Pith. sign in

REVIEW 3 major objections 4 minor 4 references

Universal Phase Contrast in Micro-CT Systems

T0 review · 3 major / 4 minor · reviewed 2026-08-03 · deepseek-v4-flash

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

Pith's one-line read Micro-CT systems with finite source and detector blur can operate in a phase-transfer regime where propagation sharpens images beyond the conventional blur limit, even when no Fresnel fringes are visible, and the detector itself can perform

desk verdict Phase-transfer micro-CT reframing with a strong dithering experiment, but Eq. (6)'s Gaussian assumption is load-bearing and the paper itself concedes it doesn't hold. read the letter →

arxiv 2607.29478 v1 pith:LPPG3WFJ submitted 2026-07-31 physics.optics physics.app-phphysics.ins-det

Grammatiki Lioliou , Alberto Astolfo , Marco Endrizzi , David Bate , Charlotte K. Hagen , Alessandro Olivo This is my paper · ORCID
classification physics.opticsphysics.app-phphysics.ins-det
keywords phasecontrastmicro-CTpropagation-basedimagingretrievaldetector-inducedspatialresolutionhardware-inducedFresnelfringes
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 argues that conventional high-resolution micro-CT is rarely purely absorption-based: whenever propagation distance and phase shifts are large enough relative to the system's source and detector blur, free-space propagation actively sharpens the image, even if no Fresnel fringes are visible. In that phase-transfer regime, the detector's blur can act as a built-in phase-retrieval filter—called detector-induced phase retrieval (DIPR)—which removes fringes and suppresses high-frequency noise while keeping the resolution gain. The authors derive analytic conditions for this regime, introduce frequency-domain metrics to identify it, and validate it numerically and experimentally on a custom and a commercial micro-CT scanner. If correct, this changes how spatial resolution and image optimization in micro-CT are interpreted, since routine scans may already be benefiting from phase contrast without showing any fringes.

What carries the argument

The load-bearing identity is the Gaussian-equivalent effective system width, σ_eff² = σ_s_eff² + σ_det_eff² − 2a², where 2a² is the negative variance contributed by free-space propagation (a² = γR′λ/4π). This expresses propagation as a sharpening that subtracts from the blur variances, predicting finer resolution whenever 2a² is non-negligible relative to σ_sys². The DIPR condition is T_prop H_det ≈ 1: the detector MTF H_det should match the inverse of the propagation transfer function—the standard single-distance phase-retrieval filter—over the usable spatial-frequency band, so the detector performs phase retrieval at acquisition. The paper also introduces two metrics: n_φ, the fraction of

What would settle it

Measure a sharp edge with a detector whose PSF is strongly non-Gaussian (Lorentzian-dominated) across a range of propagation distances, and compare the measured edge widths to Eq. (6). If the resolution minimum occurs at a different position, or the improvement is far smaller than predicted, the Gaussian cancellation that defines the DIPR resolution advantage is the point of failure. Conversely, a system with a nearly Gaussian detector PSF should reproduce the predicted resolution gain.

Watch

Extended reading notes

Core claim

The central claim is that micro-CT systems with non-negligible propagation-induced phase transfer operate in a phase-transfer regime (Regime B) in which propagation modifies the effective system response and can provide spatial resolution finer than that predicted from conventional source and detector blur alone, irrespective of whether Fresnel fringes remain visible. Within Regime B, hardware-induced phase retrieval (HIPR) occurs in degrees: under-HIPR leaves residual edge enhancement, matched-HIPR compensates propagation exactly, and over-HIPR suppresses high frequencies. Detector-induced phase retrieval (DIPR) is the matched case in which the detector's modulation transfer function approx

Load-bearing premise

The central calculation assumes the detector point-spread function can be treated as Gaussian when subtracting the propagation term in quadrature (σ_eff² = σ_s_eff² + σ_det_eff² − 2a²); real detector PSFs have significant Lorentzian tails, and the paper itself notes the Gaussian variance rule is strictly no longer applicable for such PSFs.

