Communications of the ACM, vol. 65, no. 1, pp. 99–106, 2021

    work page 2021

  4. [4]

    Beam hardening in x-ray recon- structive tomography,

    R. A. BROOKS and G. D. CHIRO, “Beam hardening in x-ray recon- structive tomography,”Physics in medicine & biology, vol. 21, no. 3, p. 390, 1976

    work page 1976

  5. [5]

    Statistical image-domain multimaterial decomposition for dual-energy ct,

    Y . Xue, R. Ruan, X. Hu, Y . Kuang, J. Wang, Y . Long, and T. Niu, “Statistical image-domain multimaterial decomposition for dual-energy ct,”Medical physics, vol. 44, no. 3, pp. 886–901, 2017

    work page 2017

  6. [6]

    A weighted polynomial based material decomposition method for spectral x-ray ct imaging,

    D. Wu, L. Zhang, X. Zhu, X. Xu, and S. Wang, “A weighted polynomial based material decomposition method for spectral x-ray ct imaging,” Physics in Medicine & Biology, vol. 61, no. 10, p. 3749, 2016

    work page 2016

  7. [7]

    Image-based dual energy ct using optimized precorrection functions: A practical new approach of material decomposition in image domain,

    C. Maaß, M. Baer, and M. Kachelrieß, “Image-based dual energy ct using optimized precorrection functions: A practical new approach of material decomposition in image domain,”Medical physics, vol. 36, no. 8, pp. 3818–3829, 2009

    work page 2009

  8. [8]

    A beam-hardening correction using dual- energy computed tomography,

    A. Coleman and M. Sinclair, “A beam-hardening correction using dual- energy computed tomography,”Physics in medicine & biology, vol. 30, no. 11, p. 1251, 1985

    work page 1985

  9. [9]

    Dual energy computed tomography for explosive detection,

    Z. Ying, R. Naidu, and C. R. Crawford, “Dual energy computed tomography for explosive detection,”Journal of X-ray Science and Technology, vol. 14, no. 4, pp. 235–256, 2006

    work page 2006

  10. [10]

    Impact of polychromatic x-ray sources on helical, cone-beam computed tomography and dual-energy methods,

    E. Y . Sidky, Y . Zou, and X. Pan, “Impact of polychromatic x-ray sources on helical, cone-beam computed tomography and dual-energy methods,” Physics in Medicine & Biology, vol. 49, no. 11, p. 2293, 2004

    work page 2004

  11. [11]

    Empirical dual energy cali- bration (edec) for cone-beam computed tomography,

    P. Stenner, T. Berkus, and M. Kachelriess, “Empirical dual energy cali- bration (edec) for cone-beam computed tomography,”Medical physics, vol. 34, no. 9, pp. 3630–3641, 2007

    work page 2007

  12. [12]

    Comparison of four dual energy image decomposition methods,

    K.-S. Chuang and H. Huang, “Comparison of four dual energy image decomposition methods,”Physics in Medicine & Biology, vol. 33, no. 4, p. 455, 1988

    work page 1988

  13. [13]

    An extended algebraic reconstruction technique (e-art) for dual spectral ct,

    Y . Zhao, X. Zhao, and P. Zhang, “An extended algebraic reconstruction technique (e-art) for dual spectral ct,”IEEE transactions on medical imaging, vol. 34, no. 3, pp. 761–768, 2014

    work page 2014

  14. [14]

    An extended simultaneous algebraic reconstruction technique (e-sart) for x-ray dual spectral computed to- mography,

    J. Hu, X. Zhao, and F. Wang, “An extended simultaneous algebraic reconstruction technique (e-sart) for x-ray dual spectral computed to- mography,”Scanning, vol. 38, no. 6, pp. 599–611, 2016

    work page 2016

  15. [15]

    Accurate iterative fbp reconstruction method for material decomposition of dual energy ct,

    M. Li, Y . Zhao, and P. Zhang, “Accurate iterative fbp reconstruction method for material decomposition of dual energy ct,”IEEE transactions on medical imaging, vol. 38, no. 3, pp. 802–812, 2018

    work page 2018

  16. [16]

    Fast iterative reconstruction for multi-spectral ct by a schmidt orthogonal modification algorithm (soma),

