pith. sign in

arxiv: 2606.20395 · v1 · pith:QUPYGXQKnew · submitted 2026-06-18 · ⚛️ physics.med-ph

Efficient and Accurate Image Reconstruction for Geometric-Inconsistent Multispectral CT with Ray-Dependent Energy Spectra

Ziqiang Zhang , Chong Chen This is my paper

Pith reviewed 2026-06-26 14:36 UTC · model grok-4.3

classification ⚛️ physics.med-ph
keywords multispectral CTimage reconstructionJacobian approximationaggregated energy spectraray-dependent spectrageometric inconsistencynonlinear forward operatorconvergence theory
0
0 comments X

The pith

Aggregated energy spectra let the Jacobian of the nonlinear forward operator in multispectral CT factor as a block product of a diagonal projection matrix and a small matrix, supporting an efficient reconstruction algorithm with convergence

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

Multispectral CT scanners often use inconsistent geometries across energy spectra, and the spectra themselves can vary by ray. The paper shows that aggregated energy spectra make it possible to approximate the Jacobian of the nonlinear forward operator at special points, such as the zero point, as the product of a diagonal matrix of projection matrices and a very small-scale matrix. This block-product structure is then used to derive a specialized reconstruction algorithm. The algorithm comes with convergence guarantees under suitable conditions and is shown in experiments to run faster and produce more accurate images than previous methods on both noiseless and noisy data.

Core claim

Using the proposed aggregated energy spectra, we approximate the Jacobian matrix of the nonlinear forward operator at certain special points (e.g., the zero point) as a block product of a diagonal matrix composed of projection matrices and a very small-scale matrix, and then based on this matrix with a special structure, propose an efficient and accurate image reconstruction algorithm tailored for geometric-inconsistent MSCT with ray-dependent energy spectra. Under appropriate conditions, we establish the convergence theory for the proposed algorithm. Furthermore, numerical experiments using both noiseless and noisy projection data are conducted to verify the performance of the proposed algo

What carries the argument

Jacobian matrix of the nonlinear forward operator approximated via aggregated energy spectra as the block product of a diagonal matrix of projection matrices and a small-scale matrix

Load-bearing premise

The Jacobian approximation at special points such as the zero point stays accurate for the full nonlinear forward operator when aggregated energy spectra are used in the presence of geometric inconsistency and ray-dependent spectra.

What would settle it

A simulation in which the proposed algorithm fails to converge or produces large reconstruction errors once the true derivative of the forward operator at the zero point deviates markedly from the claimed block-product form.

Figures

Figures reproduced from arXiv: 2606.20395 by Chong Chen, Ziqiang Zhang.

