Pith. sign in

REVIEW 3 major objections 5 minor 37 references

Bootstrap-based Hypothesis Test of 2D Contours using Elastic Shape Analysis

T0 review · 3 major / 5 minor · reviewed 2026-07-12 · grok-4.5

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

Pith's one-line read A one-sample bootstrap test for whether an image contour matches a hypothesized shape, using elastic shape distance.

desk verdict Solid first one-sample ESD test via m-out-of-N bootstrap; usable for ICF and generalizable, with free parameters and data-on-request as the main soft spots. read the letter →

arxiv 2606.04879 v2 pith:A6W2A2GU submitted 2026-06-03 stat.ME stat.AP

Susan Glenn , Justin Strait , Kelly Moran , Christopher Danly , Matthew P Selwood This is my paper · ORCID
classification stat.MEstat.AP MSC 62G0962H3562G10
keywords elasticshapeanalysisdistancebootstrapconfidenceintervalsone-samplehypothesistestpercentilecontoursinertialconfinementfusionm-out-of-Nnon-smoothfunctionals
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

Elastic shape analysis can measure how two curves differ after removing rotation, scale, translation, and reparameterization, but it has lacked a simple one-sample hypothesis test for contours extracted from images. This paper supplies that test: construct an empirical confidence interval for the elastic shape distance between a hypothesized true contour and the contour estimated from data, then reject if the hypothesized shape falls outside the interval. The interval is built by an m-out-of-N bootstrap that handles the fact that the elastic distance is not differentiable at zero. Simulations show controlled Type I error once the bootstrap sample size is chosen appropriately, and power against common deviations such as indents, ellipses, and polygons. Real neutron images from inertial confinement fusion experiments illustrate when reconstructed source shapes can or cannot be treated as circular.

What carries the argument

bootESA: map images to percentile contours, convert to square-root velocity functions, form the elastic shape distance, then build a rescaled bootstrap distribution of those distances (with m = N^ε when the ordinary bootstrap is invalid) to produce the confidence set and p-value.

What would settle it

On synthetic images whose true source is known to be a circle, run the test at the recommended m = N^0.8 and check whether the observed Type I error approaches the nominal level (e.g., 0.1) as sample size grows; systematic under- or over-rejection would falsify the calibration claim.

Watch

Extended reading notes

Core claim

The first elastic-shape-analysis one-sample test for image contours: empirical confidence intervals for the elastic shape distance between a proposed true shape and an estimated shape, obtained from an m-out-of-N bootstrap that accounts for the non-differentiability of that distance.

Load-bearing premise

The mean image must be smooth enough that the chosen percentile contour stays a single closed curve under small data changes; if topology flips or the intensity map is too rough, the bootstrap justification fails.

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 / 5 minor

Summary. The paper proposes bootESA, a one-sample hypothesis test for 2D contours under elastic shape analysis. After reconstructing a mean image from multiple observations (e.g., pinhole sub-images), a percentile contour is extracted, converted to its square-root velocity function, and compared to a hypothesized shape via the elastic shape distance (ESD). Because the ESD is non-differentiable at zero, ordinary bootstrap can be invalid; the authors therefore construct empirical confidence intervals and p-values with an m-out-of-N bootstrap (Eqs. 9–12, Sections 3.1–3.2). Validity rests on two smoothness/topology assumptions that make the image-to-SRVF map continuously differentiable. Simulations examine Type I error under different m = N^ε and percentiles (Table 3) and power against indented circles, ellipses and polygons (Figure 6); two real ICF neutron images illustrate the procedure.

Significance. Formal one-sample inference for ESA contours has been largely missing; existing ESA tests are multi-class permutation or energy tests. A bootstrap procedure that explicitly accounts for the non-smoothness of the ESD is therefore a useful methodological contribution, especially for imaging applications (ICF, TDA-style density contours) where a target shape is of scientific interest. The paper supplies concrete Type I error and power evidence under the stated assumptions and demonstrates that a practical choice ε ≈ 0.8 recovers near-nominal size for mid-percentiles. The framework is generalizable beyond ICF once a consistent mean image and bootstrap of that mean are available.

