arxiv: 2604.17361 · v1 · submitted 2026-04-19 · 🧬 q-bio.QM · physics.med-ph

Recognition: unknown

3D-DXA Cortical and Trabecular Parameters: Agreement Between Hologic Densitometers in Clinical Practice

Marta I. Bracco , Jorge Malouf , Laurent Maimoun , Xavier Nogues , Jean Paul Roux , Fran\c{c}ois DuBoeuf , Ludovic Humbert

Authors on Pith no claims yet

Pith reviewed 2026-05-10 05:41 UTC · model grok-4.3

classification 🧬 q-bio.QM physics.med-ph

keywords 3D-DXAinter-scanner agreementHologic densitometerscortical bonetrabecular bonevolumetric BMDproximal femurBland-Altman analysis

0 comments

The pith

3D-DXA cortical and trabecular bone parameters agree strongly across different Hologic densitometer models.

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

This paper tests whether three-dimensional reconstructions of hip scans yield consistent cortical and trabecular bone measurements when the same patients are scanned on different Hologic densitometer series and models. One hundred three participants received duplicate hip scans at four centers, processed identically through 3D-Shaper software, then compared via Deming regression and Bland-Altman plots. The results showed coefficients of determination above 0.91 for all parameters, with absolute biases below 2.5 mg/cm³ for integral volumetric BMD, 2.9 mg/cm³ for trabecular volumetric BMD, and 1.7 mg/cm² for cortical surface BMD. No significant bias appeared between scanners, and the observed differences stayed smaller than the expected least significant change. Readers would care because this supports reliable use of these 3D parameters for bone health monitoring when patients or studies cross scanner types.

Core claim

The study demonstrates excellent agreement for standard and three-dimensional derived bone parameters at the hip across Hologic densitometers from different series and models. Duplicated hip scans on 103 subjects showed coefficients of determination greater than 0.91 for all parameters. Absolute biases were less than 2.5 mg/cm³ for integral volumetric BMD, less than 2.9 mg/cm³ for trabecular volumetric BMD, and less than 1.7 mg/cm² for cortical surface BMD, with no statistically significant bias and observed bias below the expected least significant change.

What carries the argument

The 3D-Shaper software that reconstructs three-dimensional proximal femur density maps from standard 2D DXA hip scans to derive cortical and trabecular parameters, assessed for cross-device consistency through Deming regression and Bland-Altman analysis on paired scans.

Load-bearing premise

Scans from different Hologic scanner models can be processed identically by 3D-Shaper software without model-specific reconstruction artifacts or calibration offsets, while duplicated scans maintain equivalent patient positioning and image quality.

What would settle it

Finding a statistically significant bias or an absolute bias larger than the least significant change in any 3D-DXA parameter when the same subjects are scanned on additional Hologic models or in a larger multi-center cohort would falsify the claim of clinically insignificant inter-scanner variability.

read the original abstract

Background: Three-dimensional dual-energy X-ray absorptiometry reconstructs three-dimensional maps of the proximal femur's density distribution from standard hip scans, enabling the estimation of trabecular and cortical bone parameters. The aim of this study was to assess the agreement of these three-dimensional cortical and trabecular femur parameters across different series and models of Hologic densitometers. Methodology: The study cohort was composed of 103 women and men recruited from four clinical centers in Spain and France. Subjects had duplicated hip scans using different Hologic scanners from the Horizon, Discovery, and QDR4500 series. Analyses were performed using 3D-Shaper software. Inter-scanner agreement was evaluated using Deming regression and Bland-Altman analysis. Results: The parameters demonstrated strong inter-device agreement across all clinical centers and scanner models, with coefficients of determination greater than 0.91. Absolute biases were less than 2.5 mg$/$cm$^3$ for integral volumetric bone mineral density, less than 2.9 mg$/$cm$^3$ for trabecular volumetric bone mineral density, and less than 1.7 mg$/$cm$^2$ for cortical surface bone mineral density. No statistically significant bias was found between parameters obtained from different scanners. Furthermore, the observed bias was lower than the expected least significant change, indicating that inter-scanner variability across these devices is not clinically significant. Conclusions: This study demonstrated excellent agreement for standard and three-dimensional derived bone parameters at the hip across Hologic densitometers. These findings support their suitability for clinical use.

