Pith. sign in

REVIEW 3 major objections 5 minor 26 references

Curvature-adaptive gigapixel microscopy at submicron resolution and centimeter scale

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a single-shot re-imaging microscope with a 6×8 array of focus-adjustable cameras can capture a seamless 630-megapixel image of a 16.3 × 18.8 mm² area at 0.84 µm half-pitch resolution without mechanical scanning…

desk verdict PANORAMA is a genuine engineering step toward single-shot cm-scale submicron imaging with per-camera refocusing, but the 0.84 µm resolution claim outruns the evidence shown. read the letter →

arxiv 2507.09697 v2 pith:LDADGKXF submitted 2025-07-13 physics.optics eess.IV

Xi Yang , Haitao Chen , Lucas Kreiss , Clare B. Cook , Genevieve Kuczewski , Mark Harfouche , Martin O. Bohlen , Roarke Horstmeyer This is my paper · ORCID
classification physics.opticseess.IV
keywords gigapixelmicroscopymulti-cameraarraymicroscopere-imagingopticsfieldcurvaturecompensationsingle-shotimagingcurvedsamplefluorescencespace-bandwidthproduct
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

This paper claims that a microscope built from a large telecentric photolithography lens and a flat 6×8 array of small cameras, each with independently adjustable focus, can photograph a 16.3 × 18.8 mm² area at 0.84 µm half-pitch resolution in a single exposure. The system, called PANORAMA, produces 630-megapixel stitched images without mechanical scanning. By refocusing individual cameras, the effective focal surface can be curved to match either the lens's field curvature or the shape of a non-flat sample. If correct, this removes the usual trade-off between field of view and resolution for centimeter-scale samples and eliminates the need to tile, stitch, or focus-stack for many biological and materials imaging tasks.

What carries the argument

The machinery is a re-imaging architecture. A commercial telecentric photolithography lens with numerical aperture 0.38, designed for 436 nm and operated at 510 nm through a 10 nm bandpass filter, forms a large intermediate image of the sample. A 6×8 array of micro-cameras, each with a 14.64 mm focal-length lens and a monochrome CMOS sensor, relays local patches of that intermediate image onto the flat sensor array. Each micro-camera can be moved axially, which changes the array's effective focal surface from a plane to a curved shell; the shell can be set to cancel the objective's field curvature or to follow a curved specimen. Neighboring fields overlap by about 10–30%, and the overlaps are used to register and stitch the sub-images into one seamless gigapixel composite.

What would settle it

Image a USAF 1951 target across the full field and record the finest group resolved in both center and corner, or measure the modulation transfer function of the primary lens at 510 nm. If G8E6 (roughly 1.1 µm half-pitch) is not clearly resolved everywhere, or if the MTF at 0.84 µm half-pitch stays below the detection contrast threshold, the stated resolution claim is not supported.

Watch

Extended reading notes

Core claim

The paper's central claim is that field curvature—normally a defect—can be turned into a feature: a high-NA telecentric objective forms a curved intermediate image, and a dense array of 48 small cameras is focused patch-by-patch onto that curved surface, so a flat sensor array captures the whole field sharply in one shot. With flat samples, individually setting each camera's focus compensates the lens's field curvature; with curved samples, the same per-camera refocusing matches the sample's own surface. The authors report a 16.31 × 18.84 mm² field, 0.84 µm half-pitch resolution (0.82 µm theoretical at 510 nm), 630 MP per snapshot, and demonstrate the system on a rat brain slice and on a curved onion epidermis in both brightfield and fluorescence. The paper's comparison table places PANORAMA as the only listed system with both single-shot acquisition and curvature adaptability at sub-micrometer resolution.

Load-bearing premise

The claim rests on the primary photolithography lens being diffraction-limited at 510 nm over the full 16.3 × 18.8 mm field; the paper does not supply an MTF or wavefront measurement at that wavelength, and the included USAF target only proves about 1.1 µm half-pitch, coarser than the headline 0.84 µm.

