Reconstruction-Independent Resolution Limits in Preclinical Cone-Beam Micro-CT: A Closed-Form Analysis
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Physics. The day in general physics: instruments, methods and results across the field.
Kai: We're looking at the paper: "Reconstruction-Independent Resolution Limits in Preclinical Cone-Beam Micro-CT".
Mira: Here's the gist of the paper. Spatial resolution in preclinical cone-beam micro-CT is bounded by two reconstruction-independent limits: a photon-noise limit and an angular-sampling limit The combined limit yields a closedform scandesign optimum Defining the Resolution Bounds The spatial resolution of preclinical cone-beam micro-CT is bounded by two reconstruction-independent limits, a photon-noise limit from the projection-domain signal known exactly, background knownexactly (SKE/BKE) ideal observer with the Rose criterion, and an angularsampling limit from the Crowther criterion The photon-noise bound is a property of the input data and applies to any reconstruction algorithm The angularsampling bound applies to classical analytic and iterative reconstructions (FDK, FBP, SIRT) that do not inject signal priors These two bounds sit inside the broader cascaded-systems framework of medical-physics image quality The matched-filter SKE/BKE bound used here is the linear shift-invariant (LSI), stationary-Gaussian, noanatomicalclutter limit of that framework Characterizing the Limits The photon-noise limit is derived from the projection-domain ideal observer, where the minimum detectable disc diameter dmin is given by a closedform expression involving feature contrast and position The spatial-frequency limit derived from this bound is fnoise = 1/(2dmin) The angularsampling limit, defined by the Crowther criterion, states that to reconstruct an image of radius r with spatial-frequency content up to fmax requires at least Nθ = 2πrfmax angular samples For a fixed number of projections Nθ acquired over an arc ∆θspan, the spatial-frequency limit at radial distance r from the rotation axis is fmax(r) = Nθ/(2∆θspan r) The Effective Resolution Ceiling For a classical reconstruction, the effective spatial-frequency limit at position r is defined as feff(r) = min
fnoise(r), fCrowther(r): This combined limit yields a closedform scandesign optimum The locus where fnoise = fCrowther partitions the field of view into a noise-dominated region and an angular-dominated region Scan Design Trade-offs The authors illustrate how each acquisition parameter reshapes the resolution map through parametric sweeps over exposure (N0) and projection count (Nθ) Increasing Nθ widens the Crowther cone linearly and improves the photon-noise bound as N1/3θ in dmin The optimal allocation ridge N⋆θ scales linearly with FOV radius, meaning larger ROIs demand more views in linear proportion Practical Implications The practical rules derived from the analysis suggest that if photon noise dominates over the region of interest, both N0 and Nθ buy resolution as ∝ D1/3 tot If angular sampling dominates, Nθ is strictly preferred to N0 because it improves both bounds while N0 improves only the noise bound that is not binding The framework allows for a pixel-by-pixel map showing which anatomical region can be resolved based on local contrast The closedform bounds are implemented in an open-source browserbased calculator The paper concludes that the photon-noise component alone is the binding floor that applies regardless of prior strength The spatial resolution in preclinical cone-beam micro-CT is bounded by two reconstruction-independent limits, a photonnoise limit and an angularsampling limit The two combine into a spatially varying resolution surface that depends only on scan parameters and feature contrast, not on the choice of reconstruction Classical FDK and iterative reconstructions operate several times above this bound, leaving headroom for noise-side algorithmic improvement Reconstructions that incorporate signal priors may exceed the Crowther limit at the cost of relying on those priors The photon-noise component alone is the binding floor that applies regardless of prior strength References
1: Falk L.
Kai: Wiegmann and Nancy L.
Mira: Ford Benchmarking open-source FDK against commercial and iterative reconstruction methods for preclinical micro-CT arXiv:2604.23047
physics.med-ph: , 2026
5: Kenneth M.
Kai: Hanson Detectability in computed tomographic images Medical Physics, 6(5):441–451, 1979 doi: 10.1118/1.594534
8: K.
Paper summary: Kai: So, to wrap up this discussion on "Reconstruction-Independent Resolution Limits in Preclinical Cone-Beam Micro-CT: A Closed-Form Analysis", we're talking about those two fundamental limits we discussed.
