pith. sign in

arxiv: 2507.10253 · v2 · submitted 2025-07-14 · ⚛️ physics.ins-det · hep-ex

Optimization of a cosmic muon tomography scanner for cargo border control inspection

Z. Zaher , H. Lay , T. Dorigo , A. Giammanco , V. Gulik , C. Hrytsiuk , V. A. Kudryavtsev , M. Lagrange
show 4 more authors
T. Metspalu G. C. Strong C. Turkoglu P. Vischia
This is my paper

Pith reviewed 2026-05-19 05:02 UTC · model grok-4.3

classification ⚛️ physics.ins-det hep-ex
keywords muon tomographycosmic rayscargo inspectiondetector optimizationBayesian optimizationGEANT4 simulationborder securitysecondary particles
0
0 comments X

The pith

Bayesian optimization added to TomOpt handles noisy objectives to tune muon tomography detectors for cargo inspection.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper examines optimization of a cosmic muon tomography scanner intended for border security checks on trucks and sea containers. It adds a Bayesian Optimization module to the TomOpt software package to manage noisy objective functions that arise when optimization is driven by image reconstruction quality. This addition works alongside existing gradient-based methods and alongside full GEANT4 simulations that incorporate secondary particles to improve material discrimination. A reader would care because practical gains in detector layout could make non-invasive scanning for hidden hazardous materials more reliable at scale.

Core claim

The paper establishes that a Bayesian Optimization module introduced into TomOpt enables effective optimization of scattering tomography detector configurations even when the objective function is noisy, complementing gradient-based approaches, while GEANT4-based simulations supply higher-fidelity modeling and demonstrate that secondary particle information alongside cosmic muons improves material discrimination for cargo inspection geometries similar to those targeted by the SilentBorder project.

What carries the argument

Bayesian Optimization module within TomOpt for managing noisy objective functions during image-reconstruction-driven detector configuration tuning.

If this is right

  • Detector module designs for the SilentBorder muon tomography system can be improved through combined gradient and Bayesian optimization.
  • Inclusion of secondary particle data yields better separation of materials inside inspected cargo.
  • TomOpt becomes usable for optimization tasks where image reconstruction noise dominates the objective.
  • The two complementary strategies—differentiable programming and detailed Monte Carlo—can be applied together to balance speed and fidelity.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same Bayesian module might reduce the number of simulation runs needed when extending the scanner to new cargo types.
  • Real border data could be used to retrain or calibrate the optimization loop and close the simulation-to-reality gap.
  • The approach could transfer to other muon tomography settings such as industrial imaging or geological surveys.

Load-bearing premise

The GEANT4 simulations of muon scattering and secondary particle production accurately reflect real cosmic muon interactions and detector responses in the specific cargo inspection geometry and materials.

What would settle it

Running the optimized detector configuration in a controlled test with known cargo loads and comparing the reconstructed images and material discrimination performance against the simulation predictions would settle the claim.

Figures

Figures reproduced from arXiv: 2507.10253 by A. Giammanco, C. Hrytsiuk, C. Turkoglu, G. C. Strong, H. Lay, M. Lagrange, P. Vischia, T. Dorigo, T. Metspalu, V. A. Kudryavtsev, V. Gulik, Z. Zaher.