Figure 1
Figure 1. Figure 1: Part of the ray-dependent energy spectra used in the tests. The left figure shows the 80 kV spectra, and the right figure shows the 140 kV spectra. 5.2 Simulated noiseless data for the forbild phantom In this test, a forbild phantom is composed of the truth water and bone basis images which are both discretized as 128×128 pixels on the square area [−5, 5]× [−5, 5] cm2 and take gray values over [0, 1], as s… view at source ↗
Figure 2
Figure 2. Figure 2: Metrics REk f (left) and REk g (right) are plotted in semi-log scale over iteration numbers for reconstructing basis images of the forbild phantom by the proposed algorithm, NCPD and NKM [PITH_FULL_IMAGE:figures/full_fig_p015_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Metrics REk f (left) and REk g (right) are plotted in semi-log scale over time (in seconds) for reconstructing basis images of the forbild phantom by the proposed algorithm, NCPD and NKM. These results demonstrate that the proposed algorithm enables efficient and high-precision inversion for noiseless data. The reconstructed basis im￾ages yielded by all methods, together with the VMIs and their correspondi… view at source ↗
Figure 4
Figure 4. Figure 4: Reconstruction results in the noiseless case. From left to right: basis images of water and bone, VMIs at energies 60 keV and 100 keV of the forbild phantom. From top to bottom: the true images, the results after 100 iterations using the proposed algorithm, NCPD and NKM [PITH_FULL_IMAGE:figures/full_fig_p016_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Metrics REk f (left) and REk g (right) are plotted in semi-log scale over iteration numbers for reconstructing basis images of the realistic torso image by the proposed algorithm, NCPD and NKM. The reconstructed basis images, VMIs and their corresponding ground truth references obtained by all methods are shown in fig. 8, while the absolute dif￾ference maps are displayed in fig. 9. From these difference ma… view at source ↗
Figure 6
Figure 6. Figure 6: Metrics REk f (left) and REk g (right) are plotted in semi-log scale over time (in seconds) for reconstructing basis images of the realistic torso image by the proposed algorithm, NCPD and NKM [PITH_FULL_IMAGE:figures/full_fig_p018_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Metrics ∆k f (left) and ∆k g (right) are plotted in semi-log scale over itera￾tion numbers for reconstructing basis images of the realistic torso image by the proposed algorithm, NCPD and NKM [PITH_FULL_IMAGE:figures/full_fig_p018_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Reconstruction results in the noisy case. From left to right: basis images of water and bone, VMIs at energies 60 keV and 100 keV of the realistic torso image. From top to bottom: the true images, the results after 50 iterations using the proposed algorithm, NCPD and NKM [PITH_FULL_IMAGE:figures/full_fig_p019_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Absolute difference maps between the reconstructed results and the ground truths in the noisy case. Each column from left to right cor￾responds to the water basis image, bone basis image, VMI at 60 keV, and VMI at 100 keV, respectively. Each row from top to bottom shows the absolute difference maps for the proposed algorithm, NCPD, and NKM, rspectively [PITH_FULL_IMAGE:figures/full_fig_p020_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Metrics REk f (left) and REk g (right) are plotted in semi-log scale over iteration numbers for reconstructing basis images of from fig. 11 by the proposed algorithm with different aggregated energy spectra. from the ray indeces. With this ray-wise decoupling, we derived an approximate Jacobian matrix that can be expressed as a block product of a diagonal matrix of projection matrices and a very small-sca… view at source ↗
Figure 11
Figure 11. Figure 11: Reconstruction results in the noiseless case. From left to right: basis images of water and bone, VMIs at energies 60 keV and 100 keV of a clinical image. From top to bottom: the true images, the results after 100 iterations using the proposed algorithm with different aggregated energy spectra [PITH_FULL_IMAGE:figures/full_fig_p023_11.png] view at source ↗
read the original abstract

In practical multispectral computed tomography (MSCT), the scanning geometric parameters under different X-ray energy spectra are often inconsistent, and the distributions of the energy spectra are even ray-dependent. However, existing algorithms cannot effectively and accurately solve the associated image reconstruction problem. To address this limitation, using the proposed aggregated energy spectra, we approximate the Jacobian matrix of the nonlinear forward operator at certain special points (e.g., the zero point) as a block product of a diagonal matrix composed of projection matrices and a very small-scale matrix, and then based on this matrix with a special structure, propose an efficient and accurate image reconstruction algorithm tailored for geometric-inconsistent MSCT with ray-dependent energy spectra. Under appropriate conditions, we establish the convergence theory for the proposed algorithm. Furthermore, numerical experiments using both noiseless and noisy projection data are conducted to verify the performance of the proposed algorithm, which demonstrate that the efficiency and accuracy of this algorithm are much higher than existing algorithms, offering the flexibility and scalability to accommodate various MSCT imaging configurations.

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

1 major / 0 minor

Summary. The paper addresses image reconstruction in multispectral CT (MSCT) under geometric inconsistency across energy spectra and ray-dependent spectra. It introduces aggregated energy spectra to approximate the Jacobian of the nonlinear forward operator at special points (e.g., the zero vector) as a block product of a diagonal matrix of projection matrices and a small-scale matrix. This structure underpins an efficient reconstruction algorithm for which convergence theory is established under appropriate conditions. Numerical experiments on noiseless and noisy data are reported to show superior efficiency and accuracy relative to existing methods, with flexibility for various MSCT configurations.