major comments (3)
  1. Section 3.2.1 and Table 3: the practical selection of ε (hence m = N^ε) is left largely empirical. Table 3 shows that m = N yields Type I error near 0 for most percentiles, while ε = 0.6–0.7 can inflate it well above α = 0.1; ε = 0.8 works for p ≤ 0.85 but still under-covers at p = 0.95. The supplemental stability plots (Figure 11) are helpful but do not constitute a data-driven rule that a practitioner can apply without knowing the true distribution. A clearer, reproducible recommendation (or a diagnostic that flags when ordinary bootstrap is safe) is needed for the method to be usable outside the authors’ simulations.
  2. Assumptions 1–2 (Section 3.2) are load-bearing for continuous differentiability of the image-to-SRVF map and therefore for both ordinary and m-out-of-N bootstrap validity. Figure 4 shows that topology can break for N ≲ 30; the real-data examples use N = 55 and m = N “to be conservative,” yet no quantitative check of smoothness or single-component topology is reported for those images. The paper should either supply a practical diagnostic or quantify how mild violations affect coverage, otherwise the real-data conclusions rest on unverified assumptions.
  3. Section 5 and Figure 8: the real-data analysis tests every percentile contour separately at α = 0.05 without multiplicity adjustment, while the text notes that a single p is chosen a priori in practice. Presenting a battery of unadjusted tests invites over-interpretation of which percentiles “are circular.” Either restrict the main analysis to one pre-specified percentile or apply a multiple-testing correction and discuss the hierarchical dependence across percentiles.
minor comments (5)
  1. Notation for the bootstrap scale is inconsistent: √m appears in Eqs. 9–10 while the test statistic later uses √N (Section 3.2.1); a single, explicit statement of the scaling used for R_obs versus R*_m would help.
  2. Table 1 caption and surrounding text refer to “spherical harmonic fit” while the method is Legendre polynomial decomposition; the terminology should be aligned.
  3. Figure 5c and the corresponding violin plots in the supplement would be clearer if the asymptotic scaling (√m or √N) were applied before plotting, so that the visual comparison matches the quantities used in the hypothesis test.
  4. A few typographical issues: “bootESAfor short,” “ak-sample,” “them-out-of-N,” and missing spaces after periods appear in the introduction and Section 3.
  5. The claim of being “the first ESA-based one-sample test” is plausible given the cited multi-class work, but a brief comparison with existing Procrustes or landmark-based one-sample bootstrap procedures would better situate the contribution.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; bootESA is a standard m-out-of-N bootstrap CI/test for the established ESD functional with an externally supplied null shape.

full rationale

The paper's central construction (Sections 3.1–3.2, Eqs. 9–12) takes the elastic shape distance of Srivastava et al. as a given metric on the quotient shape space, forms the usual empirical bootstrap distribution of that functional under the image-to-contour-to-SRVF map, and replaces the ordinary n-out-of-n scale by the m-out-of-N scale of Bickel et al. precisely because the ESD map is only directionally differentiable. The hypothesized contour c0 (or its SRVF q0) is supplied externally and is never estimated from the same data that generate the bootstrap quantiles; the resulting p-value and CI are therefore not forced by construction. Assumptions 1–2 are stated explicitly as the conditions under which the continuous-mapping and delta-method steps hold, and are checked rather than assumed away. Citations to the authors’ own ICF reconstruction pipeline (Volegov, Lamb) appear only in the data-setup section and do not underwrite the validity of the bootstrap argument. No parameter is fitted and then re-used as a “prediction,” no uniqueness theorem is imported from prior work by the same authors, and no known empirical pattern is merely renamed. The derivation is therefore self-contained against the external bootstrap and ESA literature.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central claim rests on classical consistency of the sample-mean image, continuous differentiability of the image-to-SRVF map under two smoothness/topology assumptions, directional differentiability of ESD, and the known validity of m-out-of-N bootstrap for non-smooth functionals. Free parameters are the bootstrap subsample size exponent and the contour percentile; no new physical entities are postulated.

