Pith. sign in

REVIEW 2 major objections 4 minor 29 references

Surrogate distributed radiological sources III: quantitative distributed source reconstructions

T0 review · 2 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read With eight surrogate distributed sources measured from drones, the paper shows that model-based gamma-ray reconstruction recovers source shape and, after one calibration factor, total activity to within about 15 percent.

desk verdict Solid experimental benchmark for distributed-source gamma-ray imaging; the shape claim holds, the absolute-activity claim is honest but less certain than the abstract suggests. read the letter →

arxiv 2412.02926 v3 pith:RHY5A7WG submitted 2024-12-04 physics.ins-det

Jayson R. Vavrek , Jaewon Lee , Marco Salathe , Mark S. Bandstra , Daniel Hellfeld , Brian J. Quiter , Tenzing H.Y. Joshi This is my paper · ORCID
classification physics.ins-det
keywords airbornegamma-rayimagingdistributedsourcereconstructionSceneDataFusionunmannedaerialsystemradiationsurveyCu-64surrogatesourcesMAP-EML1/2regularizationtotalvariation
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 is the third in a series that built surrogate distributed radiological sources out of dense arrays of Cu-64 point sources and measured them from drones carrying omnidirectional gamma-ray detectors. It claims that maximum-likelihood image reconstruction, applied to those aerial measurements, recovers the expected shapes of eight different distributed source patterns and, after applying a single global activity calibration, puts the absolute source activity within roughly 15 percent. The work matters because quantitative maps of distributed contamination, rather than qualitative hot-versus-cold images, are what would let emergency responders or regulators compare measured activity against limits and plan ground-level dose reductions. The paper also sweeps measurement and reconstruction parameters, showing which flight altitudes, line spacings, speeds, and regularization settings still yield faithful images.

What carries the argument

The central object is the linear forward model $\lambda = Vw$: the expected counts in each dwell interval equal a system matrix $V$ times a vector of non-negative source intensities on a $125 \times 80$ grid of 1 m pixels, where $V_{ik} = \eta_{ik} t_i \exp(-\mu_{\mathrm{air}}|\vec r_{ik}|)/(4\pi|\vec r_{ik}|^2)$ combines effective area, dwell time, inverse-square falloff, and air attenuation. Reconstruction solves the Poisson negative log-likelihood, optionally with a regularizer $\beta f(w)$, by expectation maximization, yielding ML-EM or MAP-EM updates weighted by a sensitivity map. Two regularizers are studied: the sparsity-promoting $L_{1/2}$ norm $f(w)=\sum_k \sqrt{w_k}$ and total variation, which smooths while preserving edges. The performance metrics are the total-activity ratio $R_{\mathrm{tot}}$, the normalized RMSE, and the structure coefficient $s$, a Pearson-like correlation between reconstructed and ground-truth pixel intensities, where the ground-truth images are 1 m continuous interpolations of the 4 m-spaced point-source arrays with the same average activity concentration.

What would settle it

Fly the same detectors and reconstruction code over a truly continuous distributed source whose activity is known from independent metrology, without re-fitting the calibration factor, and compare total reconstructed activity and shape metrics; the central claim fails if total activity deviates by more than about 15 percent or if the structure coefficient drops below the 0.87–0.94 band reported here.

Watch

Extended reading notes

Core claim

Using the linear model $\lambda = Vw$, in which each measurement is a Poisson sample of expected counts from a grid of image weights, and solving it with maximum-likelihood expectation maximization and sparsity- or smoothness-promoting regularizers, the paper reconstructs eight distributed Cu-64 source patterns measured by two airborne detector systems. After applying the activity calibration developed in Part II, the ratio of reconstructed to true total activity lies between 0.859 and 0.919, an absolute activity accuracy near 15 percent; the normalized RMSE ranges from 0.151 to 0.380, and the structure coefficient ranges from 0.870 to 0.935, indicating the shapes are reproduced with high perceived similarity. Interior activities tend to be over-predicted and edges under-predicted, corners are rounded over a few meters, and sharp zero-activity corridors are smoothed. The paper also reports that image quality degrades monotonically with detector altitude, with raster line spacing beyond about 8 m, and with flight speed beyond about 8 m/s, while replacing the full anisotropic detector response by an isotropic model of the same total area has only a small effect.