Significance. If the Jacobian approximation remains accurate with controlled error away from the special points and the convergence theory applies to realistic iterates, the work would provide a scalable, theoretically supported solution for a practically relevant class of MSCT problems that current methods handle poorly. The explicit use of aggregated spectra to obtain a structured Jacobian, combined with convergence guarantees and numerical validation, would represent a meaningful advance in algorithmic efficiency for nonlinear inverse problems in medical imaging.

major comments (1)
  1. [Abstract] Abstract: The central algorithmic and theoretical claims rest on approximating the Jacobian at special points such as the zero vector and then using the resulting block structure for the full nonlinear reconstruction. No explicit error bound or analysis is indicated for how the approximation error grows when the current estimate moves to realistic attenuation values, which directly affects whether the efficiency claim and the established convergence theory remain valid throughout the iteration.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the thorough review and valuable feedback on our manuscript. The comment raises an important point regarding the Jacobian approximation, which we address directly below.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central algorithmic and theoretical claims rest on approximating the Jacobian at special points such as the zero vector and then using the resulting block structure for the full nonlinear reconstruction. No explicit error bound or analysis is indicated for how the approximation error grows when the current estimate moves to realistic attenuation values, which directly affects whether the efficiency claim and the established convergence theory remain valid throughout the iteration.

    Authors: We agree that the manuscript would benefit from an explicit discussion or bound on the approximation error as iterates move away from the special points (e.g., the zero vector). The convergence theory is established under conditions ensuring the approximated Jacobian preserves key structural properties (such as the block-diagonal form derived from aggregated spectra) that support the algorithm's efficiency and convergence. Numerical results on both noiseless and noisy data with realistic attenuation values demonstrate stable convergence and superior performance, indicating the approximation remains effective in practice. In the revised manuscript, we will add a dedicated subsection analyzing the error growth (theoretically where possible and empirically via additional plots of approximation error versus iterate distance) to clarify the range of validity. revision: yes

Circularity Check

0 steps flagged

No circularity: derivation introduces new approximation and convergence theory without reduction to fitted inputs or self-citations

full rationale

The paper proposes aggregated energy spectra to approximate the Jacobian of the nonlinear forward operator at special points (e.g., zero) as a block product structure, then derives an algorithm and convergence theory from this. No quoted step reduces by construction to a fitted parameter renamed as prediction, a self-defined quantity, or a load-bearing self-citation chain. The approximation is presented as novel, the theory is conditional on appropriate assumptions, and validation uses independent numerical experiments on noiseless/noisy data. The central claims remain self-contained against external benchmarks rather than tautological.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review provides insufficient detail to enumerate specific free parameters or axioms; the approach implicitly relies on the validity of the forward operator approximation and choice of special points for Jacobian evaluation.

pith-pipeline@v0.9.1-grok · 5708 in / 1040 out tokens · 19105 ms · 2026-06-26T14:36:36.284483+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

47 extracted references

  1. [1]

    Adler, H

    J. Adler, H. Kohr, A. Ringh, J. Moosmann, B. Seibaneert, M. J. Ehrhardt, G. R. Lee, K. Niinimaki, B. Gris, O. Verdier, J. Karlsson, G. Zickert, W. J. Palenstijn, O. ¨Oktem, C. Chen, H. A. Loarca, and M. Lohmann. odlgroup/odl: ODL 0.7.0, 2018

    2018

  2. [2]

    Alvarez and E

    R. Alvarez and E. Seppi. A comparison of noise and dose in conven- tional and energy selective computed tomography.IEEE Trans. Nucl. Sci., 26(2):2853–2856, 1979

    1979

  3. [3]

    R. E. Alvarez and A. Macovski. Energy-selective reconstructions in X-ray computerised tomography.Phys. Med. Biol., 21(5):733, 1976

    1976

  4. [4]

    G. Bal, R. Gong, and F. Terzioglu. An inversion algorithm for P-functions with applications to multi-energy CT.Inverse Probl., 38(3):035011, 2022

