High-Fidelity ROI CT Reconstruction with Limited Quantum Resources via Hybrid Classical-Quantum Refinement

Hyunju Lee; Jeonghwa Lee; Kyungtaek Jun

arxiv: 2605.29472 · v1 · pith:SUC3H2ECnew · submitted 2026-05-28 · 🪐 quant-ph

High-Fidelity ROI CT Reconstruction with Limited Quantum Resources via Hybrid Classical-Quantum Refinement

Hyunju Lee , Jeonghwa Lee , Kyungtaek Jun This is my paper

Pith reviewed 2026-06-29 07:11 UTC · model grok-4.3

classification 🪐 quant-ph

keywords quantum optimizationCT reconstructionregion of interesthybrid classical-quantumQUBOtomographic reconstructionresidual formulationreduced-angle acquisition

0 comments

The pith

Hybrid classical-quantum CT reconstruction applies quantum optimization only to the region of interest after a coarse global estimate.

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

The paper shows that full-image quantum CT reconstruction is limited by the number of binary variables needed, so a two-stage hybrid method first builds a coarse global image with QTR, QCSTR, FBP or SART and then runs quantum optimization solely on the ROI via a residual projection-image formulation. This shrinks the QUBO size while aiming to keep high fidelity inside the selected region. Experiments on three discrete phantoms indicate that the approach works for moderate-size cases under reduced angles and that the best ROI accuracy for the largest sample comes when a stable classical coarse reconstruction is paired with second-stage quantum refinement. The results position quantum assistance as useful mainly for local refinement rather than global reconstruction.

Core claim

A coarse global reconstruction obtained by QTR, QCSTR, FBP or SART supplies a background estimate; quantum optimization is then applied only to the ROI through a residual projection-image formulation that subtracts the known global contribution, thereby reducing effective QUBO size. On three phantom samples the pipelines QTR/QCSTR+QTR/QCSTR and SART+QTR/QCSTR recover the ROI accurately for moderate cases under reduced-angle acquisition, while for the largest sample the lowest average ROI RMSE and MAE are achieved by SART+QTR/QCSTR, confirming that the practical advantage arises when classical methods stabilize the global background and quantum optimization is reserved for local refinement.

What carries the argument

The residual projection-image formulation that isolates the ROI contribution for quantum optimization after subtracting the coarse global background.

If this is right

Quantum optimization becomes feasible for larger or more complex images by restricting it to the ROI.
SART followed by QTR or QCSTR yields the lowest average ROI reconstruction errors among the tested pipelines.
Reduced-angle acquisition still permits accurate ROI recovery when the hybrid pipeline is used on moderate-size phantoms.
Pure quantum global reconstruction is outperformed by the hybrid approach once object size and complexity increase.

Where Pith is reading between the lines

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

The same residual-formulation idea could be applied to other linear inverse problems where a cheap global approximation is available.
Quantum hardware may function best as a local accelerator rather than a full-image solver in imaging tasks.
Real medical CT data would provide a direct test of whether background accuracy from classical methods remains sufficient outside phantom settings.

Load-bearing premise

The coarse global reconstruction must supply a background estimate accurate enough that residual errors do not propagate significant inaccuracies into the quantum-refined ROI.

What would settle it

A test case in which the coarse global image contains visible background errors yet the ROI quantum stage still produces low RMSE and MAE would falsify the claim that the residual formulation prevents global errors from affecting the refined region.

read the original abstract

Quantum optimization for computed tomography (CT) reconstruction is constrained by the number of binary variables required for image representation, making direct whole-image quantum reconstruction difficult for large or structurally complex objects. We propose a hybrid region-of-interest (ROI) refinement framework in which a coarse global image is first reconstructed by quantum tomographic reconstruction (QTR) and quantum compressed sensing tomographic reconstruction (QCSTR), filtered backprojection (FBP), or simultaneous algebraic reconstruction technique (SART), and quantum optimization is then applied only to the selected ROI through a residual projection-image formulation. This strategy reduces the effective QUBO size while preserving high-fidelity reconstruction in the target region. Experiments on three discrete phantom samples show that both QTR/QCSTR+QTR/QCSTR and SART+QTR/QCSTR achieve accurate ROI reconstruction for moderate-size cases under a reduced-angle setting. For the largest and most complex sample, the quality of the coarse global estimate becomes critical, and the best result is obtained when a stable classical coarse reconstruction is combined with second-stage ROI-only QTR/QCSTR. Among the tested pipelines, SART+QTR/QCSTR achieves the lowest average ROI RMSE and MAE. The results indicate that the practical advantage of quantum-assisted CT reconstruction lies in reserving quantum optimization for local refinement while using classical reconstruction to stabilize the global background.