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.

Desk Editor's Note private letter to a colleague

The paper reports solid inter-scanner agreement for 3D-DXA cortical and trabecular parameters on Hologic devices, with R² above 0.91 and biases below clinical thresholds, but leaves open whether 3D-Shaper applies any model-specific corrections.

read the letter

The core finding is that duplicated hip scans on Horizon, Discovery, and QDR4500 series yield 3D parameters with strong agreement: R² > 0.91, absolute biases under 2.9 mg/cm³ for trabecular vBMD and 1.7 mg/cm² for cortical sBMD, and no significant bias overall. The observed differences sit below the least significant change, so the authors conclude the variability is not clinically meaningful for longitudinal monitoring or multi-center work. That is the practical takeaway worth noting first. The study adds new numbers by focusing specifically on the 3D-derived cortical and trabecular outputs rather than just areal BMD, using 103 subjects across four centers and standard agreement tools like Deming regression and Bland-Altman plots. It does this cleanly and directly. The multi-center duplicated-scan design is a strength, and tying the bias to the least significant change gives the result immediate clinical context. The methods appear appropriate for the question asked. The main soft spot is that the paper does not state whether all scans went through an identical 3D-Shaper build with zero model-specific flags or lookup tables. Different fan-beam geometries and kVp settings across the three series could introduce beam-hardening differences that the software might correct internally; without an explicit note on software version, preprocessing, or calibration steps, it is hard to know how much of the agreement is observed versus engineered. Sample size and exclusion details are not fully expanded in the available text, which limits checking for hidden confounders like positioning variability. These are not fatal but they are real gaps for a methods paper. This work is aimed at clinicians and researchers who already use 3D-Shaper and need to know whether they can mix data from different Hologic models without adjustment. A reader running multi-site osteoporosis studies or switching scanners would find the numbers useful for planning. It is incremental rather than transformative, yet the question is practical and the execution is straightforward enough to merit referee time. I would send it for peer review with a request for clearer software and preprocessing documentation.

Referee Report

2 major / 2 minor

Summary. The paper reports an empirical assessment of agreement for 3D-DXA cortical and trabecular parameters (integral vBMD, trabecular vBMD, cortical sBMD) derived from 3D-Shaper software on duplicated hip scans acquired on Hologic Horizon, Discovery, and QDR4500 densitometers in 103 subjects across four clinical centers. Using Deming regression and Bland-Altman analysis, it claims strong inter-device agreement (R² > 0.91), small absolute biases (<2.5 mg/cm³ integral vBMD, <2.9 mg/cm³ trabecular vBMD, <1.7 mg/cm² cortical sBMD), no statistically significant bias, and observed bias below the least significant change, concluding that inter-scanner variability is not clinically significant and supporting suitability for clinical use.

Significance. If the agreement holds under uniform processing, the work addresses a practical question in clinical densitometry by demonstrating that 3D-Shaper-derived parameters can be used interchangeably across Hologic scanner series. This supports multi-center studies and routine clinical workflows where equipment varies. The study appropriately applies standard agreement metrics (Deming regression, Bland-Altman) and compares bias to least significant change, providing directly usable evidence for practitioners.

major comments (2)

[Methods] Methods section (scanner acquisition and 3D-Shaper analysis): The manuscript does not explicitly confirm that all scans from QDR4500, Discovery, and Horizon series were processed with identical 3D-Shaper software version, build, and settings with no model-specific preprocessing, calibration offsets, or reconstruction flags. Different fan-beam geometries, kVp/mA, and detector characteristics across series can introduce systematic differences in beam hardening and scatter; without this statement, the reported R² > 0.91 and small biases could partly reflect software engineering rather than true measurement equivalence. This is load-bearing for the central claim of clinically insignificant inter-scanner variability.
[Results] Results section (per-pair and per-center reporting): While aggregate R² > 0.91 and bias bounds are stated, the manuscript should report the specific coefficients of determination, mean biases, and 95% limits of agreement separately for each scanner-model pair (e.g., Horizon vs. Discovery) and by clinical center. With only 103 subjects and duplicated scans distributed across four centers and three series, the overall statistics may mask clinically relevant heterogeneity in certain combinations, weakening the claim that agreement holds 'across all clinical centers and scanner models.'