free parameters (3)
  • ε (m = N^ε) = 0.8 (recommended)
    Chosen by visual stability of successive ESD distributions (Section 4.1 and Supplement); recommended value 0.8 for ICF mid-percentiles, but not derived from a data-independent rule.
  • percentile level p = 0.65–0.95 (simulations); 0.85–1.0 (real data)
    User-selected contour intensity quantile; results vary across p and multiple-testing correction is omitted.
  • B (number of bootstrap replicates) = 1000
    Monte-Carlo sample size for empirical quantiles; set to 1000 in real-data examples without sensitivity analysis.
assumptions (5)
  • domain assumption Sample-mean image converges in probability to the true source image (LLN + consistency of EM/Lucy–Richardson under Poisson model).
    Invoked in Section 3.2 to justify √N-consistency of the image estimator before mapping to shape space.
  • domain assumption Assumption 1: pixel-intensity distribution of the mean image is smooth and non-degenerate so that the percentile function is differentiable.
    Required for continuous differentiability of the image-to-contour map (Section 3.2).
  • domain assumption Assumption 2: topology of the extracted contour remains a single closed curve under small perturbations.
    Needed so that ESD is always defined between two curves of the same topology (Section 3.2).
  • standard math m-out-of-N bootstrap is consistent for non-smooth functionals when m/N → 0 (Bickel & Sakov 2008).
    Cited to justify replacing ordinary bootstrap by m-out-of-N because ESD is not differentiable at the origin.
  • standard math ESD is a true metric on the elastic shape space after optimal registration (Srivastava et al.).
    Background fact used throughout; not re-proved.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Bootstrap-based Hypothesis Test of 2D Contours using Elastic Shape Analysis." pith.science (2026). https://pith.science/paper/A6W2A2GU

@misc{pith2026260604879,
  author       = {Pith},
  title        = {Pith review of: Bootstrap-based Hypothesis Test of 2D Contours using Elastic Shape Analysis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/A6W2A2GU}},
  note         = {Machine review of arXiv:2606.04879}
}
read the original abstract

Shapes of objects in images are often complex, high-dimensional, and vary in ways not captured by standard Euclidean geometry and statistics. Statistical shape analysis encompasses methods for flexible and interpretable measurement of intrinsic shape and shape variability in geometric objects. Elastic Shape Analysis (ESA) is one such method that measures shape differences between objects, represented by contours, in a way that is invariant to rotation, scale, translation, and parameterization. Although ESA is useful for quantifying shape of objects in many image applications, formal methods for statistical inference in image-based ESA remain limited. This work introduces a hypothesis test procedure based on empirical confidence intervals for the elastic shape distance (ESD) between a proposed underlying true shape and an estimated shape. The confidence intervals are created using a bootstrap procedure for non-smooth functionals, which accounts for the non-differentiability of the ESD. The effectiveness of the method is illustrated through both numerical studies and real world image examples from inertial confinement fusion (ICF).

Figures

Figures reproduced from arXiv: 2606.04879 by the authors.

Figure 1
Figure 1. Example of neutron source with pinhole path (red line) through pinhole aperture [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. The data generating process (first row) convolves the true source image with a [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Example of pipeline to get ESD between two curves [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Simulation example with images of the mean reconstructed source by number of [PITH_FULL_IMAGE:figures/full_fig_p019_4.png]
Figure 5
Figure 5. Figure 5: Simulation results for sample mean contours ¯c [PITH_FULL_IMAGE:figures/full_fig_p021_5.png]
Figure 6
Figure 6. Figure 6: Power analysis of the hypothesis test when [PITH_FULL_IMAGE:figures/full_fig_p024_6.png]
Figure 7
Figure 7. Figure 7: Sample mean images (a) S¯ 1 and (c) S¯ 2 for two separate ICF shots. The contour color corresponds to percentile for p = {0.85, . . . , 1} where purple is for the 85th percentile and red is for the 100th percentile. Four different bootstrap sample mean images are shows…
Figure 8
Figure 8. Figure 8: Hypothesis test results for two different ICF implosions shots in (a) [PITH_FULL_IMAGE:figures/full_fig_p026_8.png]
Figure 9
Figure 9. Figure 9: Example of the data setup with (a) the true density, (b) data estimates using the [PITH_FULL_IMAGE:figures/full_fig_p029_9.png]
Figure 10
Figure 10. Figure 10: Simulation results for sample mean contours ¯c [PITH_FULL_IMAGE:figures/full_fig_p030_10.png]
Figure 11
Figure 11. Figure 11: Simulations for finding stability for m in bootstrap with the target distribution (black) and m ∈ {Nϵ} on the x−axis for ϵ ∈ [0.5, 1]. 31 [PITH_FULL_IMAGE:figures/full_fig_p031_11.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 3 linked inside Pith

  1. [1]

    M., et al

    Merrill, F. M., et al. (2012). The neutron imaging diagnostic at NIF (invited). Review of Scientific Instruments, 83(10)

    2012

  2. [2]