Load-bearing premise

The load-bearing premise is that the 4 m-spaced point-source arrays, interpolated to 1 m continuous ground-truth images, faithfully represent truly continuous distributed sources under the aerial measurement geometry; the quantitative accuracy claims are measured against that synthetic truth, and the ~1.4x activity calibration carries substantial calibration uncertainty.

Editorial extensions

If this is right

  • Regulators and emergency responders can compare reconstructed activity directly against concentration limits, not just qualitative hot/cold patterns.
  • For source extents near 1000 m², flight altitudes around 6 m, raster spacings around 5 m, and speeds at or below 8 m/s keep the quality metrics in the demonstrated range.
  • The full anisotropic detector response can be replaced by an equal-area isotropic model with only small quality loss in these airborne surveys.
  • Dense point-source arrays remain a valid experimental stand-in for continuous sources, so future imaging studies can use reconfigurable arrays and still benchmark against known truth.

Reading between the lines

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

  • Editorial extension: If the point-source-array surrogate transfers to real contamination fields, the same pipeline becomes an operational decision-support tool, letting aerial surveys feed regulatory comparisons and ground-dose planning directly.
  • Editorial extension: Because the reconstructions run in 1–2 s on a laptop GPU and degrade gracefully at a 3× speedup, near-real-time adaptive surveying, in which raster lines are re-planned while the drone is still airborne, is a plausible next step.
  • Editorial extension: The small difference between anisotropic and isotropic detector-response models suggests that in this flat-terrain geometry the largest gains in reconstruction quality would come from trajectory design and photon statistics, not from more detailed detector modeling.
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

2 major / 4 minor

Summary. This paper presents quantitative image reconstruction results from aerial measurements of eight surrogate distributed gamma-ray sources, each constructed as an array of Cu-64 point sources. The authors use singles-mode Scene Data Fusion with ML-EM and MAP-EM reconstructions, including L1/2 and total-variation regularizers, and evaluate reconstruction quality against synthetic ground-truth images using three metrics: total-activity ratio Rtot, NRMSE, and a structure coefficient. They report shape reconstructions with structure coefficients 0.87–0.94 and NRMSE 0.15–0.38, and claim that the absolute activity scale can be determined to within ~15% after applying a ~1.4× calibration factor derived in Part II. The paper also includes parametric studies of detector altitude, flight speed, raster spacing, data/response coarsening, and regularization strength.

Significance. If the stated claims hold, this is a valuable contribution to quantitative aerial gamma-ray imaging, providing a systematic experimental benchmark for distributed-source reconstruction and demonstrating that shape-faithful reconstructions are achievable with lightweight UAS-borne detectors. The use of externally designed source geometries gives independent support to the shape-accuracy claims, and the parameter sweeps offer practical guidance for survey design. However, the absolute-activity claim is not fully supported because it rests on an unquantified calibration factor, and the quantitative metrics are computed against synthetic ground truth that inherits the same calibration uncertainty. The paper is nonetheless a solid experimental study with reproducible methodology and clear presentation of the reconstruction machinery.