Editorial extensions

If this is right

  • An entire 2 cm² brain section can be captured at cellular resolution in one shot, removing the need for XY scanning in histology-style imaging.
  • Curved or unflattenable specimens, such as tissue on curved substrates, can be imaged sharply in a single exposure in both brightfield and fluorescence, eliminating focus stacks.
  • Fluorescence and brightfield can share the same optics at the same resolution, with only the illumination and emission filter changed.
  • The single-shot, no-moving-parts design increases throughput relative to tiled scanning systems and avoids stitching artifacts from sample drift between tiles.
  • The architecture scales: adding more cameras or larger sensors directly enlarges the field of view at the same resolution.

Reading between the lines

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

  • If the 0.84 µm claim holds, PANORAMA would close the gap between whole-slide scanners (fast but often slower or coarser) and high-NA microscopes (diffraction-limited but tiny FOV), enabling gigapixel histology as a snapshot.
  • The authors do not state it, but the per-camera focus control could be automated and combined with a depth sensor to create a dynamically reconfigurable focal surface for moving or deformable samples.
  • The overlapping fields of view could be leveraged for stereo or light-field depth estimation, extending the system from 2D panoramas to 3D surface measurements without hardware changes.
  • The central resolution number depends on the primary lens; if an independent MTF test shows the lens is not diffraction-limited at 510 nm across the full field, the real half-pitch may be closer to the roughly 1.1 µm demonstrated by the USAF target in the figure, a useful but more modest result.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The manuscript reports PANORAMA, a re-imaging microscope built from a Zeiss S-Planar 436 nm photolithography objective, a large-aperture tube lens, and a 6×8 array of independently focusable micro-cameras. The authors claim single-shot gigapixel imaging over a 16.3×18.8 mm field at 0.84 µm half-pitch resolution, with per-camera refocusing to keep curved or uneven samples in focus without mechanical scanning. Demonstrations include a USAF resolution target, a rat brain slice, and curved onion epidermis in both brightfield and fluorescence modes, with a direct comparison between flat-focus and multi-focus configurations.

Significance. If fully supported, this is a substantial practical advance: it would permit centimeter-scale, submicron lateral resolution in a single exposure and would add a simple per-camera refocusing mechanism for non-flat samples. The engineering description is concrete and the onion experiment is a useful controlled comparison. I also credit the paper for not using fitted parameters in its central performance claims and for reporting system numbers (FOV, overlap, DOF, magnification range) that are in principle checkable. The main weakness is that the headline 0.84 µm half-pitch resolution is not experimentally established by the data shown in the manuscript.

major comments (3)
  1. [Fig. 1(b4–b6) and system characterization paragraph] The claimed 0.84 µm half-pitch resolution is not supported by the USAF target data shown. The finest displayed elements, G8E5 and G8E6, have line widths of approximately 1.23 µm and 1.10 µm (roughly 406 and 456 lp/mm), whereas a 0.84 µm half-pitch corresponds to about 595 lp/mm. The contrast profile in Fig. 1(b6) therefore only demonstrates resolution coarser than the headline value. To support the claim, the authors should show a target with features finer than 0.84 µm (e.g., USAF group 9 elements or an equivalent high-frequency grating), or provide a measured MTF or knife-edge response at 510 nm; otherwise the stated resolution should be revised to the value actually demonstrated.
  2. [Primary lens description (Letter 2, 'we used the Carl Zeiss S-Planar 436 nm photolithography lens...')] The resolution claim assumes that the 436 nm photolithography lens is diffraction-limited at the operating wavelength of 510 nm across the full 16.3×18.8 mm field. The paper introduces a 510/10 nm bandpass filter to reduce chromatic aberration, but it presents no MTF, wavefront, or through-focus measurement at 510 nm for any field position. The Rayleigh limit with NA=0.38 and λ=510 nm is only valid if residual aberrations are negligible over the entire field. Please provide such measurements, or explicitly label the 0.84 µm figure as a design estimate rather than a measured system specification.
  3. [Supplemental document and Data availability statements] The manuscript repeatedly cites Supplement 1 (S1, S2) for 'comprehensive characterization' and for stitching details, but the supplement is not included with the posted preprint, and the data availability statement says the data are not publicly available. I have no way to verify the claimed resolution, the exact stitching procedure, or the stated overlap-based magnification calibration from the material provided. The authors should make the supplement and at least representative raw or unstitched sub-images available for review; if the supplement already exists in the journal version, it should be supplied with any resubmission.
minor comments (5)
  1. [Abstract] The phrase '16.3×18.8 mm2 FOV' uses 'mm2' as if it were a unit of area; it should read '16.3 mm × 18.8 mm FOV'.
  2. [Fig. 4 caption] The caption says 'The insects indicate the regions examined in detail'; this should be 'insets'.
  3. [Letter 3, brightfield demonstration] The text says 'the entire 2 cm2 section was captured'; this is ambiguous and should be 'a 2 cm × 2 cm section' or 'a sample of area 2 cm²'.
  4. [Table 1] The resolution values in Table 1 are listed without specifying whether they are theoretical or experimentally measured for each system; adding a footnote with the measurement method would make the comparison more meaningful.
  5. [Fig. 1(b6)] The contrast profile plot lacks axis scales and a clear indication of which group elements are being traced; adding a spatial-frequency or line-width scale would let the reader check the resolution claim directly.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the headline resolution is a standard Rayleigh calculation and the capability claims are direct measurements.