Mira: The thesis of the paper is that spatial resolution is bounded by a photon-noise limit and an angular-sampling limit, and these combine to create a spatially varying resolution surface based only on scan parameters and feature contrast.
Lev: They show how this combination leads to a closed-form scan-design optimum, which basically gives you the ideal way to acquire data for that specific anatomy.
Kai: The authors are looking at how classical analytic and iterative reconstructions, like FDK or SIRT, operate several times above this bound, meaning there's still room for improvement in the algorithms themselves.
Mira: However, they also mention that if you use reconstructions with signal priors, like neural networks trained on relevant data, those methods can exceed the Crowther limit at the cost of relying on those priors.
Lev: The paper clearly states that even when you inject priors, the photon-noise component remains the binding floor regardless of how strong your prior is.
Kai: So what this means practically is that if you are limited by noise in a certain area, increasing your exposure and projection count both help resolution up to a certain point.
Mira: If angular sampling limits you, then focusing on N theta over the exposure setting N zero is the better way to improve the overall resolution map for that region.
Lev: The paper provides an open-source browser-based calculator for these closed-form bounds so researchers can actually see how it applies to different setups.
Kai: That's what we have here, this analysis of reconstruction-independent limits in cone-beam micro-CT.
Conclusion: Kai: So, we're wrapping up this look at that paper on reconstruction-independent resolution limits in preclinical cone-beam micro-CT.
Mira: The title is pretty direct, focusing on those two limits: photon noise and angular sampling, and how they combine into a closed form.
Lev: It’s about defining the hard boundaries of what you can actually see before you even start building the reconstruction software or running an iterative algorithm.
Kai: Exactly. The authors use this math to show that whatever resolution you get is just constrained by these physical limits, not just by how good your reconstruction technique is.
Mira: That’s the core idea, right? They prove that for classical methods like FDK or FBP, you can't magically get better than this combined bound without changing the scan design itself.
Lev: And what they really show is that there’s a trade-off in how you acquire data—how many projections versus how long your exposure is.
Kai: They map out exactly where those two limits cross each other, which creates this noise-dominated region and an angular-dominated region in the field of view.
Mira: The implication for us is that if we’re seeing blurry spots in a particular anatomical area, it might not be our reconstruction algorithm failing; it might just be because the scan design didn't meet the optimal ratio for that specific spot.
Lev: It means you can use this closed-form analysis to predict exactly what kind of scan you need to get a certain level of resolution for a given ROI, which is super useful when we’re designing new hardware.
Kai: So, essentially, this paper gives us a blueprint for designing better scans by focusing on those physical constraints instead of just tweaking the reconstruction code.
Mira: It’s about understanding the physics before you get bogged down in the heavy math of iterative reconstruction methods.
Lev: And that leads right into how much room we actually have left to improve things once you hit this theoretical floor.
Falk L. Wiegmann, Nancy L. Ford
Department of Physics and Astronomy, The University of British Columbia · Department of Oral Biological and Medical Sciences, The University of British Columbia
physics.med-ph
Submitted: 2026-07-14
Updated: 2026-07-14
Comments: Accompanying open-source interactive web calculator: https://ubc-ford-lab.github.io/ideal_observer_SKE_BKE_model/
Journal ref: Med. Phys. 53(11), e70737 (2026)
DOI: 10.1002/mp.70737
Code: https://github.com/UBC-Ford-lab/ideal_observer_SKE_BKE_model
Project page: https://ubc-ford-lab.github.io/ideal_observer_SKE_BKE_model
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 82/100