    R., Fittinghoff, D

    Volegov, P., Danly, C. R., Fittinghoff, D. N., Grim, G. P., Guler, N., Izumi, N., ..., and Wilson, D. C. (2014). Neutron source reconstruction from pinhole imaging at National Ignition Facility. Review of Scientific Instruments, 85(2)

    2014

  3. [3]

    M., et al

    Lamb, K. M., et al. (2022). Bootstrap estimation of the effect of instrument response function uncertainty on the reconstruction of fusion neutron sources. Review of Scientific Instruments, 93(4)

    2022

  4. [4]

    Kritcher, A. L. et al. (2024). Design of first experiment to achieve fusion target gain 1 Physics of Plasmas, 31(7), 070502

    2024

  5. [5]

    V., Wu, X., Kent, J

    Mardia, K. V., Wu, X., Kent, J. T., Goodall, C. R., and Khambay, B. S. (2024). Asymmetry Analysis of Bilateral Shapes. arXiv preprint arXiv:2407.17225

    arXiv 2024

  6. [6]

    J., Dryden, I

    Amaral, G. J., Dryden, I. L., Patrangenaru, V., and Wood, A. T. (2010). Bootstrap confidence intervals for the planar mean shape. Journal of Statistical Planning and Inference, 140(11), 3026-3034

    2010

  7. [7]

    Amaral, G. J. (2004). Bootstrap and empirical likelihood methods in statistical shape analysis (Doctoral dissertation, University of Nottingham)

    2004

  8. [8]

    Kurtek, S., Srivastava, A., Klassen, E., and Ding, Z. (2012). Statistical modeling of curves using shapes and related features. Journal of the American Statistical Association, 107(499), 1152-1165

    2012

Show all 37 references
  1. [9]

    Samir, C., Srivastava, A., Daoudi, M., and Kurtek, S. (2009). On Analyzing Symmetry of Objects using Elastic Deformations. In VISAPP, (1) 194-200

    2009

  2. [10]

    Dogan, G., Bernal, J., and Hagwood, C. (2025). FFT-based alignment of 2d closed curves with application to elastic shape analysis. arXiv preprint arXiv:2501.17779

    2025 arXiv

  3. [11]

    Wald, A. (1949). Note on the consistency of the maximum likelihood estimate. The Annals of Mathematical Statistics, 20(4), 595-601

    1949

  4. [12]

    Liu, J., Chen, D., and Wei, G. W. (2025). Parametrization of Symmetry in Data. arXiv preprint arXiv:2508.07531

    2025 arXiv

  5. [13]

    Srivastava, A., and Klassen, E. P. (2016). Functional and shape data analysis. (Vol. 1, pp. 43-44). New York: Springer

    2016

  6. [14]

    Srivastava, A., Wu, W., Kurtek, S., Klassen, E., and Marron, J. S. (2011). Statistical analysis and modeling of elastic functions. arXiv preprint arXiv:1103.3817

    2011 arXiv

  7. [15]

    and Kurtek, S

    Bharath, K. and Kurtek, S. (2020). Analysis of shape data: From landmarks to elastic curves. Wiley Interdisciplinary Reviews: Computational Statistics, 12(3), e1495

    2020

  8. [16]

    H., Xie, Q., Klassen, E., and Laga, H

    Kurtek, S., Jermyn, I. H., Xie, Q., Klassen, E., and Laga, H. (2016). Elastic shape analysis of surfaces and images. In Riemannian computing in computer vision (pp. 257-277). Cham: Springer International Publishing

    2016

  9. [17]

    A., and Absil, P

    Huang, W., You, Y., Gallivan, K. A., and Absil, P. A. (2015). Karcher mean in elastic shape analysis. In Proceedings of the 1st International Workshop on Differential Geometry in Computer Vision for Analysis of Shapes, Images and Trajectories

    2015

  10. [18]

    T., Lecci, F., Rinaldo, A., Wasserman, L., Balakrishnan, S., and Singh, A

    Fasy, B. T., Lecci, F., Rinaldo, A., Wasserman, L., Balakrishnan, S., and Singh, A. (2014). Confidence sets for persistence diagrams

    2014

  11. [19]

    J., and Sakov, A

    Bickel, P. J., and Sakov, A. (2008). On the choice of m in the m out of n bootstrap and confidence bounds for extrema. Statistica Sinica, 967-985

    2008

  12. [20]