major comments (2)
  1. [Section II-A, Eq. (11), Conclusion] The headline quantitative claim that the absolute activity scale is determined to within ~15% is not supported at the stated confidence. The ~1.4× calibration factor is described in Section II-A as having 'substantial calibration uncertainties,' but no uncertainty is ever assigned to this factor or propagated into the reported activity ratios. Because the ground-truth images used in Eq. (11) are built from the same nominal source activities that carry this calibration uncertainty, the observed Rtot range of 0.859–0.919 demonstrates run-to-run precision and internal consistency, not accuracy relative to an independent activity standard. The discussion in Section IV noting that a subsequent JHU APL campaign required no such correction factor further indicates that this is a campaign-specific normalization. To support the 'within ~15%' claim, the authors should either provide a quantitative uncertainty on the 1.4× factor and propagate it into the reported Rtot values, or provide an independent assay of a subset of the point sources.
  2. [Section II-D] The 'true' images used in all three quantitative metrics are generated by interpolating the discrete point-source arrays to a 1-m continuous grid, and these interpolated images are treated as exact ground truth. This construction inherits both the calibration uncertainty of the nominal source activities and the modeling assumption that 4-m-spaced point sources faithfully represent a continuous distributed source at the measurement geometry. The paper should explicitly state that all reported values of Rtot, NRMSE, and the structure coefficient are relative to this synthetic surrogate truth, and should discuss how deviations from the continuity assumption would affect the metrics. This is load-bearing for the transferability of the quantitative accuracy claims to real continuous contamination and should be addressed directly.
minor comments (4)
  1. [Section II-D, Eq. (13)] The definition of the structure coefficient appears to contain a typographical error: the stabilizing constant in the numerator is written as εp and in the denominator as ε, but no distinction between these two constants is explained. Please clarify the intended formula, likely consistent with the original SSIM definition.
  2. [Section III-A] The phrase 'Binmode MAP-EM reconstructions' is used without definition; please clarify that this means bin-mode (time-binned) processing as opposed to list-mode, and ensure the terminology is introduced in Section II.
  3. [Section IV] In the sentence 'the change in source activity due to radioactive decay during a given measurement is small ( 1% over a 10 minute flight)', the inequality sign appears to be missing; it should read '<1%'.
  4. [Section III-E] The speed-emulation methodology is described as 'compressing the radiation and trajectory timestamps by a constant factor'; it would be helpful to state explicitly that this preserves the spatial sampling pattern of the trajectory while reducing counts per dwell time, and to note any possible artifacts from non-constant readout intervals.

Circularity Check

1 steps flagged · score 4.0 of 10

Absolute-activity claim is calibrated via a ~1.4× factor fitted to the same campaign; shape reconstructions are independently checked, but the 'within ~15%' result is not an independent absolute-scale test.

  1. fitted input called prediction [Section II-A (Measurement overview) and Section V (Conclusion)]
    "the count rate comparisons made in Part II Section V-A suggest that the true source activities were ~1.4x higher than the planned nominal activities, which were subject to substantial calibration uncertainties. In this work, we consider this ~1.4x scale factor to be part of the overall system calibration. ... The absolute activity scale can also be determined to within ~15% after applying the activity calibration developed in Part II."

    The paper's central quantitative claim about absolute activity is not an independent prediction: the single ~1.4x factor is fitted to count-rate comparisons from the same WSU campaign (Part II) and then adopted as the absolute scale for both the ground-truth totals and the reconstructed activity scale. Since Rtot (Eq. 11) compares reconstructed totals to ground-truth totals whose absolute scale is set by that same fitted factor, the reported near-unity Rtot values (0.859-0.919) are in-sample residuals after applying a fitted calibration.

full rationale

The image-reconstruction methodology itself is not circular: the forward model (Eqs. 1-2), ML-EM/MAP-EM inversion (Eqs. 5-7), and the shape metrics (Eqs. 11-13) are standard, and the shape comparisons are made against known on-field point-source layouts, which are externally designed configurations rather than outputs of the reconstruction. The primary circularity concern is limited to the absolute-activity claim, which depends on the ~1.4x calibration factor imported from Part II. Because that factor is derived from the same campaign and its uncertainty is unquantified, the 'absolute activity to within ~15%' statement is a calibrated in-sample result, not a blind test. The ground-truth images are obtained by interpolating the point-source arrays to a continuous 1-m grid (Sec. II-D); this is a self-consistency check of the reconstruction against the array layout rather than an independent test of the arrays' fidelity as continuous-source surrogates, but that surrogate property is established in Part I and is secondary to the imaging claims. Overall, the shape and parametric-trend studies have independent content; the absolute-scale headline is partially circular because it is fitted, so the score is 4 rather than 0 or 6.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central quantitative claims rest primarily on a global calibration scale factor (~1.4x) derived from the same experimental campaign, on the surrogate point-source array ground truth from the authors' Part I design, and on manually chosen regularization hyperparameters. No new physical entities are introduced.

