Quantum CT via Dynamic Interval Encoding and Prior-Balanced QUBO Reconstruction

Ao Wang, Yikuang Yuluo, Yujie Liu, Shuangyang Zhong, Yuwen Zhang, Zihao Wang, Fenglin Liu, Andreas Maier, Haijun Yu, Yixing Huang · 2026

Read Quantum CT via Dynamic Interval Encoding and Prior-Balanced QUBO Reconstruction by Ao Wang, Yikuang Yuluo, Yujie Liu, Shuangyang Zhong, Yuwen Zhang, Zihao Wang, Fenglin Liu, Andreas Maier, Haijun Yu, Yixing Huang (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-24561.txt.

Introduction

Wang and colleagues reconstruct CT images on quantum hardware, spending their binary variable budget only where the image is still uncertain. Quantum computed tomography casts the rebuild as a quadratic binary problem, for quantum annealers and hybrid solvers. Gray images strain the encoding: fixed bit planes balloon the problem as precision rises, while low bit encodings quantise coarsely.

Dynamic interval encoding refines iteratively. Each round encodes only the still-active pixels, within local gray level intervals around the running estimate. A boundary hit guided rule switches between widening the search and refining locally. Before forming the final problem, data agreement is balanced against an edge preserving quadratic prior, for stability. Sparse view and limited angle fan beam experiments show the framework recovering structures and gray level spreads better than the analytic, iterative, variational, and quantum rivals. Wang's 2026 team targets the dose sparing scan shapes where classical reconstruction struggles most.

Catalog record

CREATORAo Wang, Yikuang Yuluo, Yujie Liu, Shuangyang Zhong, Yuwen Zhang, Zihao Wang, Fenglin Liu, Andreas Maier, Haijun Yu, Yixing Huang
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE44 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2606.24561

