pith. sign in

arxiv: 1906.10288 · v2 · pith:QHC3T3GOnew · submitted 2019-06-25 · 📡 eess.IV · cs.CV

3DBGrowth: volumetric vertebrae segmentation and reconstruction in magnetic resonance imaging

Jonathan S. Ramos , Mirela T. Cazzolato , Bruno S. Fai\c{c}al , Marcello H. Nogueira-Barbosa , Caetano Traina Jr. , Agma J. M. Traina This is my paper

Pith reviewed 2026-05-25 16:45 UTC · model grok-4.3

classification 📡 eess.IV cs.CV
keywords vertebral body segmentationMRI 3D reconstructionsemi-automatic segmentationannotation reductionregion growingvolumetric medical imaging
0
0 comments X

The pith

3DBGrowth extends Balanced Growth to 3D MRI volumes and reconstructs vertebrae accurately while needing manual labels on only 37 percent of slices.

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

The paper introduces 3DBGrowth as a 3D extension of the existing Balanced Growth segmentation method, combined with a slope-coefficient rule drawn from annotation timing to choose which slices require manual input. Experiments on 17 MRI exams show the approach produces higher Dice scores than prior methods, maintains reconstruction quality, and tolerates imprecise initial seed points. If correct, this would mean radiologists or surgeons can obtain reliable 3D vertebral models with substantially less labeling effort and without sacrificing the speed or precision needed for treatment planning.

Core claim

By lifting the Balanced Growth procedure into three dimensions and using the slope of annotation-time curves to select a sparse subset of slices, 3DBGrowth achieves volumetric vertebral segmentation and surface reconstruction on MRI data. On a set of 17 exams the method requires annotation of only 37 percent of the slices that contain vertebral bodies on average, delivers more than 5 percent higher Dice overlap than competing techniques at comparable run time, and continues to perform well even when the initial seed points are imprecise.

What carries the argument

3DBGrowth, the 3D extension of the Balanced Growth region-growing procedure together with slope-coefficient slice selection.

If this is right

  • Surgeons obtain objective 3D vertebral models for planning with reduced specialist annotation time.
  • The same reduced-slice protocol can be applied to new MRI exams without retraining the underlying growth procedure.
  • Imprecise seed placement no longer forces full re-annotation or loss of final accuracy.

Where Pith is reading between the lines

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

  • The slice-selection rule may transfer to other volumetric medical segmentation tasks where annotation cost grows linearly with the number of slices.
  • If the 3D growth step preserves the original 2D efficiency, similar extensions could shorten labeling for organs whose cross-sections vary smoothly along one axis.

Load-bearing premise

The slope coefficient computed from annotation times can identify a smaller set of slices whose manual labeling is sufficient to keep both 3D segmentation accuracy and reconstruction quality intact.

What would settle it

On a held-out collection of MRI spine volumes, labeling only the slices chosen by the slope rule produces Dice scores that fall below those of the best competing method.

Figures

Figures reproduced from arXiv: 1906.10288 by Agma J. M. Traina, Bruno S. Fai\c{c}al, Caetano Traina Jr., Jonathan S. Ramos, Marcello H. Nogueira-Barbosa, Mirela T. Cazzolato.