free parameters (4)
  • Global activity calibration scale factor = ~1.4 (dimensionless)
    Applied to all source activities to convert nominal to true activities, derived from count-rate comparisons in Part II Section V-A. It is a global multiplicative fit to the measurement campaign and directly sets the absolute activity scale in the Rtot comparisons.
  • L1/2 regularization strength beta = 3 x 10^-3 (representative), 10^-4 and 10^-2 in sweeps
    Chosen manually from prior reconstruction experience (Section III-A); affects sharpness and total activity of reconstructions.
  • Number of MAP-EM iterations = 30 (default), 10 to 1000 in sweeps
    Pre-set number of iterations; no formal stopping criterion; under-convergence seen at 10 iterations.
  • TV regularization strength beta and stabilization epsilon = beta in [10^-7, 3x10^-7], epsilon = 2.4x10^5 gamma/s
    For the TV regularizer study (Section III-F), epsilon set per Ref [18, Section II-C], beta chosen to avoid checkerboard artifacts.
assumptions (5)
  • domain assumption The Poisson linear forward model lambda = V w with 1/r^2 attenuation and exponential air attenuation (Eqs. 1 to 3) accurately describes the measurement process.
    Standard radiation transport and detection statistics; the basis of the reconstruction.
  • domain assumption The source activity is confined to a flat plane at z=0 and the reconstructed image space is 125x80 m with 1 m pixels (Section II-B).
    Assumes no volumetric source distribution or significant terrain effects; true for the WSU flat terrain experiment but limits generalization to hilly terrain.
  • ad hoc to paper Point-source arrays spaced on a 4 m grid are faithful surrogates for continuous distributed sources for the detectors and trajectories used (Section II-D, relying on Part I design).
    The ground truth images are interpolations of these arrays; this premise is inherited from the authors' own design paper and is not independently validated against a real continuous source.
  • ad hoc to paper The ~1.4x activity calibration factor derived from count-rate comparisons in Part II is an accurate global correction for nominal source activities (Section II-A).
    The absolute activity claim depends on this factor, which has substantial calibration uncertainties.
  • domain assumption SLAM lidar co-registration with mean 6.2 cm point error is sufficient for 1-m pixel reconstruction (Section II-B).
    Co-registration error is much smaller than the pixel size, so its effect on the system matrix is assumed negligible.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Surrogate distributed radiological sources III: quantitative distributed source reconstructions." pith.science (2026). https://pith.science/paper/RHY5A7WG

@misc{pith2026241202926,
  author       = {Pith},
  title        = {Pith review of: Surrogate distributed radiological sources III: quantitative distributed source reconstructions},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RHY5A7WG}},
  note         = {Machine review of arXiv:2412.02926}
}
read the original abstract

In this third part of a multi-paper series, we present quantitative image reconstruction results from aerial measurements of eight different surrogate distributed gamma-ray sources on flat terrain. We show that our quantitative imaging methods can accurately reconstruct the expected shapes, and, after appropriate calibration, the absolute activity of the distributed sources. We conduct several studies of imaging performance versus various measurement and reconstruction parameters, including detector altitude and raster pass spacing, data and modeling fidelity, and regularization type and strength. The imaging quality performance is quantified using various quantitative image quality metrics. Our results confirm the utility of point source arrays as surrogates for truly distributed radiological sources, and advance the quantitative capabilities of Scene Data Fusion gamma-ray imaging methods.

Figures

Figures reproduced from arXiv: 2412.02926 by the authors.