    H. Pan, S. Zhao, W. Zhang, H. Zhang, and X. Zhao, “Fast iterative reconstruction for multi-spectral ct by a schmidt orthogonal modification algorithm (soma),”Inverse Problems, vol. 39, no. 8, p. 085001, 2023

    work page 2023

  17. [17]

    Image reconstruc- tion of multiple basis materials with data augmentation in dect,

    B. Chen, Z. Zhang, D. Xia, E. Y . Sidky, and X. Pan, “Image reconstruc- tion of multiple basis materials with data augmentation in dect,”IEEE Transactions on Biomedical Engineering, vol. 72, no. 5, pp. 1553–1561, 2025

    work page 2025

  18. [18]

    Multi-material decomposition using statisti- cal image reconstruction for spectral ct,

    Y . Long and J. A. Fessler, “Multi-material decomposition using statisti- cal image reconstruction for spectral ct,”IEEE transactions on medical imaging, vol. 33, no. 8, pp. 1614–1626, 2014

    work page 2014

  19. [19]

    Numerical algorithms for polyenergetic digital breast tomosynthesis reconstruction,

    J. Chung, J. G. Nagy, and I. Sechopoulos, “Numerical algorithms for polyenergetic digital breast tomosynthesis reconstruction,”SIAM Journal on Imaging Sciences, vol. 3, no. 1, pp. 133–152, 2010

    work page 2010

  20. [20]

    Numerical solution of a nonlinear least squares problem in digital breast tomosynthesis,

    G. Landi, E. L. Piccolomini, and J. Nagy, “Numerical solution of a nonlinear least squares problem in digital breast tomosynthesis,” in Journal of Physics: Conference Series, vol. 657, no. 1. IOP Publishing, 2015, p. 012006

    work page 2015

  21. [21]

    Proximal admm for multi-channel image reconstruction in spectral x-ray ct,

    A. Sawatzky, Q. Xu, C. O. Schirra, and M. A. Anastasio, “Proximal admm for multi-channel image reconstruction in spectral x-ray ct,”IEEE transactions on medical imaging, vol. 33, no. 8, pp. 1657–1668, 2014

    work page 2014

  22. [22]

    An algorithm for constrained one-step inversion of spectral ct data,

    R. F. Barber, E. Y . Sidky, T. G. Schmidt, and X. Pan, “An algorithm for constrained one-step inversion of spectral ct data,”Physics in Medicine & Biology, vol. 61, no. 10, p. 3784, 2016

    work page 2016

  23. [23]

    A spectral ct method to directly estimate basis material maps from experimental photon-counting data,

    T. G. Schmidt, R. F. Barber, and E. Y . Sidky, “A spectral ct method to directly estimate basis material maps from experimental photon-counting data,”IEEE transactions on medical imaging, vol. 36, no. 9, pp. 1808– 1819, 2017

    work page 2017

  24. [24]

    Statistical image reconstruction for polyenergetic x-ray computed tomography,

    I. A. Elbakri and J. A. Fessler, “Statistical image reconstruction for polyenergetic x-ray computed tomography,”IEEE transactions on med- ical imaging, vol. 21, no. 2, pp. 89–99, 2002

    work page 2002

  25. [25]

    Multi-energy ct decompo- sition using convolutional neural networks,

    D. P. Clark, M. Holbrook, and C. T. Badea, “Multi-energy ct decompo- sition using convolutional neural networks,” inMedical Imaging 2018: Physics of Medical Imaging, vol. 10573. SPIE, 2018, pp. 415–423

    work page 2018

  26. [26]

    A material decomposi- tion method for dual-energy ct via dual interactive wasserstein generative adversarial networks,

    Z. Shi, H. Li, Q. Cao, Z. Wang, and M. Cheng, “A material decomposi- tion method for dual-energy ct via dual interactive wasserstein generative adversarial networks,”Medical Physics, vol. 48, no. 6, pp. 2891–2905, 2021

    work page 2021

  27. [27]

    Virtual monoenergetic ct imaging via deep learning,

    W. Cong, Y . Xi, P. Fitzgerald, B. De Man, and G. Wang, “Virtual monoenergetic ct imaging via deep learning,”Patterns, vol. 1, no. 8, 2020

    work page 2020

  28. [28]