The gist: Spatial resolution in preclinical cone-beam micro-CT is bounded by two reconstruction-independent limits…
Terminology
Summary
Spatial resolution in preclinical cone-beam micro-CT is bounded by two reconstruction-independent limits: a photon-noise limit and an angular-sampling limit The combined limit yields a closedform scandesign optimum
Defining the Resolution Bounds
The spatial resolution of preclinical cone-beam micro-CT is bounded by two reconstruction-independent limits, a photon-noise limit from the projection-domain signal known exactly, background knownexactly (SKE/BKE) ideal observer with the Rose criterion, and an angularsampling limit from the Crowther criterion The photon-noise bound is a property of the input data and applies to any reconstruction algorithm The angularsampling bound applies to classical analytic and iterative reconstructions (FDK, FBP, SIRT) that do not inject signal priors These two bounds sit inside the broader cascaded-systems framework of medical-physics image quality The matched-filter SKE/BKE bound used here is the linear shift-invariant (LSI), stationary-Gaussian, noanatomicalclutter limit of that framework
Characterizing the Limits
The photon-noise limit is derived from the projection-domain ideal observer, where the minimum detectable disc diameter dmin is given by a closedform expression involving feature contrast and position The spatial-frequency limit derived from this bound is fnoise = 1/(2dmin) The angularsampling limit, defined by the Crowther criterion, states that to reconstruct an image of radius r with spatial-frequency content up to fmax requires at least Nθ = 2πrfmax angular samples For a fixed number of projections Nθ acquired over an arc ∆θspan, the spatial-frequency limit at radial distance r from the rotation axis is fmax(r) = Nθ/(2∆θspan r)
The Effective Resolution Ceiling
For a classical reconstruction, the effective spatial-frequency limit at position r is defined as feff(r) = min[fnoise(r), fCrowther(r)] This combined limit yields a closedform scandesign optimum The locus where fnoise = fCrowther partitions the field of view into a noise-dominated region and an angular-dominated region
Scan Design Trade-offs
The authors illustrate how each acquisition parameter reshapes the resolution map through parametric sweeps over exposure (N0) and projection count (Nθ) Increasing Nθ widens the Crowther cone linearly and improves the photon-noise bound as N1/3θ in dmin The optimal allocation ridge N⋆θ scales linearly with FOV radius, meaning larger ROIs demand more views in linear proportion
Practical Implications
The practical rules derived from the analysis suggest that if photon noise dominates over the region of interest, both N0 and Nθ buy resolution as ∝ D1/3 tot If angular sampling dominates, Nθ is strictly preferred to N0 because it improves both bounds while N0 improves only the noise bound that is not binding The framework allows for a pixel-by-pixel map showing which anatomical region can be resolved based on local contrast The closedform bounds are implemented in an open-source browserbased calculator The paper concludes that the photon-noise component alone is the binding floor that applies regardless of prior strength
Conclusion
The spatial resolution in preclinical cone-beam micro-CT is bounded by two reconstruction-independent limits, a photonnoise limit and an angularsampling limit The two combine into a spatially varying resolution surface that depends only on scan parameters and feature contrast, not on the choice of reconstruction Classical FDK and iterative reconstructions operate several times above this bound, leaving headroom for noise-side algorithmic improvement Reconstructions that incorporate signal priors may exceed the Crowther limit at the cost of relying on those priors The photon-noise component alone is the binding floor that applies regardless of prior strength
References
[1] Falk L. Wiegmann and Nancy L. p. 0 Ford Benchmarking open-source FDK against commercial and iterative reconstruction methods for preclinical micro-CT arXiv:2604. p. 023047 [physics. p. 0med-ph], 2026
[5] Kenneth M. Hanson Detectability in computed tomographic images Medical Physics, 6(5):441–451, 1979 doi: 10. p. 11118/1.594534