Figure 1
Figure 1. Figure 1: FIG. 1: The SilentBorder muon scanner prototype produced [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2: A visualization of a particular configuration of the [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3: Locations of reconstructed PoCA points within the [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5: The change in efficiency (top) and angular resolution [PITH_FULL_IMAGE:figures/full_fig_p005_5.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7: The total true energy deposition for each hodoscope [PITH_FULL_IMAGE:figures/full_fig_p006_7.png] view at source ↗
Figure 9
Figure 9. Figure 9: FIG. 9: The total number of hits recorded across all [PITH_FULL_IMAGE:figures/full_fig_p007_9.png] view at source ↗
Figure 8
Figure 8. Figure 8: FIG. 8: The total number of hits recorded across all [PITH_FULL_IMAGE:figures/full_fig_p007_8.png] view at source ↗
Figure 11
Figure 11. Figure 11: FIG. 11: The number of hits (split into muon hits and [PITH_FULL_IMAGE:figures/full_fig_p008_11.png] view at source ↗
Figure 10
Figure 10. Figure 10: FIG. 10: Two schematic examples of a muon passing through [PITH_FULL_IMAGE:figures/full_fig_p008_10.png] view at source ↗
Figure 13
Figure 13. Figure 13: FIG. 13: The dependence of the contamination proportion on [PITH_FULL_IMAGE:figures/full_fig_p009_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: FIG. 14: The contamination proportion for a variety of cargo [PITH_FULL_IMAGE:figures/full_fig_p009_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: FIG. 15: TomOpt optimization pipeline, as described in [10] [PITH_FULL_IMAGE:figures/full_fig_p010_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: FIG. 16: Example of the surrogate model used to describe [PITH_FULL_IMAGE:figures/full_fig_p010_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: FIG. 17: 3D volume inference using the radiation length [PITH_FULL_IMAGE:figures/full_fig_p011_17.png] view at source ↗
Figure 19
Figure 19. Figure 19: FIG. 19: Relationship between inferred quantities and [PITH_FULL_IMAGE:figures/full_fig_p012_19.png] view at source ↗
Figure 20
Figure 20. Figure 20: FIG. 20: The initial configuration of the hodoscope setup in [PITH_FULL_IMAGE:figures/full_fig_p014_20.png] view at source ↗
Figure 23
Figure 23. Figure 23: FIG. 23: The evolution of the gap size between upper [PITH_FULL_IMAGE:figures/full_fig_p015_23.png] view at source ↗
Figure 24
Figure 24. Figure 24: FIG. 24: The BO optimized configuration of the hodoscope [PITH_FULL_IMAGE:figures/full_fig_p016_24.png] view at source ↗
Figure 25
Figure 25. Figure 25: FIG. 25: Convergence of the best minimum objective across [PITH_FULL_IMAGE:figures/full_fig_p016_25.png] view at source ↗
Figure 28
Figure 28. Figure 28: FIG. 28: The BO optimized configuration of the hodoscope [PITH_FULL_IMAGE:figures/full_fig_p017_28.png] view at source ↗
Figure 29
Figure 29. Figure 29: FIG. 29: Statistical distribution of volume inference [PITH_FULL_IMAGE:figures/full_fig_p017_29.png] view at source ↗
Figure 30
Figure 30. Figure 30: FIG. 30: The configuration of the baseline detector setup in [PITH_FULL_IMAGE:figures/full_fig_p018_30.png] view at source ↗
read the original abstract

The past several decades have seen significant advancement in applications using cosmic-ray muons for tomography scanning of unknown objects. One of the most promising developments is the application of this technique in border security for the inspection of cargo inside trucks and sea containers in order to search for hazardous and illicit hidden materials. This work focuses on the optimization studies for a muon tomography system similar to that being developed within the framework of the `SilentBorder' project funded by the EU Horizon 2020 scheme. Current studies are directed toward optimizing the detector module design, following two complementary approaches. The first leverages TomOpt, a Python-based end-to-end software that employs differentiable programming to optimize scattering tomography detector configurations. While TomOpt inherently supports gradient-based optimization, a Bayesian Optimization module is introduced to better handle scenarios with noisy objective functions, particularly in image reconstruction-driven optimization tasks. The second optimization strategy relies on detailed GEANT4-based simulations, which, while more computationally intensive, offer higher physical fidelity. These simulations are also employed to study the impact of incorporating secondary particle information alongside cosmic muons for improved material discrimination. This paper presents the current status and results obtained from these optimization studies.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper reports optimization studies for a cosmic muon tomography scanner aimed at cargo border control, using two complementary methods: (i) extension of the TomOpt differentiable-programming framework with a Bayesian Optimization module to handle noisy image-reconstruction objectives, and (ii) detailed GEANT4 simulations to refine detector-module geometry and to assess the added value of secondary-particle information for material discrimination.