Editorial extensions

If this is right

  • Absence of visible Fresnel fringes cannot be taken as evidence that image formation is purely absorption-based; it may instead indicate matched hardware phase retrieval.
  • The source-to-sample distance becomes a tunable parameter for jointly optimizing phase transfer, hardware retrieval, and field of view, and the optimum is not where geometric magnification is highest.
  • System optimization should allocate as much of the required low-pass filtering to the detector as possible (DIPR) rather than to source blur, to preserve the phase-induced resolution gain and gain noise suppression.
  • Spatial resolution estimates based on source and detector blur alone will underestimate performance in Regime B, where detector sampling (and sub-pixel dithering) often becomes the limiting factor.
  • Software phase retrieval is only strictly needed in the under-HIPR sub-regime; in matched or over-HIPR, the hardware already provides the retrieval during acquisition.

Reading between the lines

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

  • If the Gaussian variance-subtraction argument is weakened by non-Gaussian detector PSFs, the quantitative resolution gain may be smaller than Eq. (6) predicts, but the qualitative existence of a phase-transfer regime would likely survive; a Lorentzian-tail-aware analysis would put numbers on this.
  • The framework suggests that many existing micro-CT datasets acquired with small focal spots and moderate propagation distances may already contain phase-induced resolution enhancement, implying that retrospective resolution estimates in the literature could be revisited.
  • A direct testable extension: scanning a sharp-edged phantom at fixed source-detector distance while sweeping source-to-sample distance should show a dip in measured edge width near the E_match minimum, exactly as the paper's experiments begin to show.
  • The DIPR concept could extend to other in-line imaging modalities with pixelated detectors, such as visible-light or electron microscopy, where a similar detector-blur-matches-propagation condition might explain resolution beyond conventional PSF predictions.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. This paper argues that propagation-based phase contrast is present and functionally important in many micro-CT systems even when Fresnel fringes are not visible. The authors classify imaging into an absorption-only Regime A and a phase-transfer Regime B, introduce hardware-induced phase retrieval (HIPR) categories including detector-induced phase retrieval (DIPR), and derive Gaussian-based analytical conditions, frequency-domain metrics nφ and Ematch, and an effective-width formula. They support the framework with simulations, custom-system scans over varying geometry, and a commercial Nikon scanner with half-pixel dithering, reporting spatial resolution finer than the conventional source/detector blur prediction.

Significance. If established, the central claim would be important: it would change the standard interpretation of image formation and spatial resolution in laboratory micro-CT, where 'no visible fringes' is often equated with attenuation-only imaging. The experimental work contains genuinely strong elements: the Nikon dithering experiment shows measured resolution (28.5 μm) below the conventional blur prediction (55.0 μm), with δ=0 simulations indicating the effect is phase-related; the PTFE fringe-amplitude measurements across geometries agree with simulations. However, the quantitative theory currently rests on a Gaussian variance-subtraction step that the authors themselves concede is not applicable to their measured mixed Gaussian-Lorentzian detector PSFs. The load-bearing quantitative predictions therefore need to be re-derived, replaced, or carefully bounded before the central claims can be accepted at the stated strength.