    2022

  5. [5]

    R. F. Barber, E. Y. Sidky, T. G. Schmidt, and X. Pan. An algorithm for constrained one-step inversion of spectral CT data.Phys. Med. Biol., 61(10):3784, 2016

    2016

  6. [6]

    Bilic, P

    P. Bilic, P. Christ, H. B. Li, E. Vorontsov, A. Ben-Cohen, G. Kaissis, et al. The liver tumor segmentation benchmark (LiTS).Med. Image Anal., 84:102680, 2023

    2023

  7. [7]

    C. Cai, T. Rodet, S. Legoupil, and A. Mohammad-Djafari. A full- spectral Bayesian reconstruction approach based on the material decom- position model applied in dual-energy computed tomography.Med. Phys., 40(11):111916, 2013

    2013

  8. [8]

    Chang, C

    S. Chang, C. Zhang, X. Mou, Q. Xu, L. He, and X. Chen. A parametric physical model-based X-ray spectrum estimation approach for CT imaging. IEEE Trans. Radiat. Plasma Med. Sci., 8(5):532–539, 2024. 6 Discussion and conclusion25

    2024

  9. [9]

    B. Chen, Z. Zhang, E. Y. Sidky, D. Xia, and X. Pan. Image reconstruc- tion and scan configurations enabled by optimization-based algorithms in multispectral CT.Phys. Med. Biol., 62(22):8763, 2017

    2017

  10. [10]

    B. Chen, Z. Zhang, D. Xia, E. Y. Sidky, and X. Pan. Non-convex primal- dual algorithm for image reconstruction in spectral CT.Comput. Med. Imaging Graph., 87:101821, 2021

    2021

  11. [11]

    C. Chen, R. Wang, C. Bajaj, and O. ¨Oktem. An efficient algorithm to compute the X-ray transform.Int. J. Comput. Math., 99(7):1325–1343, 2022

    2022

  12. [12]

    I. A. Elbakri and J. A. Fessler. Statistical image reconstruction for polyen- ergetic X-ray computed tomography.IEEE Trans. Med. Imaging, 21(2):89– 99, 2002

    2002

  13. [13]

    J. G. Fletcher, A. Inoue, A. Bratt, K. K. Horst, C. W. Koo, P. S. Rajiah, F. I. Baffour, J. P. Ko, M. Remy-Jardin, C. H. McCollough, and L. Yu. Photon-counting CT in thoracic imaging: Early clinical evidence and in- corporation into clinical practice.Radiology, 310(3):e231986, 2024

    2024

  14. [14]

    Gao and C

    Y. Gao and C. Chen. Convergence analysis of the nonlinear Kaczmarz method for systems of nonlinear equations with componentwise convex mappings and applications to image reconstruction in multispectral CT. SIAM J. Imaging Sci., 18(1):120–151, 2025

    2025

  15. [15]

    Gao and C

    Y. Gao and C. Chen. AFIRE: Accurate and fast image reconstruction algorithm for geometric-inconsistent multispectral CT.SIAM J. Imaging Sci., to appear, 2026

    2026

  16. [16]

    Y. Gao, X. Pan, and C. Chen. An extended primal-dual algorithm frame- work for nonconvex problems: application to image reconstruction in spec- tral CT.Inverse Probl., 38(8):085011, 2022

    2022

  17. [17]

    Greffier, N

    J. Greffier, N. Villani, D. Defez, D. Dabli, and S. Si-Mohamed. Spectral CT imaging: Technical principles of dual-energy CT and multi-energy photon- counting CT.Diagn. Interv. Imaging, 104(4):167–177, 2023

    2023

  18. [18]

    Guggenberger, S

    R. Guggenberger, S. Winklhofer, G. Osterhoff, G. A. Wanner, M. Fortunati, G. Andreisek, H. Alkadhi, and Stolzmann P. Metallic artefact reduction with monoenergetic dual-energy CT: systematic ex vivo evaluation of pos- terior spinal fusion implants from various vendors and different spine levels. Eur. Radiol., 22(11):2357–2364, 2012