[8] K. Faulkner and B. M. Moores Noise and contrast detection in computed tomography images Physics in Medicine and Biology, 29(4):329–339, 1984 doi: 10. p. 61088/0031-9155/29/4/003
[4] Scott S. Hsieh, Shuai Leng, Lifeng Yu, Nathan R. p. 0 Huber, and Cynthia H McCollough A minimum SNR criterion for computed tomography object detection in the projection domain Medical Physics, 49(8):4988–4998, 2022 doi: 10. p. 01002/mp.15832
[7] J. C. Gore and P. S Tofts Statistical limitations in computed tomography Physics in Medicine and Biology, 23(6):1176–1182, 1978 doi: 10. p. 01088/0031-9155/23/6/013
[9] R. A. Crowther D J DeRosier, and A Klug The reconstruction of a three-dimensional structure from projections and its application to electron microscopy Proceedings of the Royal Society A, 317(1530):319–340, 1970 doi: 10. p. 51098/rspa.1970.0119
[22] Albert Rose Vision: Human and Electronic Plenum Press, New York, 1973 doi: 10.1007/978-1-4684-2037-1
[25] Wim van Aarle Willem Jan Palenstijn Jeroen Cant Eline Janssens Folkert Bleichrodt Andrei Dabravolski Jan De Beenhouwer K Joost Batenburg and Jan Sijbers Fast and flexible X-ray tomography using the ASTRA toolbox Optics Express, 24(22):25129–25147, 2016 doi: 10.1364/OE.24.025129
[3] Harrison H Barrett and Kyle J Myers Foundations of Image Science Wiley, Hoboken, NJ, 2004 ISBN 978-0-471-15300-9
[6] D A Chesler S J Riederer, and N J Pelc Noise due to photon counting statistics in computed x ray tomography Journal of Computer Assisted Tomography, 1(1):64–74, 1977 doi: 10.1097/00004728-197701000-000-
[23] Nancy Lee Ford Xi Ren Luca Egoriti Nolan Esplen Stephanie Radel Brandon Humphries Hui-Wen Koay Thomas Planche Cornelia Hoehr Alexander Gottberg and Magdalena Bazalova-Carter Respiratory-gated microcomputed tomography imaging to measure radiation-induced lung injuries in mice following ultra-high dose-rate and conventional dose-rate radiation therapy Journal of Medical Imaging, 12(1):014002, 2025 doi: 10.1117/1.JMI.12.1.014002
[4] Xi Ren Luca Egoriti Nolan Esplen Stephanie Rädel Brandon Humphries Hui-Wen Koay Thomas Planche Cornelia Hoehr Alexander Gottberg Magdalena Bazalova-Carter Using in vivo respiratory-gated microcomputed tomography imaging to monitor pulmonary side effects in 10 MV FLASH and conventional radiotherapy In Medical Imaging 2023: Biomedical Applications in Molecular, Structural, and Functional Imaging, volume 12468 page 124680W SPIE, 2023 doi: 10.1117/1.JMI.12.2654427
[9] R A Crowther D J DeRosier, and A Klug The reconstruction of a three-dimensional structure from projections and its application to electron microscopy Proceedings of the Royal Society A, 317(1530):319–340, 1970 doi: 10.1098/rspa.1970.0119
[26] Ander Biguri Manjit Dosanjh Steven Hancock and Manuchehr Soleimani TIGRE A MATLAB-GPU toolbox for CBCT image reconstruction Biomedical Physics & Engineering Express, 2(5):055010 doi: 10.1088/2057-1976/2/5/055010
[3] Harrison H Barrett and Kyle J Myers Foundations of Image Science Wiley, Hoboken, NJ, 2004 ISBN 978-0-471-15300-9
[11] Frank Natterer The Mathematics of Computerized Tomography SIAM, Philadelphia, 2001 doi: 10.1137/9780898719284
[24] Xi Ren Luca Egoriti Nolan Esplen Stephanie Rädel Brandon Humphries Hui-Wen Koay Thomas Planche Cornelia Hoehr Alexander Gottberg Magdalena Bazalova-Carter Using in vivo respiratory-gated microcomputed tomography imaging to monitor pulmonary side effects in 10 MV FLASH and conventional radiotherapy In Medical Imaging 2023: Biomedical Applications in Molecular, Structural, and Functional Imaging, volume 12468 page 124680W SPIE, 2023 doi: 10.1117/1.JMI.12.2654427
[3] Harrison H Barrett and Kyle J Myers Foundations of Image Science Wiley, Hoboken, NJ, 2004 ISBN 978-0-471-15300-9
[11] Frank Natterer The Mathematics of Computerized Tomography SIAM, Philadelphia, 2001 doi: 10.1137/9780898719284
[24] Xi Ren Luca Egoriti Nolan Esplen Stephanie Rädel Brandon Humphries Hui-Wen Koay Thomas Planche Cornelia Hoehr Alexander Gottberg Magdalena Bazalova-Carter Using in vivo respiratory-gated microcomputed tomography imaging to monitor pulmonary side effects in 10 MV FLASH and conventional radiotherapy In Medical Imaging 2023: Biomedical Applications in Molecular, Structural, and Functional Imaging, volume 12468 page 124680W SPIE, 2023 doi: 10.1117/1.JMI.12.2654427