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.

Desk Editor's Note private letter to a colleague

The hybrid ROI refinement via residual projection-image cuts QUBO size for CT but the claimed error decoupling rests only on three phantoms without propagation bounds.

read the letter

The main takeaway is that this work restricts quantum optimization to an ROI residual after a classical global stage, which shrinks the QUBO and shows decent ROI RMSE/MAE on moderate phantoms under reduced angles. The residual formulation itself is the concrete new piece; it does not appear in the cited prior work and lets them combine QTR/QCSTR with FBP or SART.

What works is the practical framing: they test four pipelines on three discrete phantoms and report that SART plus second-stage QTR/QCSTR gives the lowest average errors when the coarse stage is stable. The experiments are run under realistic limited-angle conditions, and the abstract is clear that the largest sample needs a good classical background.

The soft spot is exactly where the stress-test flags it. The central claim needs the residual operator to isolate ROI variables so global errors do not leak in, yet there is no derivation, sensitivity sweep, or bound on how discretization mismatch or mask incompleteness affects the QUBO. Success is shown only on the three phantoms; the abstract itself notes that the largest case is sensitive to the coarse estimate. No error bars, exclusion rules, or post-hoc checks are described in the supplied text, so the quantitative edge over pure classical ROI methods stays empirical rather than bounded.

This is for groups already running quantum optimization on imaging tasks who want a resource-reduction trick that still needs a solid classical front end. It is not yet a general method. The work shows clear thinking on the hybrid split and honest experimental reporting, so it deserves a serious referee rather than a desk reject.

Referee Report

1 major / 2 minor

Summary. The manuscript introduces a hybrid classical-quantum framework for high-fidelity region-of-interest (ROI) CT reconstruction under limited quantum resources. A coarse global reconstruction is performed using methods such as FBP, SART, QTR, or QCSTR, followed by quantum optimization applied solely to the ROI via a residual projection-image formulation that reduces the QUBO size. Experiments on three discrete phantom samples under reduced-angle settings indicate that the SART+QTR/QCSTR pipeline yields the lowest average ROI RMSE and MAE, with the observation that coarse global estimate quality is critical for larger, more complex samples.

Significance. Should the residual formulation prove robust to global errors, this work could meaningfully advance practical applications of quantum optimization in CT by confining quantum computation to local refinements while relying on classical methods for global stability. The reported empirical results on phantoms provide initial support for moderate cases, highlighting a potential division of labor between classical and quantum stages.

major comments (1)

[residual projection-image formulation (as described in the methods)] The description of the residual projection-image formulation provides no error-propagation analysis, bound, or sensitivity study to demonstrate that global reconstruction errors do not significantly bias the second-stage QUBO solution for the ROI. This is load-bearing for the central claim of effective separation of concerns, especially given the abstract's statement that 'for the largest and most complex sample, the quality of the coarse global estimate becomes critical' and that best results require a stable classical coarse stage.

minor comments (2)

Include quantitative details on error bars, number of runs, and exact parameter choices for the QUBO encoding to strengthen the experimental claims.
Clarify the precise conditions under which the residual operator perfectly decouples global components, including any assumptions on discretization and ROI mask support.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address the single major comment below.

read point-by-point responses

Referee: [residual projection-image formulation (as described in the methods)] The description of the residual projection-image formulation provides no error-propagation analysis, bound, or sensitivity study to demonstrate that global reconstruction errors do not significantly bias the second-stage QUBO solution for the ROI. This is load-bearing for the central claim of effective separation of concerns, especially given the abstract's statement that 'for the largest and most complex sample, the quality of the coarse global estimate becomes critical' and that best results require a stable classical coarse stage.