major comments (3)
  1. [Methods, Eq. (6); Results (intrinsic resolutions)] Equation (6), σ_eff² = σ_s_eff² + σ_det_eff² − 2a², is obtained by combining the quadratic propagation transfer T_prop(f)=1+a²(2πf)² with Gaussian MTFs. Even for perfectly Gaussian PSFs this is a small-frequency expansion, not an exact identity. More importantly, it is undefined for the measured detector PSFs: Methods reports the custom detector as 67.9% Lorentzian (50 μm HWHM) plus 32.1% Gaussian, and the Nikon detector as 40% Lorentzian (120 μm HWHM) plus 60% Gaussian; a Lorentzian MTF has an infinite second moment, so variance subtraction has no well-defined meaning. The paper explicitly concedes in Methods that Gaussian σ is 'strictly, no longer applicable' in this case. Yet Eq. (6) is used to predict the Regime A/B transition, the intrinsic DIPR resolutions (5.9 μm custom, 14.2 μm Nikon), and the phase-induced reduction in effective system width. Please replace Eq. (6) with a calcul
  2. [Methods, Eq. (14); Results, Fig. 1(c)] DIPR is defined as the condition H_det ≈ H_Pag, and Ematch is defined as the RMS difference between H_det and H_Pag over the usable frequency band. Finding that Ematch has a minimum, and calling that geometry the DIPR optimum, is therefore partly a restatement of the metric's definition. The independent experimental evidence (especially the dithering scan) is valuable, but the theoretical 'optimality' claim needs a separate argument that the minimum of this particular RMS metric is the optimum operating point in a resolution/SNR tradeoff sense. This is especially important because H_Pag contains the sample-dependent propagation parameter a², and the integration band f1..fmax is defined system-by-system.
  3. [Results, Scan 3 and Scan II; Table I] Several of the measured resolutions coincide exactly with the sampling limit, so they are upper bounds rather than direct measurements of the intrinsic phase-enhanced width. Scan 3's 'improvement from 100 μm to 54.4 μm FWHM' equals the two-pixel sample-plane scale at that geometry (27.2 μm pixel), and Scan II's 28.5 μm is exactly the dithered sampling limit at 14.3 μm pixels. The comparison with the conventional blur prediction (99.9 μm and 55.0 μm, respectively) is still informative, but the text should state that the measured LSF is sampling-limited at ≤54.4/≤28.5 μm, not that the intrinsic DIPR resolution was directly measured at those values. Please report fitted LSF FWHM values with uncertainties and show that the edge profiles are not truncated by the sampling grid.
minor comments (4)
  1. [Methods, Eq. (8)] Equation (8) appears to contain a typo: the detector contribution is written as a factor (−2π²|f|²(...)) rather than as an exponential. It should presumably be exp(−2π²|f|²(1/M)²σ_det²).
  2. [Results, first subsection] Typo: 'freqiencies' should be 'frequencies'.
  3. [Methods, Eq. (11) vs Eq. (2)] The forward model in Eq. (2) uses the transport-of-intensity transfer function T_prop(f)=1+a²(2π|f|)², while the phase-transfer metric in Eq. (11) uses sin(πλR′f²), a different Fresnel weak-phase form. Please clarify which model is used for which prediction and confirm that the difference does not affect the comparison with the Fresnel-Kirchhoff simulations.
  4. [Abstract and Discussion] The word 'universal' in the title/abstract is stronger than what two systems can establish. Consider tempering to 'common' or 'prevalent' unless a broader survey is provided.

Circularity Check

1 steps flagged · score 2.0 of 10

Minor definitional circularity in the Ematch/DIPR optimum; the central phase-enhanced resolution claim is independently supported by experiments.

  1. self definitional [Results, Fig. 1(c); Methods, Eq. (14)]
    "Figure 1 (c) presents the detector compatibility metric Ematch, Eq. (14), which quantifies the mismatch between the detector MTF and the Paganin filter over the usable spatial-frequency range. ... The minimum of Ematch defines the optimum DIPR geometry. ... It quantifies the similarity between the detector’s MTF and the Paganin filter using the Root Mean Square (RMS) match: Ematch(r1)= sqrt(∫[Hdet(f)−HPag(f)]^2 df/(fmax−f1))."