    2012

  19. [19]

    W. Ha, E. Y. Sidky, R. F. Barber, T. G. Schmidt, and X. Pan. Estimating the spectrum in computed tomography via Kullback–Leibler divergence constrained optimization.Med. Phys., 46(1):81–92, 2019

    2019

  20. [20]

    G. N. Hounsfield. Computerized transverse axial scanning (tomography): Part I. Description of system.Br. J. Radiol., 68(815):H166–72, 1973. 6 Discussion and conclusion26

    1973

  21. [21]

    Hsieh.Computed Tomography: Principles, Design, Artifacts, and Recent Advances

    J. Hsieh.Computed Tomography: Principles, Design, Artifacts, and Recent Advances. SPIE Press, Bellingham, Washington, 4 edition, 2022

    2022

  22. [22]

    J. H. Hubbell and S. M. Seltzer. Tables of X-ray mass attenuation coef- ficients and mass energy-absorption coefficients 1 keV to 20 MeV for el- ements Z = 1 to 92 and 48 additional substances of dosimetric interest. Technical report, National Institute of Standards and Technology (NIST), Gaithersburg, MD, United States, 1995

    1995

  23. [23]

    D. Lee, J. Lee, J. Min, B. Lee, B. Lee, K. Oh, J. Kim, and S. Cho. Efficient material decomposition method for dual-energy X-ray cargo inspection sys- tem.Nucl. Instrum. Methods Phys. Res. A, 884:105–112, 2018

    2018

  24. [24]

    Leinweber, J

    C. Leinweber, J. Maier, and M. Kachelrieß. X-ray spectrum estimation for accurate attenuation simulation.Med. Phys., 44(12):6183–6194, 2017

    2017

  25. [25]

    Long and J

    Y. Long and J. A. Fessler. Multi-material decomposition using statisti- cal image reconstruction for spectral CT.IEEE Trans. Med. Imaging, 33(8):1614–1626, 2014

    2014

  26. [26]

    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.Med. Phys., 36(8):3818–3829, 2009

    2009

  27. [27]

    C. Maaß, E. Meyer, and M. Kachelrieß. Exact dual energy material de- composition from inconsistent rays (MDIR).Med. Phys., 38(2):691–700, 2011

    2011

  28. [28]

    Macovski, R

    A. Macovski, R. E. Alvarez, J. L.-H. Chan, J. P. Stonestrom, and L. M. Zatz. Energy dependent reconstruction in X-ray computerized tomography. Comput. Biol. Med., 6(4):325–336, 1976

    1976

  29. [29]

    C. H. McCollough, S. Leng, L. Yu, and J. G. Fletcher. Dual- and multi- energy CT: Principles, technical approaches, and clinical applications.Ra- diology, 276(3):637–653, 2015

    2015

  30. [30]

    Mechlem, S

    K. Mechlem, S. Ehn, T. Sellerer, E. Braig, D. M¨ unzel, F. Pfeiffer, and P. B. No¨ el. Joint statistical iterative material image reconstruction for spectral computed tomography using a semi-empirical forward model.IEEE Trans. Med. Imaging, 37(1):68–80, 2018

    2018

  31. [31]

    C. Mory, B. Sixou, S. Si-Mohamed, L. Boussel, and S. Rit. Comparison of five one-step reconstruction algorithms for spectral CT.Phys. Med. Biol., 63(23):235001, 2018

    2018

  32. [32]

    Natterer.The Mathematics of Computerized Tomography

    F. Natterer.The Mathematics of Computerized Tomography. Society for Industrial and Applied Mathematics, Philadelphia, PA, 2001

    2001

  33. [33]

    T. Niu, X. Dong, M. Petrongolo, and L. Zhu. Iterative image-domain decomposition for dual-energy CT.Med. Phys., 41(4):041901, 2014. 6 Discussion and conclusion27

    2014

  34. [34]