Significance. If the GEANT4 modeling of scattering and secondary yields is shown to be sufficiently accurate for the relevant cargo materials and geometries, the work could supply concrete, simulation-validated detector configurations and a practical demonstration that secondary-particle channels improve discrimination in a realistic border-inspection setting.

major comments (2)
  1. [GEANT4 simulation and results sections] The central claim that the Bayesian Optimization module in TomOpt yields improved configurations for noisy reconstruction-driven objectives rests on the fidelity of the underlying GEANT4 scattering and secondary-particle distributions; however, the manuscript provides no quantitative validation of these distributions against experimental data for the specific truck/container materials and detector layout (see the GEANT4 simulation section and the results on material discrimination).
  2. [Optimization results] No baseline comparisons (e.g., to gradient-based TomOpt optimization or to a non-optimized reference geometry) or uncertainty estimates on the reported image-quality metrics are presented, which prevents assessment of whether the claimed gains in material discrimination are statistically significant or practically meaningful.
minor comments (2)
  1. [TomOpt Bayesian Optimization module] Clarify the precise definition of the objective function used inside the Bayesian Optimization loop and how it relates to the image-reconstruction metric.
  2. [Discussion] Add a short discussion of computational cost trade-offs between the TomOpt and full GEANT4 approaches to help readers judge practicality for iterative design studies.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed report. We address each major comment below, indicating the revisions we intend to make to strengthen the manuscript while remaining faithful to its simulation-focused scope.

read point-by-point responses

  1. Referee: [GEANT4 simulation and results sections] The central claim that the Bayesian Optimization module in TomOpt yields improved configurations for noisy reconstruction-driven objectives rests on the fidelity of the underlying GEANT4 scattering and secondary-particle distributions; however, the manuscript provides no quantitative validation of these distributions against experimental data for the specific truck/container materials and detector layout (see the GEANT4 simulation section and the results on material discrimination).

    Authors: We acknowledge that the manuscript does not contain new quantitative comparisons of GEANT4 distributions to experimental data for the exact cargo materials and detector layout under study. The work presented is a simulation-based optimization study; the GEANT4 component employs the standard QGSP_BERT physics list together with muon-specific processes that have been benchmarked in prior muon-tomography literature. In the revised manuscript we will expand the GEANT4 section to cite these existing validation studies, explicitly state the limitations of the current simulation-only results, and clarify that the Bayesian-optimization gains are demonstrated within the simulation framework rather than claimed as experimentally validated. Full experimental benchmarking remains a goal of the ongoing SilentBorder project and is outside the scope of this paper. revision: partial

  2. Referee: [Optimization results] No baseline comparisons (e.g., to gradient-based TomOpt optimization or to a non-optimized reference geometry) or uncertainty estimates on the reported image-quality metrics are presented, which prevents assessment of whether the claimed gains in material discrimination are statistically significant or practically meaningful.

    Authors: We agree that the absence of explicit baselines and uncertainty estimates limits the interpretability of the reported gains. In the revised manuscript we will add (i) a direct comparison of the Bayesian-optimized geometries against a standard non-optimized reference layout and, where computationally feasible, against the gradient-based TomOpt optimizer, and (ii) uncertainty estimates on the image-quality and discrimination metrics obtained from multiple independent simulation runs with varied random seeds. These additions will allow readers to evaluate both the magnitude and the statistical robustness of the improvements. revision: yes

