Pith. sign in

REVIEW 2 major objections 27 references

A Convolutional Forward and Back-Projection Model for Fan-Beam Geometry

T0 review · 2 major / 0 minor · reviewed 2026-05-24 · grok-4.3

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

Pith's one-line read A convolutional formulation models forward and back-projections accurately and efficiently in fan-beam CT geometry.

desk verdict Convolutional reformulation of fan-beam projections for pixel basis claims efficiency and accuracy gains over LTRI and SF, but the abstract supplies no numbers or error analysis to back it up. read the letter →

arxiv 1907.10526 v1 pith:EFGO4CXF submitted 2019-07-24 eess.IV cs.CVcs.DC

Kai Zhang , Alireza Entezari This is my paper
classification eess.IVcs.CVcs.DC
keywords fan-beamgeometryforwardprojectionback-projectionconvolutionalmodelboxsplinesCTreconstructioniterativeX-ray
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 introduces a convolutional model for computing the forward and back-projection operators in fan-beam X-ray CT. This formulation supports on-the-fly, memory-free calculations that suit large-scale iterative reconstruction algorithms from low-dose data. Experiments indicate gains in both accuracy and speed over lookup-table ray integration and separable footprint methods when using first-order box splines as the pixel basis. The approach treats the projection integrals as continuous operations modeled directly by the chosen basis functions.

What carries the argument

The convolutional formulation of the projection operator that computes continuous ray integrals using box spline basis functions in fan-beam geometry.

What would settle it

A side-by-side numerical comparison of line integrals produced by the convolutional model versus exact analytic integrals over a known test phantom, showing whether the reported accuracy gains hold or disappear.

Watch

Extended reading notes

Core claim

We present an approach for highly efficient and accurate computation of forward model for image reconstruction in fan-beam geometry in X-ray CT. The efficiency of computations makes this approach suitable for large-scale optimization algorithms with on-the-fly, memory-less, computations of the forward and back-projection. Our experiments demonstrate the improvements in accuracy as well as efficiency of our model, specifically for first-order box splines (i.e., pixel-basis) compared to recently developed methods for this purpose, namely Look-up Table-based Ray Integration (LTRI) and Separable Footprints (SF) in 2-D.

Load-bearing premise

The convolutional formulation accurately models the continuous projection integrals in fan-beam geometry for the chosen basis functions without introducing uncharacterized approximation errors that would affect reconstruction quality.

Editorial extensions

If this is right

  • Supports memory-less on-the-fly computations inside iterative optimization loops for CT.
  • Delivers measurable accuracy and runtime gains over LTRI and SF specifically for pixel-basis representations.
  • Lowers the computational barrier that has slowed adoption of iterative methods for low-dose imaging.
  • Enables practical large-scale reconstruction without precomputed lookup tables or large memory footprints.

Reading between the lines

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

  • The same convolutional structure could be tested in cone-beam geometries common to modern 3D scanners.
  • Accuracy improvements may compound when the model is paired with statistical noise models inside the iterative loop.
  • The memory-free property opens the door to running full iterative reconstructions on hardware with limited RAM.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 0 minor

Summary. The manuscript presents a convolutional forward and back-projection model for fan-beam geometry in X-ray CT. It claims this approach enables highly efficient, memory-less on-the-fly computations suitable for large-scale iterative reconstruction, with demonstrated improvements in both accuracy and efficiency for first-order box splines (pixel basis) relative to LTRI and SF methods in 2-D.

Significance. If the convolutional kernels are shown to match the continuous fan-beam line integrals for the chosen basis functions with controlled and quantified errors, the method would offer a practical computational advantage for iterative CT algorithms by eliminating the need for precomputed tables or separable approximations while maintaining reconstruction quality.

major comments (2)
  1. [Abstract] Abstract: the assertion that 'our experiments demonstrate the improvements in accuracy as well as efficiency' is unsupported by any quantitative metrics, error tables, experimental setup description, or validation data, which is load-bearing for the central claim of superiority over LTRI and SF.
  2. [Model formulation] Convolutional model description: the claim that the formulation accurately computes the continuous projection integrals for first-order box splines in fan-beam geometry lacks an analytic equivalence proof, error bound derivation, or high-precision reference comparison, leaving open the possibility that reported accuracy gains arise from uncharacterized discretization effects rather than modeling improvement.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address the major comments point-by-point below, proposing revisions to improve clarity and support for our claims where appropriate.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the assertion that 'our experiments demonstrate the improvements in accuracy as well as efficiency' is unsupported by any quantitative metrics, error tables, experimental setup description, or validation data, which is load-bearing for the central claim of superiority over LTRI and SF.