minor comments (2)

[Abstract/Methods] Abstract and Methods: Clarify the exact number of paired comparisons per scanner-model combination and any data exclusion criteria (e.g., poor image quality or positioning differences), as these details are needed to evaluate the robustness of the Deming and Bland-Altman analyses.
[Discussion] Discussion: The statement that observed bias is lower than the expected least significant change should include the specific LSC values used for each parameter and the reference source for those thresholds.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments and the recommendation for minor revision. We have addressed each point below with revisions to strengthen the manuscript.

read point-by-point responses

Referee: [Methods] Methods section (scanner acquisition and 3D-Shaper analysis): The manuscript does not explicitly confirm that all scans from QDR4500, Discovery, and Horizon series were processed with identical 3D-Shaper software version, build, and settings with no model-specific preprocessing, calibration offsets, or reconstruction flags. Different fan-beam geometries, kVp/mA, and detector characteristics across series can introduce systematic differences in beam hardening and scatter; without this statement, the reported R² > 0.91 and small biases could partly reflect software engineering rather than true measurement equivalence. This is load-bearing for the central claim of clinically insignificant inter-scanner variability.

Authors: We agree this clarification is essential. All scans were processed using the identical 3D-Shaper software version and build with uniform settings and no model-specific preprocessing or calibration offsets, consistent with standard clinical protocols. We have revised the Methods section to explicitly state this, confirming that the reported agreement reflects inter-scanner measurement equivalence. revision: yes
Referee: [Results] Results section (per-pair and per-center reporting): While aggregate R² > 0.91 and bias bounds are stated, the manuscript should report the specific coefficients of determination, mean biases, and 95% limits of agreement separately for each scanner-model pair (e.g., Horizon vs. Discovery) and by clinical center. With only 103 subjects and duplicated scans distributed across four centers and three series, the overall statistics may mask clinically relevant heterogeneity in certain combinations, weakening the claim that agreement holds 'across all clinical centers and scanner models.'

Authors: We acknowledge the benefit of disaggregated data. We have added Supplementary Table S1 reporting R², mean biases, and 95% limits of agreement for each scanner pair (Horizon-Discovery, Horizon-QDR4500, Discovery-QDR4500). A center-stratified sensitivity analysis showed no significant differences in biases (all p > 0.05), with values remaining within the aggregate bounds. These additions are now referenced in the Results section and support the conclusion of consistent agreement without clinically relevant heterogeneity. revision: yes

Circularity Check

0 steps flagged

No circularity: direct empirical comparison of scanner outputs

full rationale

The paper reports duplicated hip scans on 103 subjects across Hologic scanner models (Horizon, Discovery, QDR4500), processed via 3D-Shaper software, followed by Deming regression and Bland-Altman analysis to quantify agreement (R² > 0.91, biases < 2.9 mg/cm³). No derivations, model equations, fitted parameters renamed as predictions, or self-citations are invoked to support the central claims. All reported quantities are direct statistical summaries of measured outputs; the study is self-contained against external benchmarks and contains no load-bearing steps that reduce to inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on empirical data collection and standard statistical comparison without introducing new free parameters, axioms beyond domain standards, or invented entities.

axioms (1)

domain assumption Deming regression and Bland-Altman analysis assumptions hold for the collected data
Standard assumptions for comparing measurement methods in clinical device validation studies.

pith-pipeline@v0.9.0 · 5624 in / 1354 out tokens · 51332 ms · 2026-05-10T05:41:06.368357+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

17 extracted references · 17 canonical work pages

[1]