full rationale

The paper's central numbers—16.3×18.8 mm² FOV, 630 MP images, and 0.84 µm half-pitch resolution—are either directly measured from the assembled system or computed from the stated NA and wavelength via the standard Rayleigh criterion (0.61λ/NA = 0.61×0.510/0.38 ≈ 0.82 µm). No parameter is fitted to the headline output and then renamed as a prediction. The per-camera magnification estimates and Hugin stitching transformations are used only for image composition, not to generate the resolution or FOV claims. Self-citations to M-FAST (Ref. 13) and MCAM work (Refs. 1, 2, 4, 10–12) provide background and incremental context; none is invoked as the proof of a target result. The validation evidence for resolution is a USAF target whose finest displayed elements (G8E5/G8E6) imply roughly 1.1–1.2 µm half-pitch, coarser than the theoretical 0.84 µm. That is a validation gap—an unverified assumption that the photolithography lens is diffraction-limited at 510 nm across the full field—but it is not circular reasoning. The missing supplement and unavailable data are transparency concerns, not circularity. Therefore no circular steps are identified.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

The claimed performance relies on standard diffraction-limit optics and on three domain assumptions: that the 436 nm photolithography lens works at 510 nm, that Hugin stitching corrects focus-induced magnification changes, and that each camera's sub-field is locally flat within the 45 um DOF. No free parameters are fitted to make the central claim work, and no new physical entities are introduced.

assumptions (4)
  • domain assumption The Zeiss S-Planar 436 nm photolithography lens is diffraction-limited at 510 nm across the full 16.3 x 18.8 mm field, with negligible chromatic aberration after bandpass filtering.
    The system operates at 510 nm but uses a lens designed for 436 nm. The letter states a theoretical resolution of 0.82 um from NA but provides no MTF or wavefront data at 510 nm to support performance across the FOV.
  • domain assumption Hugin stitching can correct the magnification and distortion changes introduced by per-camera refocusing without producing visible seams.
    The paper reports magnification differences up to 10% and notes stitching errors become noticeable above 15% (Letter 3), but no quantitative stitching error or seam metrics are given.
  • domain assumption Each micro-camera's portion of the sample can be treated as flat within the 45 um depth of field after refocusing.
    The maximum curvature of 6.6 m^-1 is stated as a consequence of the 45 um DOF, but the derivation is not shown and no validation across sample geometries is provided.
  • standard math Rayleigh resolution criterion: resolution = 0.61 lambda / NA.
    Used to compute the theoretical resolution of 0.82 um at 510 nm and NA 0.38.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Curvature-adaptive gigapixel microscopy at submicron resolution and centimeter scale." pith.science (2026). https://pith.science/paper/LDADGKXF

@misc{pith2026250709697,
  author       = {Pith},
  title        = {Pith review of: Curvature-adaptive gigapixel microscopy at submicron resolution and centimeter scale},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LDADGKXF}},
  note         = {Machine review of arXiv:2507.09697}
}
abstract