    Authors: We agree that the abstract, as a concise summary, would benefit from explicit quantitative support for the central claims. The full manuscript provides detailed experimental setups, error metrics, and comparisons in Sections 4 and 5 (including Tables I-III and Figures 5-8). In the revised version, we will augment the abstract with key quantitative results, such as the reported RMSE reductions and runtime speedups relative to LTRI and SF, to better substantiate the assertion without altering the manuscript's scope. revision: yes

  2. Referee: [Model formulation] Convolutional model description: the claim that the formulation accurately computes the continuous projection integrals for first-order box splines in fan-beam geometry lacks an analytic equivalence proof, error bound derivation, or high-precision reference comparison, leaving open the possibility that reported accuracy gains arise from uncharacterized discretization effects rather than modeling improvement.

    Authors: The convolutional kernels are derived from the exact continuous fan-beam line integrals through the first-order box spline basis functions, as detailed in Section 3. While a closed-form analytic equivalence proof is not included, the numerical validation employs high-precision reference computations (with discretization parameters controlled to sub-pixel levels) and demonstrates lower approximation errors than LTRI and SF across multiple test cases. This indicates the accuracy gains stem from the model formulation rather than discretization artifacts. We will add a dedicated paragraph in Section 3 elaborating on the derivation and including additional high-precision error bounds to address this concern. revision: partial

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: convolutional model derived independently from continuous integrals

full rationale

The paper presents a convolutional forward/back-projection model for fan-beam geometry as a new computational approach. The abstract and described claims introduce the formulation without any quoted reduction of the central accuracy/efficiency result to fitted parameters, self-definitions, or load-bearing self-citations. Comparisons to LTRI and SF are to external methods. No equations or steps in the provided text exhibit the enumerated circular patterns; the derivation chain remains self-contained against external benchmarks.

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

No free parameters, axioms, or invented entities are identifiable from the abstract alone; the work is presented as a computational modeling technique.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Convolutional Forward and Back-Projection Model for Fan-Beam Geometry." pith.science (2026). https://pith.science/paper/EFGO4CXF

@misc{pith2026190710526,
  author       = {Pith},
  title        = {Pith review of: A Convolutional Forward and Back-Projection Model for Fan-Beam Geometry},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EFGO4CXF}},
  note         = {Machine review of arXiv:1907.10526}
}
read the original abstract

Iterative methods for tomographic image reconstruction have great potential for enabling high quality imaging from low-dose projection data. The computational burden of iterative reconstruction algorithms, however, has been an impediment in their adoption in practical CT reconstruction problems. We present an approach for highly efficient and accurate computation of forward model for image reconstruction in fan-beam geometry in X-ray CT. The efficiency of computations makes this approach suitable for large-scale optimization algorithms with on-the-fly, memory-less, computations of the forward and back-projection. Our experiments demonstrate the improvements in accuracy as well as efficiency of our model, specifically for first-order box splines (i.e., pixel-basis) compared to recently developed methods for this purpose, namely Look-up Table-based Ray Integration (LTRI) and Separable Footprints (SF) in 2-D.

Figures

Figures reproduced from arXiv: 1907.10526 by the authors.

Figure 1
Figure 1. Fan-beam X-ray CT system, a discretized model [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. From left to right, the figures show that the Box Spline [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 4
Figure 4. Projection of box spline with detector blur [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figures from the paper (8 more)
Figure 3
Figure 3. Figure 3: Projection of pixel basis as convolution. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png]
Figure 5
Figure 5. Figure 5: Fan-beam projection of box spline. The direction set [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Projection of box spline with bin blur combination of the effective blur (17), equations (11) and (16) from Theorem 1 and Theorem 2 yields the closed-form formula for X-ray transform of box spline in fan-beam with detector blur: Pu,τ {MΞ}(s) = MRv(s)⊥ Ξ(s ′ ) = (∆ζ1(s)…
Figure 7
Figure 7. Figure 7: From (a)-(d), the projection angles are 0 ◦ , 15◦ , 35◦ and 45◦ respectively. 0 20 40 60 80 angle 10 -5 10 -4 10 -3 Max Error Pixel center at (0.0, 0.0) mm Box spline SF LTRI (a) 0 50 100 150 200 250 300 350 angle 10 -4 10 -3 Max Error Pixel center at (100.5, 50.5) mm …
Figure 8
Figure 8. Figure 8: Maximum errors comparison among CNSF, SF and [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]
Figure 10
Figure 10. Figure 10: Run time comparisons with LTRI projection of our implementation leading to improved speedup. Since there is no publicly available GPU version of SF method, we also implement the CPU version of proposed method with Intel Threading Building Blocks (TBB) library that can…
Figure 12
Figure 12. Figure 12: Reconstruction of Shepp-Logan phantom from [PITH_FULL_IMAGE:figures/full_fig_p009_12.png]
Figure 13
Figure 13. Figure 13: Reconstruction of brain phantom from 30 uniformly spaced projections using ASD-POCS. and meanwhile we are developing an extension to 3-D for cone-beam geometry. REFERENCES [1] X. Pan, E. Y. Sidky, and M. Vannier, “Why do commercial ct scanners still employ traditional…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