Figure 1
Figure 1. Figure 1: Steps in a semi-automatic segmentation schema. plied for the classification of a given anomaly (9, 10, 11) or for Content-Based Image Retrieval (CBIR) (12, 13). Interactive segmentation tools can be meaning￾ful during the training and education of new radiolo￾gists (14). Students can learn how to correctly segment each vertebra and to detect spine pathologies (15). This kind of training may avoid potential… view at source ↗
Figure 2
Figure 2. Figure 2: Examples of slices annotation for a single vertebral body (Exam AKa2, L5) and 3DBGrowth's iterations. Ground￾truth, interior and exterior annotation in red, magenta and blue, respectively. Several fully automatic vertebrae segmentation meth￾ods have been proposed (19, 20). However, they take too much processing time, which may not suit clinical practice (21). More recently, a novel approach called Balanced… view at source ↗
Figure 3
Figure 3. Figure 3: Example of sloppy annotation for a few vertebral bodies in one slice (Aka2, slice 8): ground-truth, interior and exterior annotations in red, magenta and blue, respectively. 3.4 Computational set-up The experiments were performed on a 2.40GHz In￾tel(R) Core(TM) i7 CPU and 8GB RAM machine, using Matlab(R) version 2018a. The maximum number of it￾erations was set to 50 for GrowCut and 3DBGrowth. No pre or pos… view at source ↗
Figure 4
Figure 4. Figure 4: Quality comparison between 3DBGrowth and GrowCut over variations on the number of slices manually an￾notated: (a) annotation time and running time results; (b) Dice (DSC) and Jaccard (JAC). (a) Original (b) Ground-Truth (c) Annotation [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Example of seed points for a single vertebrae (St1, slice 10, L2): ground-truth (GT), interior and exterior annota￾tions in red, magenta and blue, respectively [PITH_FULL_IMAGE:figures/full_fig_p004_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Comparison of results for L2 on exam AKa2: three slices, out of 7, were annotated. To further investigate the results presented in [PITH_FULL_IMAGE:figures/full_fig_p005_6.png] view at source ↗
read the original abstract

Segmentation of medical images is critical for making several processes of analysis and classification more reliable. With the growing number of people presenting back pain and related problems, the semi-automatic segmentation and 3D reconstruction of vertebral bodies became even more important to support decision making. A 3D reconstruction allows a fast and objective analysis of each vertebrae condition, which may play a major role in surgical planning and evaluation of suitable treatments. In this paper, we propose 3DBGrowth, which develops a 3D reconstruction over the efficient Balanced Growth method for 2D images. We also take advantage of the slope coefficient from the annotation time to reduce the total number of annotated slices, reducing the time spent on manual annotation. We show experimental results on a representative dataset with 17 MRI exams demonstrating that our approach significantly outperforms the competitors and, on average, only 37% of the total slices with vertebral body content must be annotated without losing performance/accuracy. Compared to the state-of-the-art methods, we have achieved a Dice Score gain of over 5% with comparable processing time. Moreover, 3DBGrowth works well with imprecise seed points, which reduces the time spent on manual annotation by the specialist.

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 / 1 minor

Summary. The paper proposes 3DBGrowth, a 3D volumetric extension of the 2D Balanced Growth segmentation method for MRI vertebral bodies. It uses a slope coefficient derived from annotation time to select an average of only 37% of vertebral slices for manual annotation while claiming no loss in segmentation accuracy or reconstruction quality. On a dataset of 17 MRI exams, it reports >5% Dice score improvement over state-of-the-art methods at comparable runtime and robustness to imprecise seed points.

Significance. If the empirical claims hold with proper validation, the work could reduce annotation burden for 3D medical segmentation tasks relevant to surgical planning. The extension of an existing 2D method and the slice-reduction heuristic are potentially useful, but the provided abstract supplies no implementation details, error bars, statistical tests, baseline tables, or ablation studies, so the significance cannot be assessed from the given text.

major comments (2)
  1. Abstract: the central claim that the annotation-time slope coefficient reliably selects a 37% slice subset without degrading 3D growth accuracy or reconstruction quality is asserted without any supporting ablation, equivalence test, or analysis of inter-slice dependency; this is load-bearing for the efficiency contribution.
  2. Abstract: no error bars, statistical significance tests, or per-exam breakdowns are supplied for the reported >5% Dice gain, preventing verification that the 3D extension preserves the 2D method's properties or that performance is not driven by dataset-specific factors.
minor comments (1)
  1. Abstract: the phrase 'significantly outperforms' lacks any mention of the statistical test used.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address each major comment below and outline the revisions we will make to improve the clarity and support for our claims.

read point-by-point responses

  1. Referee: Abstract: the central claim that the annotation-time slope coefficient reliably selects a 37% slice subset without degrading 3D growth accuracy or reconstruction quality is asserted without any supporting ablation, equivalence test, or analysis of inter-slice dependency; this is load-bearing for the efficiency contribution.