Large-area microscopy with submicron resolution is limited by tradeoffs between field of view (FOV), resolution, and imaging speed. Samples are rarely flat across centimeter-scale FOV, which often requires existing solutions to use mechanical scanning to ensure focused capture at reduced throughput. Here, we present PANORAMA, a single-shot, re-imaging microscope that achieves seamless, gigapixel imaging over a 16.3$\times$18.8 $\text{mm}^2$ FOV at 0.84 um resolution without mechanical scanning. By using a telecentric photolithography lens, a large-aperture tube lens, and a flat micro-camera array with adaptive per-camera focus control, PANORAMA maintains submicron focus across flat, curved or uneven samples that span centimeters. This approach improves imaging throughput and adaptability, enabling gigapixel multi-modal microscopy of large flat and non-flat samples in one shot, thus broadening its applications in biomedical and materials imaging.

Figures

Figures reproduced from arXiv: 2507.09697 by the authors.

Figure 1
Figure 1. Optical setup figure and the system resolution performance. (a) Schematic of the PANORAMA. The system is designed to re￾lay and capture high-resolution images of the sample (S) placed at the object plane. Insert: Curvature-adaptive focusing of MCAM lenses by ∆z via per-camera focus control. (b) shows the system resolution and FOV. (b1) Full-field image of a USAF resolution tar￾get, demonstrating the large field of v… view at source ↗
Figure 3
Figure 3. Brightfield image of a rat brain slice acquired using a 48-camera reimaging system in one shot. Insets and zoom￾ins show cellular-level structures such as dendrites (top right, yellow boxes) and hippocampal morphology (bottom right, pink boxes) over a large field of view. camera differences up to ∼10%; empirically, stitching errors be￾came noticeable when differences exceeded ∼15%. Third, for steep curvature, periph… view at source ↗
Figure 2
Figure 2. Imaging configurations for different sample geometries and focusing strategies. (a) Concave sample with single focus yields near-flat re-imaging. (b) Flat sample with single focus shows defocus from field curvature. (c) Multi-focus corrects defocus for flat samples. (d) Multi-focus enables uniform focus for convex samples. (e) Excessive depth variation within a camera’s FOV causes local defocus. (f) Large inter-came… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Multi-focus snapshot imaging of curved biological samples using a tiled sub-camera array microscope. (a) Schematic of an onion epidermal sheet placed on a gently curved substrate. The system supports adaptive per-camera focus to match sample topology. (b) Whole-sample …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

26 extracted references · 26 canonical work pages

  1. [1]

    K. C. Zhou, M. Harfouche, M. Zheng, et al., J. Big Data 11, 62 (2024)

    work page 2024

  2. [2]

    K. Kim, A. Chaware, C. B. Cook, et al., npj Imaging 2, 39 (2024)

    work page 2024

  3. [3]

    J. Fan, J. Suo, J. Wu, et al., Nat. Photonics 13, 809 (2019)

    work page 2019

  4. [4]

    K. C. Zhou, M. Harfouche, C. L. Cooke, et al., Nat. photonics 17, 442 (2023)

    work page 2023

  5. [5]

    H. Chen, K. Liu, Y . Jiang, et al., Opt. Express 33, 22062 (2025)

    work page 2025

  6. [6]

    McConnell, J

    G. McConnell, J. Trägårdh, R. Amor, et al., Elife 5, e18659 (2016)

    work page 2016

  7. [7]

    R. Shi, X. Chen, J. Deng, et al., Nat. Photonics 18, 721 (2024)

    work page 2024

  8. [8]

    Ichimura, T

    T. Ichimura, T. Kakizuka, Y . Taniguchi,et al., Elife 13, RP93633 (2025)

    work page 2025

Show all 26 references
  1. [9]

    D. J. Brady, M. E. Gehm, R. A. Stack, et al., Nature 486, 386 (2012)

    2012

  2. [10]

    Harfouche, K

    M. Harfouche, K. Kim, K. C. Zhou, et al., Optica 10, 471 (2023)

    2023

  3. [11]

    K. C. Zhou, C. Cook, A. Chakraborty, et al., Optica 12, 674 (2025)

    2025

  4. [12]

    Kreiss, W

    L. Kreiss, W. Tang, R. Balla, et al. , Biomed. Opt. Express 16, 617 (2025)

    2025

  5. [13]