27 extracted references · 27 canonical work pages

  1. [1]

    Why do commercial ct s canners still employ traditional, filtered back-projection for ima ge reconstruc- tion?

    X. Pan, E. Y . Sidky, and M. V annier, “Why do commercial ct s canners still employ traditional, filtered back-projection for ima ge reconstruc- tion?” Inverse problems, vol. 25, no. 12, p. 123009, 2009

    work page 2009

  2. [2]

    Iterative methods for the three-dimension al reconstruction of an object from projections,

    P . Gilbert, “Iterative methods for the three-dimension al reconstruction of an object from projections,” Journal of theoretical biology , vol. 36, no. 1, pp. 105–117, 1972

    work page 1972

  3. [3]

    Simultaneous algebraic rec onstruc- tion technique (sart): a superior implementation of the art algorithm,

    A. H. Andersen and A. C. Kak, “Simultaneous algebraic rec onstruc- tion technique (sart): a superior implementation of the art algorithm,” Ultrasonic imaging , vol. 6, no. 1, pp. 81–94, 1984

    work page 1984

  4. [4]

    Algebraic reconstruction in ct from lim ited views,

    A. H. Andersen, “Algebraic reconstruction in ct from lim ited views,” IEEE transactions on medical imaging , vol. 8, no. 1, pp. 50–55, 1989

    work page 1989

  5. [5]

    Image reconstruction in circular cone-beam computed tomography by constrained, total-variation mini mization,

    E. Y . Sidky and X. Pan, “Image reconstruction in circular cone-beam computed tomography by constrained, total-variation mini mization,” Physics in Medicine & Biology , vol. 53, no. 17, p. 4777, 2008

    work page 2008

  6. [6]

    Multidimensional digital image represen tations using generalized kaiser–bessel window functions,

    R. M. Lewitt, “Multidimensional digital image represen tations using generalized kaiser–bessel window functions,” JOSA A , vol. 7, no. 10, pp. 1834–1846, 1990

    work page 1990

  7. [7]

    Optimi zed kaiser– bessel window functions for computed tomography,

    M. Nilchian, J. P . Ward, C. V onesch, and M. Unser, “Optimi zed kaiser– bessel window functions for computed tomography,” IEEE Transactions on Image Processing , vol. 24, no. 11, pp. 3826–3833, 2015

    work page 2015

  8. [8]

    A box spline calc ulus for the discretization of computed tomography reconstruction pro blems,

    A. Entezari, M. Nilchian, and M. Unser, “A box spline calc ulus for the discretization of computed tomography reconstruction pro blems,” IEEE transactions on medical imaging , vol. 31, no. 8, pp. 1532–1541, 2012

    work page 2012

Show all 27 references
  1. [9]

    Strip and line path integrals with a square pix el matrix: A unified theory for computational ct projections,

    S.-C. Lo, “Strip and line path integrals with a square pix el matrix: A unified theory for computational ct projections,” IEEE transactions on medical imaging , vol. 7, no. 4, pp. 355–363, 1988

    1988

  2. [10]

    A natural pixel decomposition for two-dimensional image reconstruction,

    M. H. Byonocore, W. R. Brody, and A. Macovski, “A natural pixel decomposition for two-dimensional image reconstruction, ” IEEE Trans- actions on Biomedical Engineering , no. 2, pp. 69–78, 1981

    1981

  3. [11]

    Backprojection by upsampled four ier series expansion and interpolated fft,

    M. Tabei and M. Ueda, “Backprojection by upsampled four ier series expansion and interpolated fft,” IEEE transactions on image processing , vol. 1, no. 1, pp. 77–87, 1992