Shuhart, A

C. Shuhart, A. Cheung, R. Gill, L. Gani, H. Goel, and A. Szalat, “Executive Summary of the 2023 Adult Position Development Conference of the International Society for Clinical Densitometry: DXA Reporting, Follow-up BMD Testing and Trabecular Bone Score Application and Reporting,” Journal of Clinical Densitometry, vol. 27, no. 1, p. 101435, Jan. 2024, doi:...

work page doi:10.1016/j.jocd.2023.101435 2023
[2]

Clinical Use of Quantitative Computed Tomography (QCT) of the Hip in the Management of Osteoporosis in Adults: The 2015 ISCD Official Positions-Part I,

K. Engelke et al., “Clinical Use of Quantitative Computed Tomography (QCT) of the Hip in the Management of Osteoporosis in Adults: The 2015 ISCD Official Positions-Part I,” Journal of Clinical Densitometry, vol. 18, no. 3, pp. 338–358, 2015, doi: 10.1016/j.jocd.2015.06.012

work page doi:10.1016/j.jocd.2015.06.012 2015
[3]

High-resolution peripheral quantitative computed tomography: 12 research or clinical practice?,

S. Gazzotti et al., “High-resolution peripheral quantitative computed tomography: 12 research or clinical practice?,” 2023, British Institute of Radiology. doi: 10.1259/bjr.20221016

work page doi:10.1259/bjr.20221016 2023
[4]

3D-DXA: Assessing the Femoral Shape, the Trabecular Macrostructure and the Cortex in 3D from DXA images,

L. Humbert et al., “3D-DXA: Assessing the Femoral Shape, the Trabecular Macrostructure and the Cortex in 3D from DXA images,” IEEE Trans. Med. Imaging, vol. 36, no. 1, pp. 27– 39, Jan. 2017, doi: 10.1109/TMI.2016.2593346

work page doi:10.1109/tmi.2016.2593346 2017
[5]

Generation of 3D shape, density, cortical thickness and finite element mesh of proximal femur from a DXA image,

S. P. Väänänen, L. Grassi, G. Flivik, J. S. Jurvelin, and H. Isaksson, “Generation of 3D shape, density, cortical thickness and finite element mesh of proximal femur from a DXA image,” Med. Image Anal., vol. 24, no. 1, pp. 125–134, Aug. 2015, doi: 10.1016/j.media.2015.06.001

work page doi:10.1016/j.media.2015.06.001 2015
[6]

Prediction of femoral strength using 3D finite element models reconstructed from DXA images: validation against experiments,

L. Grassi, S. P. Väänänen, M. Ristinmaa, J. S. Jurvelin, and H. Isaksson, “Prediction of femoral strength using 3D finite element models reconstructed from DXA images: validation against experiments,” Biomech. Model. Mechanobiol., vol. 16, no. 3, pp. 989– 1000, Jun. 2017, doi: 10.1007/s10237-016-0866-2

work page doi:10.1007/s10237-016-0866-2 2017
[7]

3D Analysis of Cortical and Trabecular Bone From Hip DXA:Precision and Trend Assessment Interval in Postmenopausal Women,

L. Humbert et al., “3D Analysis of Cortical and Trabecular Bone From Hip DXA:Precision and Trend Assessment Interval in Postmenopausal Women,” Journal of Clinical Densitometry, vol. 22, no. 2, pp. 214–218, Apr. 2019, doi: 10.1016/j.jocd.2018.05.001

work page doi:10.1016/j.jocd.2018.05.001 2019
[8]

3D-DXA Cortical and Trabecular Parameters; Agreement and Precision Between GE Healthcare Prodigy and iDXA Densitometers,

D. Krueger et al., “3D-DXA Cortical and Trabecular Parameters; Agreement and Precision Between GE Healthcare Prodigy and iDXA Densitometers,” Mar. 04, 2026. doi: 10.64898/2026.03.04.26347524

work page doi:10.64898/2026.03.04.26347524 2026
[9]

Modification of bone mineral density, bone geometry and volumetric BMD in young women with obesity,