Circularity Check

0 steps flagged

No circularity: optimization objectives defined externally via image metrics and GEANT4 fidelity

full rationale

The paper describes introducing a Bayesian Optimization module into TomOpt to handle noisy reconstruction-driven objectives, alongside GEANT4 simulations for secondary-particle studies. No derivation reduces by construction to its inputs: targets are image quality metrics and material discrimination goals set independently of the fitted configurations. TomOpt and GEANT4 are treated as external tools with stated assumptions about scattering and secondary yields; no self-definitional loop, fitted-input-as-prediction, or load-bearing self-citation chain appears in the abstract or described workflow. The central results are simulation outputs, not tautological renamings or uniqueness theorems imported from the authors' prior work.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

The work relies on standard assumptions from cosmic ray physics and detector simulation without introducing new entities or many free parameters beyond typical optimization hyperparameters.

free parameters (1)
  • Bayesian optimization hyperparameters
    Tuning parameters for the Bayesian module to handle noisy image reconstruction objectives.
axioms (2)
  • domain assumption GEANT4 accurately models muon scattering and secondary particle production in cargo materials and detector volumes
    Invoked when using simulations to study material discrimination and optimize configurations.
  • standard math Cosmic muon flux and energy spectrum follow established models for the relevant latitude and altitude
    Background assumption for all tomography simulations.

pith-pipeline@v0.9.0 · 5789 in / 1458 out tokens · 27301 ms · 2026-05-19T05:02:11.483240+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

33 extracted references · 33 canonical work pages · 1 internal anchor

  1. [1]

    merlin.mbs aapmrev4-1.bst 2010-07-25 4.21a (PWD, AO, DPC) hacked

    FUNCTION id.bst "merlin.mbs aapmrev4-1.bst 2010-07-25 4.21a (PWD, AO, DPC) hacked" ENTRY address archive archivePrefix author bookaddress booktitle chapter collaboration doi edition editor eid eprint howpublished institution isbn issn journal key language month note number organization pages primaryClass publisher school SLACcitation series title translat...

    work page 2010

  2. [2]

    merlin.mbs aipauth4-1.bst 2010-07-25 4.21a (PWD, AO, DPC) hacked

    FUNCTION id.bst "merlin.mbs aipauth4-1.bst 2010-07-25 4.21a (PWD, AO, DPC) hacked" ENTRY address archive archivePrefix author bookaddress booktitle chapter collaboration doi edition editor eid eprint howpublished institution isbn issn journal key language month note number organization pages primaryClass publisher school SLACcitation series title translat...

    work page 2010

  3. [3]

    merlin.mbs aipnum4-1.bst 2010-07-25 4.21a (PWD, AO, DPC) hacked

    FUNCTION id.bst "merlin.mbs aipnum4-1.bst 2010-07-25 4.21a (PWD, AO, DPC) hacked" ENTRY address archive archivePrefix author bookaddress booktitle chapter collaboration doi edition editor eid eprint howpublished institution isbn issn journal key language month note number organization pages primaryClass publisher school SLACcitation series title translati...

    work page 2010

  4. [4]

    merlin.mbs apsrev4-1.bst 2010-07-25 4.21a (PWD, AO, DPC) hacked

    FUNCTION id.bst "merlin.mbs apsrev4-1.bst 2010-07-25 4.21a (PWD, AO, DPC) hacked" ENTRY address archive archivePrefix author bookaddress booktitle chapter collaboration doi edition editor eid eprint howpublished institution isbn issn journal key language month note number organization pages primaryClass publisher school SLACcitation series title translati...

    work page 2010

  5. [5]

    merlin.mbs apsrmp4-1.bst 2010-07-25 4.21a (PWD, AO, DPC) hacked

    FUNCTION id.bst "merlin.mbs apsrmp4-1.bst 2010-07-25 4.21a (PWD, AO, DPC) hacked" ENTRY address archive archivePrefix author bookaddress booktitle chapter collaboration doi edition editor eid eprint howpublished institution isbn issn journal key language month note number organization pages primaryClass publisher school SLACcitation series title translati...

    work page 2010

  6. [6]