Authors: We agree that the manuscript currently lacks a formal error-propagation analysis, theoretical bound, or dedicated sensitivity study on how global reconstruction errors propagate into the ROI QUBO. This is a substantive gap for the separation-of-concerns claim. The empirical results on the three phantoms already show that the residual formulation succeeds for moderate-size cases and that, as stated in the abstract, coarse-stage quality becomes critical for the largest sample; however, these observations are not accompanied by a systematic robustness assessment. In the revised manuscript we will add an explicit sensitivity study that intentionally varies global-reconstruction quality (e.g., by injecting controlled noise or substituting weaker classical initializers) and reports the resulting change in ROI RMSE/MAE. This will supply the quantitative evidence requested while preserving the existing empirical findings. revision: yes

Circularity Check

0 steps flagged

No circularity; empirical method validated on phantoms without self-referential derivations

full rationale

The paper presents a hybrid classical-quantum ROI refinement framework for CT reconstruction, first obtaining a coarse global image via FBP, SART, QTR or QCSTR then applying quantum optimization only to the ROI via residual projection-image formulation. All reported results consist of comparative RMSE/MAE metrics on three discrete phantom samples under reduced-angle settings; no equations, uniqueness theorems, or fitted parameters are shown that reduce by construction to the target claims. The central separation-of-concerns argument is supported solely by these external phantom experiments rather than any self-definitional loop or load-bearing self-citation chain.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

The approach rests on domain assumptions about residual accuracy and choices for ROI definition and QUBO encoding that are not derived from first principles within the abstract.

free parameters (2)

ROI selection and size
Choice of which region receives quantum refinement is a modeling decision not fixed by the method itself.
QUBO encoding parameters
Formulation of the binary optimization problem for the residual likely involves tunable or chosen parameters.

axioms (1)

domain assumption Residual between coarse global projections and measurements isolates the ROI error sufficiently for quantum refinement to succeed independently.
Invoked to justify applying quantum optimization only to the ROI after the first stage.

pith-pipeline@v0.9.1-grok · 5773 in / 1261 out tokens · 33650 ms · 2026-06-29T07:11:36.015575+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

13 extracted references · 2 canonical work pages

[1]

Interior region -of-interest reconstruction using a small, nearly complete, truncated dataset,

K. Taguchi, H. Yu, D. Kim, E. Kapadia, O. V. Leach, and G. Wang, “Interior region -of-interest reconstruction using a small, nearly complete, truncated dataset,” Med. Phys., vol. 38, no. 3, pp. 1307–1316, 2011

2011
[2]

The evolution of image reconstruction for CT —from filtered back projection to artificial intelligence,

M. J. Willemink and P. B. Noël, “The evolution of image reconstruction for CT —from filtered back projection to artificial intelligence,” Eur. Radiol., vol. 29, pp. 2185–2195, 2019

2019
[3]

Image reconstruction for sparse-view CT and interior CT— Introduction to compressed sensing and differentiated backprojection,

H. Kudo, “Image reconstruction for sparse-view CT and interior CT— Introduction to compressed sensing and differentiated backprojection,” Quant. Imaging Med. Surg., vol. 3, no. 3, pp. 147–161, 2013

2013
[4]

Limited -angle computed tomography with deep image and physics priors,

S. Barutcu et al., “Limited -angle computed tomography with deep image and physics priors,” Sci. Rep., vol. 11, Art. no. 18794, 2021

2021
[5]

A highly accurate quantum optimization algorithm for CT image reconstruction based on sinogram patterns,

K. Jun, “A highly accurate quantum optimization algorithm for CT image reconstruction based on sinogram patterns,” Sci. Rep., vol. 13, Art. no. 14407, 2023

2023
[6]

Quantum optimization algorithms for CT image segmentation from X-ray data,

K. Jun and H. Lee, “Quantum optimization algorithms for CT image segmentation from X-ray data,” Sci. Rep., vol. 15, Art. no. 20649, 2025

2025
[7]

Quantum -Assisted Tomographic Image Refinement with Limited Qubits for High -Resolution Imaging,

H. Lee and K. Jun, “Quantum -Assisted Tomographic Image Refinement with Limited Qubits for High -Resolution Imaging,” arXiv:2504.20654, 2025

work page arXiv 2025
[8]

Quantum compressed sensing tomographic reconstruction algorithm,

A. Ryou, K. Kim, and K. Jun, “Quantum compressed sensing tomographic reconstruction algorithm,” arXiv:2505.11286, 2025

work page arXiv 2025
[9]

Evidence of scaling advantage for the quantum approximate optimization algorithm on a classically intractable problem,

R. Shaydulin et al., “Evidence of scaling advantage for the quantum approximate optimization algorithm on a classically intractable problem,” Sci. Adv., vol. 10, no. 22, eadm6761, 2024

2024
[10]

Chemistry beyond the scale of exact diagonalization on a quantum -centric supercomputer,

J. Robledo -Moreno et al., “Chemistry beyond the scale of exact diagonalization on a quantum -centric supercomputer,” Sci. Adv., vol. 11, no. 25, adu9991, 2025

2025
[11]