Figure 1
Figure 1. Left: top-down view of the (un-leveled) square source reference point cloud in CloudCompare, with points colorized by height. The larger yellow [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. 35 lidar SLAM point clouds from the WSU aerial measurement campaign, all co-registered and aligned to the idealized field coordinate system of Part I. Top: xy projection. The darker spots in the center-right of the field are traffic cones placed around the boundary of the 10 × 10 square and L-shape sources. The circular patterns near (x, y) = (20, 10) and (70, 75) m are lidar artifacts from takeoff and landing. Midd… view at source ↗
Figure 3
Figure 3. Imaging study for four surrogate distributed sources (left to right: the [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: Like Fig. 3, but for the other four source distributions, each of which featured areas of heightened source activity. Left to right: the plume, the [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Top: MAP-EM-reconstructed (black) and measured (orange) counts [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: 3D Scene Data Fusion perspective render of the reconstruction result for the [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 8
Figure 8. Figure 8: Replication study with the square source. Note that the top left [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]
Figure 10
Figure 10. Figure 10: Speed study with the 12 m-separation source [PITH_FULL_IMAGE:figures/full_fig_p009_10.png]
Figure 9
Figure 9. Figure 9: Coarse-graining study with the 12 m-separation source. The left column shows image results with increasingly long time-discretization intervals ti for an analysis of all MiniPRISM detector modules’ angular sensitivity η treated individually, while the right column show…
Figure 12
Figure 12. Figure 12: L1/2 regularizer study with the L-shape source. discussion of the TV vs L1/2 regularizer performance is given in Section IV [PITH_FULL_IMAGE:figures/full_fig_p010_12.png]
Figure 13
Figure 13. Figure 13: TV regularizer study with the hot/coldspot source. [PITH_FULL_IMAGE:figures/full_fig_p010_13.png]
Figure 14
Figure 14. Figure 14: Raster spacing study with the L-shape source, showing the reconstructed weights (left and center columns) and sensitivities (right column) when the raster lines are successively cut. Note that the color scales vary among rows and columns in order to better highlight t…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

29 extracted references · 27 canonical work pages

  1. [1]

    Surrogate distributed radiological sources I: point- source array design methods

    Jayson R Vavrek, Mark S Bandstra, Daniel Hellfeld, Brian J Quiter, and Tenzing HY Joshi. Surrogate distributed radiological sources I: point- source array design methods. IEEE Transactions on Nuclear Science , 2024

    work page 2024

  2. [2]

    Surrogate distributed radiological sources II: aerial measurement campaign

    Jayson R Vavrek, C Corey Hines, Mark S Bandstra, Daniel Hellfeld, Maddison A Heine, Zachariah M Heiden, Nick R Mann, Brian J Quiter, and Tenzing HY Joshi. Surrogate distributed radiological sources II: aerial measurement campaign. IEEE Transactions on Nuclear Science , 2024

    work page 2024

  3. [3]

    Free-moving quantitative gamma-ray imaging

    D Hellfeld, MS Bandstra, JR Vavrek, DL Gunter, JC Curtis, M Salathe, R Pavlovsky, V Negut, PJ Barton, JW Cates, et al. Free-moving quantitative gamma-ray imaging. Scientific reports, 11(1):1–14, 2021

    work page 2021

  4. [4]

    Advances in nuclear radiation sensing: Enabling 3-D gamma-ray vision

    Kai Vetter, Ross Barnowski, Joshua W Cates, Andrew Haefner, Ten- zing HY Joshi, Ryan Pavlovsky, and Brian J Quiter. Advances in nuclear radiation sensing: Enabling 3-D gamma-ray vision. Sensors, 19(11):2541, 2019

    work page 2019

  5. [5]

    Tomographic reconstruction of a spatially-extended source from the perimeter of a restricted-access zone using a SCoTSS compton gamma imager

    NJ Murtha, LE Sinclair, PRB Saull, A McCann, and AML MacLeod. Tomographic reconstruction of a spatially-extended source from the perimeter of a restricted-access zone using a SCoTSS compton gamma imager. Journal of Environmental Radioactivity , 240:106758, 2021

    work page 2021

  6. [6]

    3-D reconstruction of extended sources during dispersal trial with a Compton imager