    Borozdin et al

    Konstantin N. Borozdin et al. Radiographic imaging with cosmic-ray muons. Nature , 422:277, 2003

    work page 2003

  7. [7]

    Bongi, A

    Raffaello D'Alessandro, Fabio Ambrosino, Guglielmo Baccani, Lorenzo Bonechi, M. Bongi, A. Caputo, Roberto Ciaranfi, Luigi Cimmino, V. Ciulli, M. D'Errico, Flora Giudicepietro, Sandro Gonzi, Giovanni Macedonio, V. Masone, Barbara Melon, Newton Mori, Pasquale Noli, Massimo Orazi, P. Passeggio, and Lorenzo Viliani. Volcanoes in Italy and the role of muon rad...

    work page 2018

  8. [8]

    Alvarez et al

    Luis W. Alvarez et al. Search for hidden chambers in the pyramids. Science , 167:832, 1970

    work page 1970

  9. [9]

    Muon imaging: Present status and emerging applications

    International Atomic Energy Agency . Muon imaging: Present status and emerging applications. IAEA TECDOC 2012, IAEA, Vienna, 2022

    work page 2012

  10. [10]

    Cosmic-ray tomography for border security

    Sarah Barnes, Anzori Georgadze, Andrea Giammanco, Madis Kiisk, Vitaly A Kudryavtsev, Maxime Lagrange, and Olin Lyod Pinto. Cosmic-ray tomography for border security. Instruments , 7(1):13, 2023

    work page 2023

  11. [11]

    Navas et al

    S. Navas et al. Review of particle physics . Phys. Rev. D , 110(3):030001, 2024

    work page 2024

  12. [12]

    Toward the end-to-end optimization of particle physics instruments with differentiable programming

    Tommaso Dorigo et al. Toward the end-to-end optimization of particle physics instruments with differentiable programming . Rev. Phys. , 10:100085, 2023

    work page 2023

  13. [13]

    Progress in End-to-End Optimization of Detectors for Fundamental Physics with Differentiable Programming , 9 2023

    Max Aehle et al. Progress in End-to-End Optimization of Detectors for Fundamental Physics with Differentiable Programming , 9 2023

    work page 2023

  14. [14]

    Agostinelli et al

    S. Agostinelli et al. GEANT4 --- a simulation toolkit. Nucl. Inst. Meth. A , 506:250, 2003

    work page 2003

  15. [15]

    Strong et al

    Giles C. Strong et al. TomOpt: differential optimisation for task- and constraint-aware design of particle detectors in the context of muon tomography . Mach. Learn. Sci. Tech. , 5(3):035002, 2024

    work page 2024

  16. [16]

    Geometry description markup language for physics simulation and analysis applications

    Radovan Chytracek, Jeremy McCormick, Witold Pokorski, and Giovanni Santin. Geometry description markup language for physics simulation and analysis applications. Nuclear Science, IEEE Transactions on , 53:2892 -- 2896, 11 2006

    work page 2006

  17. [17]

    Blackwell and V.A

    T.B. Blackwell and V.A. Kudryavtsev. Simulation study into the identification of nuclear materials in cargo containers using cosmic rays. Journal of Instrumentation , 10(04):T04003, apr 2015

    work page 2015

  18. [18]

    Analysis of secondary particles as a complement to muon scattering measurements

    Maximilian Pérez Prada, Sarah Barnes, and Maurice Stephan. Analysis of secondary particles as a complement to muon scattering measurements. Instruments , 6(4):66, 2022

    work page 2022

  19. [19]

    Imaging by muons and their induced secondary particles—a novel technique

    G Galg \'o czi, D Mrdja, I Bikit, K Bikit, J Slivka, J Hansman, L Ol \'a h, G Hamar, and D Varga. Imaging by muons and their induced secondary particles—a novel technique. Journal of Instrumentation , 15(06):C06014, 2020

    work page 2020

  20. [20]