    Ematch is constructed as the RMS difference between Hdet and HPag, while DIPR is defined as the case in which the detector approximates the Paganin filter (Hdet≈HPag). Saying that the minimum of Ematch 'defines the optimum DIPR geometry' is therefore a restatement of the definition: the optimum is, by construction, wherever this mismatch metric is smallest. It is not an independent physical prediction derived from image-formation theory. However, the paper's central resolution-enhancement claim does not reduce to this; it is supported by the fringe-amplitude measurements and the Nikon dithering experiment, which do not rely on Ematch.

full rationale

The main derivation is a forward model: Eqs. (1)-(6) follow from the transport-of-intensity equation and Gaussian-blur assumptions, and no fitted parameter is later renamed as a prediction. The Regime A/B classification is a threshold applied to the same model, and the experimental validations (PTFE fringe amplitudes, SNR, dithering) use independently measured PSFs and forward wave-optics simulations, so they can in principle falsify the model. The only clear definitional element is the Ematch/DIPR optimum: since Ematch is defined as RMS(Hdet−HPag), its minimum is tautologically the 'matched' DIPR geometry. This does not invalidate the physical claim that detector-side low-pass filtering can preserve phase-induced sharpening while reducing noise, which is argued separately and supported by the dithering result. Self-citations (e.g., ref. 40 for δ/β of the biological sample) are used for material parameters rather than to justify the central derivation. The non-Gaussian PSF caveat concerning Eq. (6) is a correctness risk, not a circularity, and therefore does not raise the score beyond a minor definitional issue.

Assumptions & free parameters 7 free parameters · 5 assumptions · 0 invented entities

The paper's contributions rest on standard in-line phase-contrast theory (Gureyev et al.) rather than new physical entities. Its quantitative predictions depend on measured/estimated PSF parameters and a fitted monoenergetic energy; the Gaussian-variance cancellation is the most fragile modeling choice.

free parameters (7)
  • Custom source FWHM (small focus) = 12 +/- 3 um
    Estimated focal spot size used in simulations and theoretical metrics; not independently measured in this work.
  • Custom source FWHM (large focus) = 63 +/- 7 um
    Used for large-focus mode calculations in Supplementary Figures.
  • Nikon source FWHM = 17 um
    Estimated Gaussian focal spot from Supplementary Fig. 10; input to system-blur and DIPR predictions.
  • Custom detector PSF parameters = Lorentzian fraction 67.9%, HWHM 50 um; Gaussian sigma 130 um
    Measured mixed Gaussian-Lorentzian model; determines E_match and resolution predictions.
  • Nikon detector PSF parameters = Lorentzian fraction 40%, HWHM 120 um; Gaussian sigma 286 um
    Measured model; determines E_match and resolution predictions.
  • Simulation monoenergetic energies = 17.6 keV (custom), 25.0 keV (Nikon)
    Selected to reproduce experimentally measured PTFE attenuation, i.e., fitted to the validation data.
  • Biological sample delta/beta = ~1000
    Taken from previous work (ref 40) at 60 kVp; used in the qualitative interpretation of the biological sample, not in the main PTFE validation.
assumptions (5)
  • domain assumption Near-field TIE with Laplacian phase-to-intensity transfer (monomorphous object, slowly varying phase).
    Eq. (1) assumes this approximation; it underlies the a^2(2*pi*f)^2 propagation transfer and the variance subtraction in Eq. (6).
  • domain assumption Source and detector blur can be treated as Gaussian for the variance-combination argument.
    Eqs. (3)-(6) combine sigma_s and sigma_det in quadrature and subtract 2a^2; the authors later acknowledge the detector PSF is Lorentzian-dominated and that Eq. (3) is then not strictly valid.
  • domain assumption Source self-noise is negligible; detector PSF smoothing is the only noise-correlating mechanism.
    Invoked in Methods to argue that detector blur, not source blur, gives the SNR benefit; relies on cross-spectral purity.
  • standard math Paganin filter is the appropriate reference for 'matched' retrieval.
    The DIPR criterion is defined against the single-distance Paganin transfer function; this is a conventional, well-established retrieval choice.
  • domain assumption Finite-source coherence criterion Eq. (9) holds for all investigated geometries.
    Used to ensure residual phase transfer remains for detector-based retrieval; stated to be fulfilled but not directly measured per geometry.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Universal Phase Contrast in Micro-CT Systems." pith.science (2026). https://pith.science/paper/LPPG3WFJ