    Patino, A

    M. Patino, A. Prochowski, M. D. Agrawal, F. J. Simeone, R. Gupta, P. F. Hahn, and D. V. Sahani. Material separation using dual-energy CT: Cur- rent and emerging applications.RadioGraphics, 36(4):1087–1105, 2016

    2016

  35. [35]

    Philippe

    D. Philippe. SpectrumGUI.https://sourceforge.net/projects/ spectrumgui/, 2015

    2015

  36. [36]

    Y. Ren, H. Xie, W. Long, X. Yang, and X. Tang. On the conditioning of spectral channelization (energy binning) and its impact on multi-material decomposition based spectral imaging in photon-counting CT.IEEE Trans. Biomed. Eng., 68(9):2678–2688, 2021

    2021

  37. [37]

    Sahbaee, C

    P. Sahbaee, C. Dunning, A. Primak, E. Shanblatt, K. Hallam, C. Choi, G. S. K. Fung, L. Severance, T. O’Donnell, J. O’Doherty, and J. C. Ramirez-Giraldo. Toward broader clinical adoption of photon-counting CT: Where we stand according to high-impact literature.Radiology, 319(1):e251496, 2026

    2026

  38. [38]

    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 Trans. Med. Imaging, 36(9):1808–1819, 2017

    2017

  39. [39]

    L. Shen, Y. Xing, and L. Zhang. Joint reconstruction and spectrum refine- ment for photon-counting-detector spectral CT.IEEE Trans. Med. Imag- ing, 42(9):2653–2665, 2023

    2023

  40. [40]

    L. A. Shepp and J. B. Kruskal. Computerized tomography: The new med- ical X-ray technology.Am. Math. Mon., 85(6):420–439, 1978

    1978

  41. [41]

    S. A. Si-Mohamed, J. Miailhes, P. Rodesch, S. Boccalini, H. Lacombe, V. Leitman, V. Cottin, L. Boussel, and P. Douek. Spectral photon-counting CT technology in chest imaging.J. Clin. Med., 10(24):5757, 2021

    2021

  42. [42]

    X-map 2.0

    K. Taguchi, T. Itoh, M. K. Fuld, E. Fournie, O. Lee, and K. Noguchi. “X-map 2.0” and “iBHC2” for consistent edema signal enhancement for acute ischemic stroke using non-contrast-enhanced dual-energy CT. In5th Meeting, Salt Lake City, 2018, pages 86–89, N¨ urnberg, 2018. Society of High Performance Computational Imaging (SHPCI) e.V

    2018

  43. [43]

    K. J. Treb, J. F. Marsh, E. Shanblatt, T. Nowak, C. A. S. Dunning, G. Fung, B. Schmidt, C. H. McCollough, and S. Leng. Technical perfor- mance of dual-contrast-agent K-edge imaging with four energy thresholds on a commercial photon counting detector CT.Med. Phys., 52(10):e70063, 2025

    2025

  44. [44]

    X. Xu, Y. Xing, S. Wang, and L. Zhang. Systematic implementation of spectral CT with a photon counting detector for liquid security inspection. Nucl. Instrum. Methods Phys. Res. A, 893:99–108, 2018. 6 Discussion and conclusion28

    2018

  45. [45]

    Zhang, J.-B

    R. Zhang, J.-B. Thibault, C. A. Bouman, K. D. Sauer, and J. Hsieh. Model-based iterative reconstruction for dual-energy X-ray CT using a joint quadratic likelihood model.IEEE Trans. Med. Imaging, 33(1):117–134, 2014

    2014

  46. [46]

    W. Zhao, K. Niu, S. Schafer, and K. Royalty. An indirect transmission measurement-based spectrum estimation method for computed tomogra- phy.Phys. Med. Biol., 60(1):339, 2014

    2014

  47. [47]

    Y. Zhao, X. Zhao, and P. Zhang. An extended algebraic reconstruction technique (E-ART) for dual spectral CT.IEEE Trans. Med. Imaging, 34(3):761–768, 2015

    2015