    Image domain dual material decomposition for dual-energy ct using butterfly network,

    W. Zhang, H. Zhang, L. Wang, X. Wang, X. Hu, A. Cai, L. Li, T. Niu, and B. Yan, “Image domain dual material decomposition for dual-energy ct using butterfly network,”Medical physics, vol. 46, no. 5, pp. 2037– 2051, 2019

    work page 2037

  29. [29]

    Image reconstruction by domain-transform manifold learning,

    B. Zhu, J. Z. Liu, S. F. Cauley, B. R. Rosen, and M. S. Rosen, “Image reconstruction by domain-transform manifold learning,”Nature, vol. 555, no. 7697, pp. 487–492, 2018

    work page 2018

  30. [30]

    Material decomposition in spectral ct using deep learning: a sim2real transfer approach,

    J. F. Abascal, N. Ducros, V . Pronina, S. Rit, P.-A. Rodesch, T. Broussaud, S. Bussod, P. C. Douek, A. Hauptmann, and S. Arridge, “Material decomposition in spectral ct using deep learning: a sim2real transfer approach,”IEEE Access, vol. 9, pp. 25 632–25 647, 2021

    work page 2021

  31. [31]

    Deep-learning-based direct inversion for material decomposi- tion,

    H. Gong, S. Tao, K. Rajendran, W. Zhou, C. H. McCollough, and S. Leng, “Deep-learning-based direct inversion for material decomposi- tion,”Medical physics, vol. 47, no. 12, pp. 6294–6309, 2020

    work page 2020

  32. [32]

    Implicit neural representation in medical imaging: A comparative survey,

    A. Molaei, A. Aminimehr, A. Tavakoli, A. Kazerouni, B. Azad, R. Azad, and D. Merhof, “Implicit neural representation in medical imaging: A comparative survey,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 2381–2391

    work page 2023

  33. [33]

    Coil: Coordinate-based internal learning for tomographic imaging,

    Y . Sun, J. Liu, M. Xie, B. Wohlberg, and U. S. Kamilov, “Coil: Coordinate-based internal learning for tomographic imaging,”IEEE Transactions on Computational Imaging, vol. 7, pp. 1400–1412, 2021

    work page 2021

  34. [34]

    Intratomo: self-supervised learning-based tomography via sinogram synthesis and prediction,

    G. Zang, R. Idoughi, R. Li, P. Wonka, and W. Heidrich, “Intratomo: self-supervised learning-based tomography via sinogram synthesis and prediction,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 1960–1970

    work page 2021

  35. [35]

    Neat: Neural adaptive tomography,

    D. R ¨uckert, Y . Wang, R. Li, R. Idoughi, and W. Heidrich, “Neat: Neural adaptive tomography,”ACM Transactions on Graphics (TOG), vol. 41, no. 4, pp. 1–13, 2022

    work page 2022

  36. [36]

    Nerp: implicit neural representation learning with prior embedding for sparsely sampled image reconstruc- tion,

    L. Shen, J. Pauly, and L. Xing, “Nerp: implicit neural representation learning with prior embedding for sparsely sampled image reconstruc- tion,”IEEE Transactions on Neural Networks and Learning Systems, 2022

    work page 2022

  37. [37]

    Dynamic ct reconstruction from limited views with implicit neural representations and parametric motion fields,

    A. W. Reed, H. Kim, R. Anirudh, K. A. Mohan, K. Champley, J. Kang, and S. Jayasuriya, “Dynamic ct reconstruction from limited views with implicit neural representations and parametric motion fields,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 2258–2268

    work page 2021

  38. [38]

    Naf: Neural attenuation fields for sparse- view cbct reconstruction,

    R. Zha, Y . Zhang, and H. Li, “Naf: Neural attenuation fields for sparse- view cbct reconstruction,” inInternational Conference on Medical Image Computing and Computer-Assisted Intervention. Springer, 2022, pp. 442–452

    work page 2022

  39. [39]

    Uncertainr: Uncertainty quantification of end-to-end implicit neural representations for computed tomography,

    F. Vasconcelos, B. He, N. Singh, and Y . W. Teh, “Uncertainr: Uncertainty quantification of end-to-end implicit neural representations for computed tomography,”arXiv preprint arXiv:2202.10847, 2022

    work page arXiv 2022

  40. [40]