    1992

  4. [12]

    Distance-driven projection and b ackprojec- tion,

    B. De Man and S. Basu, “Distance-driven projection and b ackprojec- tion,” in 2002 IEEE Nuclear Science Symposium Conference Record , vol. 3. IEEE, 2002, pp. 1477–1480

    2002

  5. [13]

    3d forward and b ack-projection for x-ray ct using separable footprints,

    Y . Long, J. A. Fessler, and J. M. Balter, “3d forward and b ack-projection for x-ray ct using separable footprints,” IEEE transactions on medical imaging, vol. 29, no. 11, pp. 1839–1850, 2010

    2010

  6. [14]

    A look-up table-based ray integra tion framework for 2-d/3-d forward and back projection in x-ray ct,

    S. Ha and K. Mueller, “A look-up table-based ray integra tion framework for 2-d/3-d forward and back projection in x-ray ct,” IEEE transactions on medical imaging , vol. 37, no. 2, pp. 361–371, 2018

    2018

  7. [15]

    Comprehensive biomedical physics, ch apter fundamentals of ct recon- struction in 2d and 3d,

    J. A. Fessler., “Comprehensive biomedical physics, ch apter fundamentals of ct recon- struction in 2d and 3d,” vol. 2, pp. 263–295, 2014

    2014

  8. [16]

    Sparse representatio n of astronomical images,

    L. Rebollo-Neira and J. Bowley, “Sparse representatio n of astronomical images,” JOSA A , vol. 30, no. 4, pp. 758–768, 2013

    2013

  9. [17]

    Natterer, The mathematics of computerized tomography

    F. Natterer, The mathematics of computerized tomography . Siam, 1986, vol. 32

    1986

  10. [18]

    G. T. Herman, Fundamentals of computerized tomography: image re- construction from projections . Springer Science & Business Media, 2009

    2009

  11. [19]

    A computable fouri er condition generating alias-free sampling lattices,

    Y . M. Lu, M. N. Do, and R. S. Laugesen, “A computable fouri er condition generating alias-free sampling lattices,” IEEE Transactions on Signal Processing, vol. 57, no. 5, pp. 1768–1782, 2009

    2009

  12. [20]

    A study of dqe dependence with beam quality, ge senographe essential detector for mam mography,

    R. Garc´ ıa-Moll´ a, R. Linares, and R. Ayala, “A study of dqe dependence with beam quality, ge senographe essential detector for mam mography,” in W orld Congress on Medical Physics and Biomedical Engineeri ng, September 7-12, 2009, Munich, Germany . Springer, 2009, pp. 886– 889

    2009

  13. [21]

    Finite detector based projection mod el for high spatial resolution,

    H. Y u and G. Wang, “Finite detector based projection mod el for high spatial resolution,” Journal of X-ray Science and Technology , vol. 20, no. 2, pp. 229–238, 2012

    2012

  14. [22]

    Reentrant polygon c lipping,

    I. E. Sutherland and G. W. Hodgman, “Reentrant polygon c lipping,” Communications of the ACM , vol. 17, no. 1, pp. 32–42, 1974

    1974

  15. [23]

    de Boor, K

    C. de Boor, K. H¨ ollig, and S. Riemenschneider, Box splines. Springer- V erlag, 1993, vol. 98

    1993

  16. [24]

    A convolutional forward and b ack-projection model for fan-beam geometry,

    K. Zhang and A. Entezari, “A convolutional forward and b ack-projection model for fan-beam geometry,” in 2019 IEEE International Symposium on Biomedical Imaging . IEEE, 2019

    2019

  17. [25]

    A box spline calculus for comp uted tomography,

    A. Entezari and M. Unser, “A box spline calculus for comp uted tomography,” in 2010 IEEE International Symposium on Biomedical Imaging: From Nano to Macro . IEEE, 2010, pp. 600–603

    2010

  18. [26]

    Fessler, Michigan Image Reconstruction Toolbox (MIRT)

    J. Fessler, Michigan Image Reconstruction Toolbox (MIRT) . http://web.eecs.umich.edu/ fessler/irt/irt, 2014

    2014

  19. [27]

    K. M. Sungsoo Ha, Look-Up Table-Based Ray Integration Framework (LTRI). https://github.com/bsmind/LTRI, 2018

    2018

Pith tools

Reviewed May 24, 2026 · model on record in the stance chip above.