@misc{pith2026260729478,
  author       = {Pith},
  title        = {Pith review of: Universal Phase Contrast in Micro-CT Systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LPPG3WFJ}},
  note         = {Machine review of arXiv:2607.29478}
}
read the original abstract

Conventional high-resolution micro-CT systems are regarded as attenuation-based unless visible Fresnel fringes reveal the presence of propagation-based phase contrast. Here we show that this interpretation is incomplete. When propagation-induced phase transfer is non-negligible relative to the system blur, micro-CT operates in a phase-transfer regime in which propagation improves spatial resolution relative to that expected from source and detector blur alone, irrespective of whether Fresnel fringes remain visible. Because micro-CT systems possess finite source and detector blur, this phase transfer is accompanied by a degree of hardware-induced phase retrieval (HIPR), ranging from under-HIPR to matched-HIPR and over-HIPR. We further identify detector-induced phase retrieval (DIPR) as the optimal case in which the detector provides all of the filtering required for phase retrieval, thereby preserving the phase-induced resolution enhancement while introducing spatial correlations between detected photons that reduce high-spatial-frequency noise. We derive analytical conditions for HIPR and DIPR, introduce frequency-domain metrics quantifying preserved phase transfer and detector compatibility, and validate the framework numerically with simulations and experimentally using custom and commercial micro-CT systems. Our results demonstrate that phase-induced resolution enhancement can occur in conventional micro-CT systems even in the absence of visible Fresnel fringes, changing how image formation, spatial resolution, and optimization should be interpreted.

Figures

Figures reproduced from arXiv: 2607.29478 by the authors.

Figure 1
Figure 1. (a) Finite-source coherence criterion, (b) normalized pre-detector phase-transfer metric, nϕ, and (c) detector compatibility metric, Ematch, as functions of source-to-sample distance, R1, for our system (black solid line) and the Nikon scanner (black dots). Finally, the conventional system blur and its implications for spatial resolution are evaluated. The FWHM corresponding to the equivalent Gaussian system width, … view at source ↗
Figure 5
Figure 5. [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

4 extracted references

  1. [1]

    T., Tajziehchi, G

    1 Ghavami-Lahiji, M., Davalloo, R. T., Tajziehchi, G. & Shams, P. Micro-computed tomography in preventive and restorative dental research: A review. Imaging Sci Dent 51, 341–350 (2021). 2 Vásárhelyi, L., Kónya, Z., Kukovecz, Á. & Vajtai, R. Microcomputed tomography– based characterization of advanced materials: a review. Materials Today Advances 8, 100084...

    2021

  2. [3]

    J Synchrotron Radiat 31, 896–909 (2024)

    Basic theory, numerical simulations and planar experimental images. J Synchrotron Radiat 31, 896–909 (2024). 34 Gureyev, T. E., Paganin, D. M. & Quiney, H. M. Signal -to-noise and spatial resolution in in -line imaging

    2024

  3. [4]

    unreasonable

    Optimization using a simple model. Journal of Synchrotron Radiation 33, 395–408 (2026). 35 Gureyev, T. E., Stevenson, A. W., Nesterets, Y. I. & Wilkins, S. W. Image deblurring by means of defocus. Optics Communications 240, 81–88 (2004). 36 Gureyev, T. E. et al. On the van Cittert-Zernike theorem for intensity correlations and its applications. J. Opt. So...

    2026

  4. [2030]

    Size, Share & Trends Analysis Report By Product (In -vivo, Ex-vivo), By Application (Life Science, Bones, Dentistry, Plants & Food), By End -use, By Region, And Segment Forecasts. (2025). https://www.grandviewresearch.com/industry- analysis/micro-computed-tomography-market-report 5 Velde, G. V. Longitudinal micro-CT provides biomarkers of lung disease tha...

    2025

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.