L. Maïmoun et al., “Modification of bone mineral density, bone geometry and volumetric BMD in young women with obesity,” Bone, vol. 150, p. 116005, Sep. 2021, doi: 10.1016/j.bone.2021.116005. 13

work page doi:10.1016/j.bone.2021.116005 2021
[10]

Effect of Anorexia Nervosa on Volumetric Bone Mineral Density.,

L. Maïmoun et al., “Effect of Anorexia Nervosa on Volumetric Bone Mineral Density.,” Calcif. Tissue Int., vol. 116, no. 1, p. 132, Oct. 2025, doi: 10.1007/s00223-025-01442-1

work page doi:10.1007/s00223-025-01442-1 2025
[11]

Analysis of the evolution of cortical and trabecular bone compartments in the proximal femur after spinal cord injury by 3D-DXA,

L. Gifre et al., “Analysis of the evolution of cortical and trabecular bone compartments in the proximal femur after spinal cord injury by 3D-DXA,” Osteoporosis International, vol. 29, no. 1, pp. 201–209, Jan. 2018, doi: 10.1007/s00198-017-4268-9

work page doi:10.1007/s00198-017-4268-9 2018
[12]

Alteration of Volumetric Bone Mineral Density Parameters in Men with Spinal Cord Injury,

L. Maïmoun et al., “Alteration of Volumetric Bone Mineral Density Parameters in Men with Spinal Cord Injury,” Calcif. Tissue Int., vol. 113, no. 3, pp. 304–316, Jun. 2023, doi: 10.1007/s00223-023-01110-2

work page doi:10.1007/s00223-023-01110-2 2023
[13]

Modification of bone architecture following sleeve gastrectomy: a five-year follow-up,

L. Maïmoun et al., “Modification of bone architecture following sleeve gastrectomy: a five-year follow-up,” Journal of Bone and Mineral Research, vol. 40, no. 2, pp. 251–261, Feb. 2025, doi: 10.1093/jbmr/zjae202

work page doi:10.1093/jbmr/zjae202 2025
[14]

Bone loss after bariatric surgery is observed mainly in the hip trabecular compartment and after hypoabsorptive techniques,

C. Gómez-Vaquero et al., “Bone loss after bariatric surgery is observed mainly in the hip trabecular compartment and after hypoabsorptive techniques,” Bone, vol. 190, Jan. 2025, doi: 10.1016/j.bone.2024.117270

work page doi:10.1016/j.bone.2024.117270 2025
[15]

BMD Measurement and Precision: A Comparison of GE Lunar Prodigy and iDXA Densitometers,

D. Krueger, N. Vallarta-Ast, M. Checovich, D. Gemar, and N. Binkley, “BMD Measurement and Precision: A Comparison of GE Lunar Prodigy and iDXA Densitometers,” Journal of Clinical Densitometry, vol. 15, no. 1, pp. 21–25, 2012, doi: 10.1016/j.jocd.2011.08.003

work page doi:10.1016/j.jocd.2011.08.003 2012
[16]

Importance of cross-calibration when replacing DXA Scanners: QDR4500W and Discovery Wi,

M. K. Covey, D. L. Smith, J. K. Berry, and E. D. Hacker, “Importance of cross-calibration when replacing DXA Scanners: QDR4500W and Discovery Wi,” J. Nurs. Meas., vol. 16, no. 3, pp. 155–170, 2008, doi: 10.1891/1061-3749.16.3.155

work page doi:10.1891/1061-3749.16.3.155 2008
[17]

Cross calibration between two dual energy X-ray absorptiometry systems: Horizon A and Discovery A,

S. Promma, “Cross calibration between two dual energy X-ray absorptiometry systems: Horizon A and Discovery A,” Chulalongkorn Medical Journal, vol. 66, no. 2, pp. 145–151, 14 Apr. 2022, doi: 10.58837/CHULA.CMJ.66.2.4. Figure 1. Deming regression (left) & Bland-Altman plot (right) between Hologic Discovery W and Horizon A devices at Hospital Sant Pau (Barc...

work page doi:10.58837/chula.cmj.66.2.4 2022