    Bacon, Konstantin N

    Jeffrey D. Bacon, Konstantin N. Borozdin, Joseph M. Fabritius, II, Christopher Morris, and John O. Perry. Muon induced fission neutrons in coincidence with muon tomography. Technical report, Los Alamos National Laboratory (LANL), Los Alamos, NM (United States), 10 2013

    work page 2013

  21. [21]

    Novel approach to imaging by cosmic-ray muons

    Istvan Bikit, Dusan Mrdja, Kristina Bikit, Jaroslav Slivka, Nikola Jovancevic, László Oláh, Gergő Hamar, and Dezső Varga. Novel approach to imaging by cosmic-ray muons. Europhysics Letters , 113(5):58001, mar 2016

    work page 2016

  22. [22]

    Python programming language

    Python Software Foundation . Python programming language. https://www.python.org/

    work page

  23. [23]

    Paszke et al

    A. Paszke et al. PyTorch: An Imperative Style, High-Performance Deep Learning Library . In Advances in Neural Information Processing Systems 32 , page 8024. Curran Associates, Inc., 2019

    work page 2019

  24. [24]

    A parametrization of the cosmic-ray muon flux at sea-level, 2015

    Mengyun Guan, Ming-Chung Chu, Jun Cao, Kam-Biu Luk, and Changgen Yang. A parametrization of the cosmic-ray muon flux at sea-level, 2015

    work page 2015

  25. [25]

    Energy and angular distributions of atmospheric muons at the earth, 2018

    Prashant Shukla and Sundaresh Sankrith. Energy and angular distributions of atmospheric muons at the earth, 2018

    work page 2018

  26. [26]

    Zaher et al

    Z. Zaher et al. Optimisation of muon portals for border controls using TomOpt . In MARESEC 2024 . Zenodo, June 2024

    work page 2024

  27. [27]

    Thomay et al

    C. Thomay et al. A binned clustering algorithm to detect high- Z material using cosmic muons. JINST , 8(10):P10013, 2013

    work page 2013

  28. [28]

    Strong, Pietro Vischia, and Haitham Zaraket

    Zahraa Zaher, Samuel Alvarez, Tommaso Dorigo, Andrea Giammanco, Maxime Lagrange, Giles C. Strong, Pietro Vischia, and Haitham Zaraket. Optimisation of muon tomography scanners for border control using tomopt. Particles , 8(2), 2025

    work page 2025

  29. [29]

    Mockus, Vytautas Tiesis, and Antanas Zilinskas

    J. Mockus, Vytautas Tiesis, and Antanas Zilinskas. The application of Bayesian methods for seeking the extremum , volume 2, pages 117--128. Springer Nature, 09 1978

    work page 1978

  30. [30]

    A Tutorial on Bayesian Optimization of Expensive Cost Functions, with Application to Active User Modeling and Hierarchical Reinforcement Learning

    Eric Brochu, Vlad Cora, and Nando Freitas. A tutorial on bayesian optimization of expensive cost functions, with application to active user modeling and hierarchical reinforcement learning. CoRR , abs/1012.2599, 12 2010

    work page internal anchor Pith review Pith/arXiv arXiv 2010

  31. [31]

    Jones, Matthias Schonlau, and William J

    Donald R. Jones, Matthias Schonlau, and William J. Welch. Efficient global optimization of expensive black-box functions. Journal of Global Optimization , 13(4):455--492, Dec 1998

    work page 1998

  32. [32]

    https://ax.dev/

    Meta Open Source . https://ax.dev/

    work page

  33. [33]

    Jiang, Samuel Daulton, Benjamin Letham, Andrew Gordon Wilson, and Eytan Bakshy

    Maximilian Balandat, Brian Karrer, Daniel R. Jiang, Samuel Daulton, Benjamin Letham, Andrew Gordon Wilson, and Eytan Bakshy. BoTorch: A Framework for Efficient Monte-Carlo Bayesian Optimization . In Advances in Neural Information Processing Systems 33 , 2020

    work page 2020