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Quantum CT via Dynamic Interval Encoding and Prior-Balanced QUBO Reconstruction Ao Wang, Yikuang Yuluo, Yujie Liu, Shuangyang Zhong, Yuwen Zhang, Zihao Wang, Fenglin Liu, Andreas Maier, Haijun Yu, Yixing Huang Abstract Quadratic unconstrained binary optimization (QUBO)-based quantum computed tomography (CT) casts reconstruction as a binary quadratic problem for quantum annealing and hybrid quantum–classical solvers. For grayscale CT, however, image encoding is constrained by the binary-variable budget: fixed global bit-plane encodings increase QUBO size and coupling complexity as gray-level precision improves, whereas low-bit encodings introduce quantization error. We propose a QUBO-based grayscale CT reconstruction framework that combines dynamic interval encoding with prior-balanced optimization. Each refinement round encodes active pixels only within local gray-level intervals around the current estimate, and a boundary-hit-guided update rule adaptively switches between search expansion and local refinement. To improve optimization stability, the method balances projection-domain data consistency and an edge-preserving quadratic prior before forming the final QUBO. Sparse-view and limited-angle fan-beam CT experiments show that the proposed method recovers structures and gray-level distributions more faithfully than the evaluated analytic, iterative, variational, and representation-based baselines. Expressivity analysis and ablation studies further indicate that the improvement mainly arises from effective gray-level representation through dynamic local encoding and more stable data-fidelity–prior coupling. Experiments on the D-Wave hybrid binary quadratic model (BQM) solver further demonstrate that the formulation is executable on a hardware-backed hybrid quantum–classical backend. I Introduction Quadratic unconstrained binary optimization (QUBO)-based quantum computed tomography (CT) casts tomographic reconstruction as a binary quadratic problem for quantum annealing and hybrid quantum–classical solvers [ 17 , 21 ] . In this framework, image intensities are encoded by binary variables, and projection consistency is expressed as a quadratic objective. Unlike continuous-variable reconstruction, QUBO-based reconstruction makes the image representation part of the optimization problem itself: improving gray-level precision usually requires more binary variables and increases the size, coupling density, and coefficient range of the QUBO. Therefore, for grayscale CT, reconstruction quality is limited not only by projection data and regularization, but also by the binary-variable budget available for image encoding. Existing QUBO-based tomographic studies have shown that CT reconstruction can be cast as a binary optimization problem solvable by annealing or hybrid solvers [ 23 , 11 , 16 , 8 , 15 ] . These studies also indicate that image encoding is a key factor governing scalability and reconstruction accuracy. For binary images, low-integer-valued images, or material labels, a small number of candidate states may be sufficient to describe pixel values. Grayscale CT, however, contains continuously varying attenuation coefficients and subtle structural transitions, requiring higher effective gray-level expressivity. A direct fixed global bit-plane encoding increases the number of variables, quadratic couplings, and coefficient dynamic range as the bit depth grows, whereas a low-bit encoding introduces substantial quantization error. We formulate this trade-off as the binary-variable budget problem in grayscale QUBO reconstruction. The binary-variable budget denotes the number of binary variables assigned to image encoding in a single QUBO optimization round. Under a fixed global encoding, this budget is mainly determined by the image size and the number of bits per pixel, so improving gray-level precision typically requires enlarging the QUBO instance directly. The basic idea of this work is to replace a global high-bit representation with a local low-bit representation updated across reconstruction rounds: each round uses only a small number of binary variables to represent local gray-level candidates around the current estimate, and the candidate intervals are iteratively updated to improve effective gray-level resolution. Based on this idea, we propose a QUBO-based grayscale CT reconstruction framework that combines dynamic interval encoding with prior-balanced optimization. Dynamic interval encoding constructs local gray-level candidate sets around the current estimate of each active pixel, allowing each QUBO round to use low-bit variables while improving effective expressivity through interval updates across rounds. A boundary-hit-guided update rule detects insufficient local interval coverage and adaptively switches between search expansion and local refinement. To improve optimization stability, we further combine projection-domain data consistency with an edge-preserving quadratic prior and balance the two objective terms before forming the final QUBO. The main contributions of this work are threefold. First, we identify the binary-variable budget as a representation bottleneck in grayscale QUBO CT reconstruction and introduce dynamic interval encoding to improve effective gray-level expressivity without increasing the per-round bit depth. Second, we develop a boundary-hit-guided interval update mechanism and a prior-balanced QUBO objective, enabling more stable joint optimization of data consistency and an edge-preserving prior. Third, we validate the method on sparse-view and limited-angle fan-beam CT tasks, and further demonstrate in the sparse-view experiments that the formulation can be executed on a D-Wave hybrid binary quadratic model solver. II Related Work II-A CT Reconstruction Algorithms CT reconstruction is commonly based on a forward model that maps an attenuation image to projection data, and existing algorithms can be broadly grouped into analytic, iterative, and prior-driven methods. Analytic methods such as filtered backprojection (FBP) [ 18 ] and the Feldkamp–Davis–Kress (FDK) algorithm [ 9 ] are computationally efficient and remain important baselines when angular sampling is sufficient, but they are prone to artifacts when the data are incomplete or angular coverage is limited. Algebraic and iterative methods, including algebraic reconstruction technique (ART) [ 10 ] , simultaneous ART (SART) [ 2 ] , and conjugate gradient for normal equations (CGN) [ 13 , 3 ] , improve reconstruction stability by repeatedly enforcing projection-domain data consistency and reduce the dependence on analytic inversion conditions. These methods form the conventional algorithmic basis of CT reconstruction and provide classical references for the baselines evaluated in this work. Beyond conventional iterative frameworks, regularized, learning-based, and representation-based methods further improve CT reconstruction by introducing image priors or alternative parameterizations. Total variation (TV) [ 24 ] and compressed-sensing methods [ 7 ] exploit gradient sparsity for edge-preserving reconstruction, leading to representative algorithms such as SART-TV [ 26 ] and TV-regularized primal-dual hybrid gradient (TV-PDHG) [ 4 ] . Deep learning methods learn reconstruction mappings, iterative updates, or implicit priors from training data or individual images, including convolutional neural network (CNN)-based inverse-problem networks [ 14 ] , framelet-based sparse-view CT networks [ 12 ] , learned primal-dual reconstruction [ 1 ] , projection-domain learning for limited-angle CT [ 33 ] , LEARN (learned experts’ assessment-based reconstruction network) [ 5 ] , and deep image prior [ 29 ] . Recent implicit neural, hash-grid, and Gaussian representations further show that the image representation itself can substantially affect reconstruction quality and parameter efficiency [ 27 , 20 , 32 ] . These methods provide an important context for this work: reconstruction performance depends not only on the objective function, but also on how the image is parameterized. Unlike these continuous-parameter representations, this work focuses on binary image representations compatible with QUBO optimization. II-B QUBO-Based Quantum CT Reconstruction QUBO-based quantum CT differs from quantum-state image reconstruction in that it seeks binary optimization solutions that can be decoded into classical CT images, rather than quantum-state outputs with nontrivial pixel-wise readout [ 19 ] . In this class of methods, pixels or image states are encoded by binary variables, and projection-domain errors or related reconstruction objectives are transformed into quadratic forms compatible with quantum annealing, adiabatic computation, or hybrid quantum–classical solvers. Existing studies have demonstrated the

Read the complete text