    A MacLeod, N Murtha, P Saull, L Sinclair, and M Andrew. 3-D reconstruction of extended sources during dispersal trial with a Compton imager. In 2023 IEEE Nuclear Science Symposium, Medical Imaging Conference and International Symposium on Room-Temperature Semi- conductor Detectors (NSS MIC RTSD) . IEEE, 2023

    work page 2023

  7. [7]

    3D Compton mapping of point and distributed sources with moving detectors

    CG Wahl, R Sobota, and D Goodman. 3D Compton mapping of point and distributed sources with moving detectors. In 2023 IEEE Nuclear Science Symposium, Medical Imaging Conference and International Symposium on Room-Temperature Semiconductor Detectors (NSS MIC RTSD). IEEE, 2023

    work page 2023

  8. [8]

    Extended sources reconstructions by means of coded mask aperture systems and deep learning algorithm

    G Daniel and O Limousin. Extended sources reconstructions by means of coded mask aperture systems and deep learning algorithm. Nuclear Instruments and Methods in Physics Research Section A: Accelera- tors, Spectrometers, Detectors and Associated Equipment, 1012:165600, 2021. IEEE TRANSACTIONS ON NUCLEAR SCIENCE 13

    work page 2021

Show all 29 references
  1. [9]

    3D Gamma- ray and Neutron Mapping in Real-Time with the Localization and Mapping Platform from Unmanned Aerial Systems and Man-Portable Configurations

    R Pavlovsky, JW Cates, WJ Vanderlip, THY Joshi, A Haefner, E Suzuki, R Barnowski, V Negut, A Moran, K Vetter, et al. 3D Gamma- ray and Neutron Mapping in Real-Time with the Localization and Mapping Platform from Unmanned Aerial Systems and Man-Portable Configurations. arXiv:19...

    1908 arXiv

  2. [10]

    R. T. Pavlovsky, J. W. Cates, M. Turqueti, D. Hellfeld, V . Negut, A. Moran, P. J. Barton, K. Vetter, and B. J. Quiter. MiniPRISM: 3D Realtime Gamma-ray Mapping from Small Unmanned Aerial Systems and Handheld Scenarios. In 2019 IEEE Nuclear Science Symposium and Medical Imagin...

    2019

  3. [11]

    Simultaneous localization and mapping: part I

    Hugh Durrant-Whyte and Tim Bailey. Simultaneous localization and mapping: part I. IEEE robotics & automation magazine , 13(2):99–110, 2006

    2006

  4. [12]

    Simultaneous localization and mapping (SLAM): Part II

    Tim Bailey and Hugh Durrant-Whyte. Simultaneous localization and mapping (SLAM): Part II. IEEE robotics & automation magazine , 13(3):108–117, 2006

    2006

  5. [13]

    Open3D: A modern library for 3D data processing

    Qian-Yi Zhou, Jaesik Park, and Vladlen Koltun. Open3D: A modern library for 3D data processing. arXiv:1801.09847, 2018

    2018 arXiv

  6. [14]

    L. A. Shepp and Y . Vardi. Maximum Likelihood Reconstruction for Emission Tomography. IEEE Trans. on Medical Imaging , 1(2):113– 122, 1982

    1982

  7. [15]

    EM reconstruction algorithms for emission and transmission tomography

    Kenneth Lange, Richard Carson, et al. EM reconstruction algorithms for emission and transmission tomography. J Comput Assist Tomogr , 8(2):306–16, 1984

    1984

  8. [16]

    L1/2 regularization

    ZongBen Xu, Hai Zhang, Yao Wang, XiangYu Chang, and Yong Liang. L1/2 regularization. Science China Information Sciences , 53(6):1159– 1169, 2010

    2010

  9. [17]

    Nonlinear total varia- tion based noise removal algorithms

    Leonid I Rudin, Stanley Osher, and Emad Fatemi. Nonlinear total varia- tion based noise removal algorithms. Physica D: Nonlinear Phenomena, 60(1-4):259–268, 1992

    1992

  10. [18]