    J., Götze, F., and van Zwet, W

    Bickel, P. J., Götze, F., and van Zwet, W. R. (2011). Resampling fewer than n observations: gains, losses, and remedies for losses. In Selected works of Willem van Zwet (pp. 267-297). New York, NY: Springer New York

    2011

  13. [21]

    Subsampling

    Politis, D.N., Romano, J.P., and Wolf, M. Subsampling. Springer, New York, 1999

    1999

  14. [22]

    and Mitteroecker, P

    Gunz, P. and Mitteroecker, P. (2013). Semilandmarks: a method for quantifying curves and surfaces. Hystrix, the Italian journal of mammalogy, 24(1), 103-109

    2013

  15. [23]

    and Zhou, C

    Jiang, Z. and Zhou, C. (2024). Comprehensive study on shape representation methods for shape-based object recognition. Journal of Optics, 53(3), 1890-1896

    2024

  16. [24]

    (2012, June)

    Bryner, D., Srivastava, A., and Klassen, E. (2012, June). Affine-invariant, elastic shape analysis of planar contours. In 2012 IEEE Conference on Computer Vision and Pattern Recognition (pp. 390-397). IEEE

    2012

  17. [25]

    (2008, October)

    Cates, J., Fletcher, T., and Whitaker, R. (2008, October). A hypothesis testing framework for high-dimensional shape models. In 2nd MICCAI workshop on mathematical foundations of computational anatomy (pp. 170-181)

    2008

  18. [26]

    and Woods, R.E

    Gonzalez, R.C. and Woods, R.E. (2018) Digital Image Processing. 4th Edition, Pearson Education, New York, 1022 p

    2018

  19. [27]

    and Strait, J.D

    Luo, H. and Strait, J.D. (2024). Multiple closed curve modeling with uncertainty quantification for shape analysis. SIAM/ASA Journal on Uncertainty Quantification, 12(4), 1192-1212

    2024

  20. [28]

    and Greven, S

    Steyer, L., Stocker, A. and Greven, S. (2023). Elastic analysis of irregularly or sparsely sampled curves. Biometrics, 79(3), 2103-2115

    2023

  21. [29]

    Strait, J.D., Kurtek, S.A., Bartha, E., and MacEachern, S.N. (2017). Landmark-constrained elastic shape analysis of planar curves. Journal of the American Statistical Association, 112(518), 521-533

    2017

  22. [30]

    Zhang, R., Ogden, R.T., Picard, M., and Srivastava, A. (2021). Nonparametric k -sample test on shape spaces with applications to mitochondrial shape analysis. Journal of the Royal Statistical Society, Series C, 71, 51-69

    2021

  23. [31]

    H., and Jermyn, I

    Srivastava, A., Klassen, E., Joshi, S. H., and Jermyn, I. H. (2010). Shape analysis of elastic curves in euclidean spaces. IEEE transactions on pattern analysis and machine intelligence, 33(7), 1415-1428

    2010

  24. [32]

    Abboud, M., Benzinou, A., and Nasreddine, K. (2020). A robust tangent PCA via shape restoration for shape variability analysis. Pattern Analysis and Applications, 23(2), 653-671

    2020

  25. [33]

    and Mardia, K.V

    Dryden, I.L. and Mardia, K.V. (2016). Statistical Shape Analysis: With Applications in R. Second edition, Wiley, Chichester

    2016

  26. [34]

    Kendall, D.G. (1984). Shape manifolds, Procrustean metrics, and complex projective spaces. Bulletin of the London Mathematical Society, 16(2), 81--121

    1984

  27. [35]

    Bookstein, F.L. (1991). Morphometric Tools for Landmark Data: Geometry and Biology. Cambridge University Press, Cambridge

    1991

  28. [36]

    E., Ross, J

    Ralph, J. E., Ross, J. S., Zylstra, A. B., Kritcher, A. L., Robey, H. F., Young, C. V., ... and Wilde, C. H. (2024). The impact of low-mode symmetry on inertial fusion energy output in the burning plasma state. Nature communications, 15(1), 2975

    2024

  29. [37]

    J., Archuleta, T

    Guler, N., Aragonez, R. J., Archuleta, T. N., Batha, S. H., Clark, D. D., Clark, D. J., ... and Traille, A. (2013). First downscattered neutron images from Inertial Confinement Fusion experiments at the National Ignition Facility. In EPJ Web of Conferences (Vol. 59, p. 13018)....

    2013

Pith tools

Reviewed July 12, 2026 · model on record in the stance chip above.