    Y ang, M

    X. Y ang, M. Harfouche, K. C. Zhou, et al., Opt. Lett. 48, 1658 (2023). Letter 5 FULL REFERENCES

    2023

  6. [14]

    Computational 3d topo- graphic microscopy from terabytes of data per sample,

    K. C. Zhou, M. Harfouche, M. Zheng, et al., “Computational 3d topo- graphic microscopy from terabytes of data per sample,” J. Big Data 11, 62 (2024)

    2024

  7. [15]

    Rapid 3d imaging at cellular resolution for digital cytopathology with a multi-camera array scanner (mcas),

    K. Kim, A. Chaware, C. B. Cook, et al., “Rapid 3d imaging at cellular resolution for digital cytopathology with a multi-camera array scanner (mcas),” npj Imaging 2, 39 (2024)

    2024

  8. [16]

    Video-rate imaging of biological dynamics at centimetre scale and micrometre resolution,

    J. Fan, J. Suo, J. Wu, et al., “Video-rate imaging of biological dynamics at centimetre scale and micrometre resolution,” Nat. Photonics 13, 809–816 (2019)

    2019

  9. [17]

    Parallelized compu- tational 3d video microscopy of freely moving organisms at multiple gigapixels per second,

    K. C. Zhou, M. Harfouche, C. L. Cooke, et al. , “Parallelized compu- tational 3d video microscopy of freely moving organisms at multiple gigapixels per second,” Nat. photonics 17, 442–450 (2023)

    2023

  10. [18]

    Multierror learning enhanced fluores- cence diffusion tomography in reflection geometry,

    H. Chen, K. Liu, Y . Jiang,et al., “Multierror learning enhanced fluores- cence diffusion tomography in reflection geometry,” Opt. Express33, 22062–22077 (2025)

    2025

  11. [19]

    A novel optical micro- scope for imaging large embryos and tissue volumes with sub-cellular resolution throughout,

    G. McConnell, J. Trägårdh, R. Amor, et al. , “A novel optical micro- scope for imaging large embryos and tissue volumes with sub-cellular resolution throughout,” Elife 5, e18659 (2016)

    2016

  12. [20]

    Random-access wide-field mesoscopy for centimetre-scale imaging of biodynamics with subcellular resolution,

    R. Shi, X. Chen, J. Deng,et al., “Random-access wide-field mesoscopy for centimetre-scale imaging of biodynamics with subcellular resolution,” Nat. Photonics 18, 721–730 (2024)

    2024

  13. [21]

    Volumetric trans-scale imaging of massive quantity of heterogeneous cell populations in centimeter-wide tissue and embryo,

    T. Ichimura, T. Kakizuka, Y . Taniguchi,et al., “Volumetric trans-scale imaging of massive quantity of heterogeneous cell populations in centimeter-wide tissue and embryo,” Elife 13, RP93633 (2025)

    2025

  14. [22]

    Multiscale gigapixel photography,

    D. J. Brady, M. E. Gehm, R. A. Stack, et al. , “Multiscale gigapixel photography,” Nature486, 386–389 (2012)

    2012

  15. [23]

    Imaging across multiple spatial scales with the multi-camera array microscope,

    M. Harfouche, K. Kim, K. C. Zhou, et al. , “Imaging across multiple spatial scales with the multi-camera array microscope,” Optica 10, 471–480 (2023)

    2023

  16. [24]

    High-speed 4d fluores- cence light field tomography of whole freely moving organisms,

    K. C. Zhou, C. Cook, A. Chakraborty, et al., “High-speed 4d fluores- cence light field tomography of whole freely moving organisms,” Optica 12, 674–684 (2025)

    2025

  17. [25]

    Recording dynamic facial micro- expressions with a multi-focus camera array,

    L. Kreiss, W. Tang, R. Balla, et al., “Recording dynamic facial micro- expressions with a multi-focus camera array,” Biomed. Opt. Express 16, 617–627 (2025)

    2025

  18. [26]

    Multi-modal imaging using a cascaded microscope design,

    X. Y ang, M. Harfouche, K. C. Zhou,et al., “Multi-modal imaging using a cascaded microscope design,” Opt. Lett. 48, 1658–1661 (2023)

    2023

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.