    Total variation regulated EM algorithm

    VY Panin, GL Zeng, and GT Gullberg. Total variation regulated EM algorithm. In 1998 IEEE Nuclear Science Symposium Conference Record, volume 3, pages 1562–1566. IEEE, 1998

    1998

  11. [19]

    Multi-modal Free- moving Data Fusion (MFDF) v1.0, 2020

    THY Joshi, BJ Quiter, J Curtis, MS Bandstra, R Cooper, D Hellfeld, M Salathe, A Moran, J Vavrek, Department of Homeland Security, DOD Defense Threat Reduction Agency, and USDOE. Multi-modal Free- moving Data Fusion (MFDF) v1.0, 2020

    2020

  12. [20]

    radkit v1.2

    Tenzing Joshi, Brian Quiter, Joey Curtis, Micah Folsom, Mark Bandstra, Nicolas Abgrall, Reynold Cooper, Daniel Hellfeld, Marco Salathe, Alex Moran, et al. radkit v1.2. Technical report, Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States), 2021

    2021

  13. [21]

    Image quality assessment: from error visibility to structural similarity

    Zhou Wang, Alan C Bovik, Hamid R Sheikh, and Eero P Simoncelli. Image quality assessment: from error visibility to structural similarity. IEEE transactions on image processing , 13(4):600–612, 2004

    2004

  14. [22]

    Demonstration of a new CLLBC-based gamma- and neutron- sensitive free-moving omnidirectional imaging detector

    JR Vavrek, R Pavlovsky, V Negut, D Hellfeld, THY Joshi, BJ Quiter, and JW Cates. Demonstration of a new CLLBC-based gamma- and neutron- sensitive free-moving omnidirectional imaging detector. In preparation for Sensors, 2024

    2024

  15. [23]

    Higher degree total variation (HDTV) reg- ularization for image recovery

    Yue Hu and Mathews Jacob. Higher degree total variation (HDTV) reg- ularization for image recovery. IEEE Transactions on Image Processing, 21(5):2559–2571, 2012

    2012

  16. [24]

    Stopping criteria for ending autonomous, single detector radiological source searches

    Gregory R Romanchek and Shiva Abbaszadeh. Stopping criteria for ending autonomous, single detector radiological source searches. PLOS One, 16(6):e0253211, 2021

    2021

  17. [25]

    A successive-elimination approach to adaptive robotic source seeking

    Esther Rolf, David Fridovich-Keil, Max Simchowitz, Benjamin Recht, and Claire Tomlin. A successive-elimination approach to adaptive robotic source seeking. IEEE Transactions on Robotics , 37(1):34–47, 2020

    2020

  18. [26]

    Multi-sensor optimal motion planning for radiological contamination surveys by using prediction-difference maps

    Tony H Shin, Daniel T Wakeford, and Suzanne F Nowicki. Multi-sensor optimal motion planning for radiological contamination surveys by using prediction-difference maps. Applied Sciences, 12(11):5627, 2022

    2022

  19. [27]

    A statistical stopping rule for MLEM reconstructions in PET

    Nicolai Bissantz, Bernard A Mair, and Axel Munk. A statistical stopping rule for MLEM reconstructions in PET. In 2008 IEEE Nuclear Science Symposium Conference Record , pages 4198–4200. IEEE, 2008

    2008

  20. [28]

    A novel MLEM stopping criterion for unfold- ing neutron fluence spectra in radiation therapy

    Logan Montgomery, Anthony Landry, Georges Al Makdessi, Felix Mathew, and John Kildea. A novel MLEM stopping criterion for unfold- ing neutron fluence spectra in radiation therapy. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detec...

    2020

  21. [29]

    A heuristic statistical stopping rule for iterative reconstruction in emis- sion tomography

    F Ben Bouall `egue, Jean-Francois Crouzet, and Denis Mariano-Goulart. A heuristic statistical stopping rule for iterative reconstruction in emis- sion tomography. Annals of nuclear medicine , 27:84–95, 2013

    2013

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.