[3] Harrison H Barrett and Kyle J Myers Foundations of Image Science Wiley, Hoboken, NJ, 2004 ISBN 978-0-471-15300-9
[11] Frank Natterer The Mathematics of Computerized Tomography SIAM, Philadelphia, 2001 doi: 10.1137/9780898719284
[24] Xi Ren Luca Egoriti Nolan Esplen Stephanie Rädel Brandon Humphries Hui-Wen Koay Thomas Planche Cornelia Hoehr Alexander Gottberg Magdalena Bazalova-Carter Using in vivo respiratory-gated microcomputed tomography imaging to monitor pulmonary side effects in 10 MV FLASH and conventional radiotherapy In Medical Imaging 2023: Biomedical Applications in Molecular, Structural, and Functional Imaging, volume 12468 page 124680W SPIE, 2023 doi: 10.1117/1.JMI.12.2654427
[3] Harrison H Barrett and Kyle J Myers Foundations of Image Science Wiley, Hoboken, NJ, 2004 ISBN 978-0-471-15300-9
[11] Frank Natterer The Mathematics of Computerized Tomography SIAM, Philadelphia, 2001 doi: 10.1137/9780898719284
[24] Xi Ren Luca Egoriti Nolan Esplen Stephanie Rädel Brandon Humphries Hui-Wen Koay Thomas Planche Cornelia Hoehr Alexander Gottberg Magdalena Bazalova-Carter Using in vivo respiratory-gated microcomputed tomography imaging to monitor pulmonary side effects in 10 MV FLASH and conventional radiotherapy In Medical Imaging 2023: Biomedical Applications in Molecular, Structural, and Functional Imaging, volume 12468 page 124680W SPIE, 2023 doi: 10.1117/1.JMI.12.2654427
[3] Harrison H Barrett and Kyle J Myers Foundations of Image Science Wiley, Hoboken, NJ, 2004 ISBN 978-0-471-15300-9
[11] Frank Natterer The Mathematics of Computerized Tomography SIAM, Philadelphia, 2001 doi: 10.1137/9780898719284
[24] Xi Ren Luca Egoriti Nolan Esplen Stephanie Rädel Brandon Humphries Hui-Wen Koay Thomas Planche Cornelia Hoehr Alexander Gottberg Magdalena Bazalova-Carter Using in vivo respiratory-gated microcomputed tomography imaging to monitor pulmonary side effects in 10 MV FLASH and conventional radiotherapy In Medical Imaging 2023: Biomedical Applications in Molecular, Structural, and Functional Imaging, volume 12468 page 124680W SPIE, 2023 doi: 10.1117/1.JMI.12.5]
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
System can determine if a given set of scan parameters (flux N0 and projection count Ntheta) is optimal for a target spatial resolution by comparing it against an analytical rule sets how to split that budget between flux and views.
-
The AI system can predict the worst-case effective resolution limit feff(r) at any point in the reconstructed field of view by calculating min[fnoise(r), fCrowther(r)], allowing for a heuristic for the binding limit of classical reconstructions.
-
The system can guide scan design by identifying crossover contours where fnoise equals fCrowther, indicating which limit dominates at specific radial distances r from the rotation axis, such as
a noise-dominated region (typically near the centre of attenuating phantoms).
-
For a fixed total dose budget Phi, the AI can calculate the optimal allocation ridge Nstar theta proportional to sqrt(N0) for a given target radius r star, providing a precise rule for splitting photon budget between N0 and Ntheta.
-
The system can inform automated ROI selection in reconstructed images by sampling the region-resolved detectability map feff(x, y) at every pixel, answering
can the features that live here be resolved here?
based on local contrast scale Delta C(x; y).
Sources
- Benchmarking Open-Source FDK Against Commercial and Iterative Reconstruction Methods for Preclinical Micro-CBCT
- Resolution-Noise Characteristics of Common FDK Filter Kernels: A Practical Reference for Preclinical Cone-Beam Micro-CT
Related papers
- BART Online Open-Source Sequence Toolbox for Computational MRI
- Constrained estimation of rotational invariants of the cumulant expansion (RICE) for rapid tensor-valued diffusion MRI
- Improved Myelin Water Fraction Estimation and Uncertainty Mapping in Multicomponent T2 Relaxometry via NNLS Bootstrap Aggregation
- Radiation damage to normal mammalian tissue in vivo with laser-driven protons at ultra-high instantaneous dose rate
- Fast ungated five-dimensional cardiac MRI on a 1.5 T MR-linac for MRI-guided radiotherapy
- BEAM3R: Beam's-eye-view architecture with Mamba-3 for implicit dose reconstruction