    Ring artifacts removal based on implicit neural representation of sinogram data,

    L. Shi, X. Jiang, Y . Liu, C. Liu, P. Yang, S. Guo, and X. Zhao, “Ring artifacts removal based on implicit neural representation of sinogram data,”IEEE Transactions on Image Processing, vol. 34, pp. 4080–4091, 2025

    work page 2025

  41. [41]

    Recovery of continuous 3d refractive index maps from discrete intensity-only measurements using neural fields,

    R. Liu, Y . Sun, J. Zhu, L. Tian, and U. S. Kamilov, “Recovery of continuous 3d refractive index maps from discrete intensity-only measurements using neural fields,”Nature Machine Intelligence, vol. 4, no. 9, pp. 781–791, 2022. SHELLet al.: A SAMPLE ARTICLE USING IEEETRAN.CLS FOR IEEE JOURNALS 11

    work page 2022

  42. [42]

    Coin: Compression with implicit neural representations,

    E. Dupont, A. Goli ´nski, M. Alizadeh, Y . W. Teh, and A. Doucet, “Coin: Compression with implicit neural representations,”arXiv preprint arXiv:2103.03123, 2021

    work page arXiv 2021

  43. [43]

    On the spectral bias of neural networks,

    N. Rahaman, A. Baratin, D. Arpit, F. Draxler, M. Lin, F. Hamprecht, Y . Bengio, and A. Courville, “On the spectral bias of neural networks,” inInternational conference on machine learning. PMLR, 2019, pp. 5301–5310

    work page 2019

  44. [44]

    Where do we stand with implicit neural representations? a technical and performance survey,

    A. Essakine, Y . Cheng, C.-W. Cheng, L. Zhang, Z. Deng, L. Zhu, C.-B. Sch ¨onlieb, and A. I. Aviles-Rivero, “Where do we stand with implicit neural representations? a technical and performance survey,” Transactions on Machine Learning Research, 2025, survey Certification. [Online]. Available: https://openreview.net/forum?id=QTsJXSvAI2

    work page 2025

  45. [45]

    Neural fields in visual computing and beyond,

    Y . Xie, T. Takikawa, S. Saito, O. Litany, S. Yan, N. Khan, F. Tombari, J. Tompkin, V . Sitzmann, and S. Sridhar, “Neural fields in visual computing and beyond,” inComputer Graphics F orum, vol. 41, no. 2. Wiley Online Library, 2022, pp. 641–676

    work page 2022

  46. [46]

    Implicit neural representations with periodic activation functions,

    V . Sitzmann, J. Martel, A. Bergman, D. Lindell, and G. Wetzstein, “Implicit neural representations with periodic activation functions,” Advances in neural information processing systems, vol. 33, pp. 7462– 7473, 2020

    work page 2020

  47. [47]

    Incode: Implicit neural conditioning with prior knowledge embeddings,

    A. Kazerouni, R. Azad, A. Hosseini, D. Merhof, and U. Bagci, “Incode: Implicit neural conditioning with prior knowledge embeddings,” in 2024 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2024, pp. 1287–1296

    work page 2024

  48. [48]

    Learned initializations for optimizing coordinate- based neural representations,

    M. Tancik, B. Mildenhall, T. Wang, D. Schmidt, P. P. Srinivasan, J. T. Barron, and R. Ng, “Learned initializations for optimizing coordinate- based neural representations,” inProceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition, 2021, pp. 2846–2855

    work page 2021

  49. [49]

    Sourbelle

    K. Sourbelle. FORBILD thorax phantom. [Online]. Available: https://web.archive.org/web/20070611224607/http://www. imp.uni-erlangen.de/forbild/deutsch/results/thorax/thorax.htm

    work page arXiv

  50. [50]

    X-ray mass attenuation coefficients,

    J. H. Hubbell and S. M. Seltzer, “X-ray mass attenuation coefficients,” 2004. [Online]. Available: https://physics.nist.gov/ PhysRefData/XrayMassCoef/tab4.html [51]Spectrum GUI, 1st ed., SourceForge, 2007. [Online]. Available: https://sourceforge.net/projects/spectrumgui/

    work page 2004