Quantum annealing -assisted lattice optimization,

Z. Xu et al., “Quantum annealing -assisted lattice optimization,” npj Comput. Mater., vol. 11, Art. no. 4, 2025

2025
[12]

Natterer, The Mathematics of Computerized Tomography

F. Natterer, The Mathematics of Computerized Tomography. Philadelphia, PA, USA: SIAM, 2001

2001
[13]

Simultaneous algebraic reconstruction technique (SART): A superior implementation of the ART algorithm,

A. H. Andersen and A. C. Kak, “Simultaneous algebraic reconstruction technique (SART): A superior implementation of the ART algorithm,” Ultrason. Imaging, vol. 6, no. 1, pp. 81–94, 1984

1984

[1] [1]

Interior region -of-interest reconstruction using a small, nearly complete, truncated dataset,

K. Taguchi, H. Yu, D. Kim, E. Kapadia, O. V. Leach, and G. Wang, “Interior region -of-interest reconstruction using a small, nearly complete, truncated dataset,” Med. Phys., vol. 38, no. 3, pp. 1307–1316, 2011

2011

[2] [2]

The evolution of image reconstruction for CT —from filtered back projection to artificial intelligence,

M. J. Willemink and P. B. Noël, “The evolution of image reconstruction for CT —from filtered back projection to artificial intelligence,” Eur. Radiol., vol. 29, pp. 2185–2195, 2019

2019

[3] [3]

Image reconstruction for sparse-view CT and interior CT— Introduction to compressed sensing and differentiated backprojection,

H. Kudo, “Image reconstruction for sparse-view CT and interior CT— Introduction to compressed sensing and differentiated backprojection,” Quant. Imaging Med. Surg., vol. 3, no. 3, pp. 147–161, 2013

2013

[4] [4]

Limited -angle computed tomography with deep image and physics priors,

S. Barutcu et al., “Limited -angle computed tomography with deep image and physics priors,” Sci. Rep., vol. 11, Art. no. 18794, 2021

2021

[5] [5]

A highly accurate quantum optimization algorithm for CT image reconstruction based on sinogram patterns,

K. Jun, “A highly accurate quantum optimization algorithm for CT image reconstruction based on sinogram patterns,” Sci. Rep., vol. 13, Art. no. 14407, 2023

2023

[6] [6]

Quantum optimization algorithms for CT image segmentation from X-ray data,

K. Jun and H. Lee, “Quantum optimization algorithms for CT image segmentation from X-ray data,” Sci. Rep., vol. 15, Art. no. 20649, 2025

2025

[7] [7]

Quantum -Assisted Tomographic Image Refinement with Limited Qubits for High -Resolution Imaging,

H. Lee and K. Jun, “Quantum -Assisted Tomographic Image Refinement with Limited Qubits for High -Resolution Imaging,” arXiv:2504.20654, 2025

work page arXiv 2025

[8] [8]

Quantum compressed sensing tomographic reconstruction algorithm,

A. Ryou, K. Kim, and K. Jun, “Quantum compressed sensing tomographic reconstruction algorithm,” arXiv:2505.11286, 2025

work page arXiv 2025

[9] [9]

Evidence of scaling advantage for the quantum approximate optimization algorithm on a classically intractable problem,

R. Shaydulin et al., “Evidence of scaling advantage for the quantum approximate optimization algorithm on a classically intractable problem,” Sci. Adv., vol. 10, no. 22, eadm6761, 2024

2024

[10] [10]

Chemistry beyond the scale of exact diagonalization on a quantum -centric supercomputer,

J. Robledo -Moreno et al., “Chemistry beyond the scale of exact diagonalization on a quantum -centric supercomputer,” Sci. Adv., vol. 11, no. 25, adu9991, 2025

2025

[11] [11]

Quantum annealing -assisted lattice optimization,

Z. Xu et al., “Quantum annealing -assisted lattice optimization,” npj Comput. Mater., vol. 11, Art. no. 4, 2025

2025

[12] [12]

Natterer, The Mathematics of Computerized Tomography

F. Natterer, The Mathematics of Computerized Tomography. Philadelphia, PA, USA: SIAM, 2001

2001

[13] [13]

Simultaneous algebraic reconstruction technique (SART): A superior implementation of the ART algorithm,

A. H. Andersen and A. C. Kak, “Simultaneous algebraic reconstruction technique (SART): A superior implementation of the ART algorithm,” Ultrason. Imaging, vol. 6, no. 1, pp. 81–94, 1984

1984