    Authors: The manuscript body reports experimental results on 17 MRI exams showing that the slope-coefficient heuristic yields 37% annotation with no loss in Dice score or reconstruction quality relative to full annotation. We will revise the abstract to explicitly reference these supporting experiments. The 3D growth mechanism itself incorporates inter-slice volumetric consistency, which is analyzed in the methods and results sections. revision: partial

  2. Referee: Abstract: no error bars, statistical significance tests, or per-exam breakdowns are supplied for the reported >5% Dice gain, preventing verification that the 3D extension preserves the 2D method's properties or that performance is not driven by dataset-specific factors.

    Authors: We agree that error bars, statistical tests, and per-exam detail would strengthen verifiability. The >5% Dice improvement is obtained from direct comparison against state-of-the-art baselines on the 17-exam dataset; the full paper already contains the aggregate tables. In revision we will add error bars to the reported metrics, include statistical significance tests, and supply per-exam breakdowns (or move them to supplementary material if space-constrained). revision: yes

Circularity Check

0 steps flagged

No circularity; empirical claims rest on experimental comparison, not self-referential fitting or derivation

full rationale

The paper extends a prior 2D Balanced Growth method to 3D and uses a slope coefficient derived from annotation times to select ~37% of slices, claiming equivalent Dice/reconstruction quality on 17 MRI exams plus >5% gain over SOTA. These are presented as experimental outcomes with direct comparisons, not as predictions forced by fitting parameters to the target metric or by self-citation chains. No equations, uniqueness theorems, or ansatzes appear that reduce the central claims to inputs by construction. The slope-based slice selection is a heuristic whose accuracy preservation is asserted via measured results rather than definitional equivalence.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review supplies no explicit free parameters, axioms, or invented entities; the method is described only at the level of extending an existing 2D technique and using an annotation-time slope heuristic.

pith-pipeline@v0.9.0 · 5781 in / 1284 out tokens · 47877 ms · 2026-05-25T16:45:49.190985+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

28 extracted references · 28 canonical work pages

  1. [1]

    FEHLINGS et al

    Fehlings et al. FEHLINGS et al. 2015 FEHLINGS, M. G. et al. The aging of the global populationthe changing epidemiology of disease and spinal disorders. Neurosurgery, v. 77, n. 1, p. 1--5, 2015

    work page 2015

  2. [2]

    o nnies RAK; T \

    Rak e T \"o nnies RAK; T \"O NNIES 2016 RAK, M.; T \"O NNIES, K. D. On computerized methods for spine analysis in MRI : a systematic review. Int. J. Comput. Assist. Radiol. Surg., v. 11, n. 8, p. 1445--1465, Aug 2016. ISSN 1861-6429

    work page 2016

  3. [3]

    WANG et al

    Wang et al. WANG et al. 2017 WANG, Y. X. J. et al. I dentifying osteoporotic vertebral endplate and cortex fractures . Quant Imaging Med Surg, v. 7, n. 5, p. 555--591, Oct 2017

    work page 2017

  4. [4]

    HAMMERNIK et al

    Hammernik et al. HAMMERNIK et al. 2015 HAMMERNIK, K. et al. Vertebrae segmentation in 3D CT images based on a variational framework. In: YAO, J. et al. (Ed.). Recent Advances in Computational Methods and Clinical Applications for Spine Imaging. Cham: Springer International Publishing, 2015. p. 227--233. ISBN 978-3-319-14148-0

    work page 2015

  5. [5]

    DZENAN et al

    Dzenan et al. DZENAN et al. 2014 DZENAN, Z. et al. Robust detection and segmentation for diagnosis of vertebral diseases using routine MR images. Computer Graphics Forum, v. 33, n. 6, p. 190--204, 2014

    work page 2014

  6. [6]

    J.; KINAHAN, P

    Gillies, Kinahan e Hricak GILLIES; KINAHAN; HRICAK 2016 GILLIES, R. J.; KINAHAN, P. E.; HRICAK, H. R adiomics: images are more than pictures, they are data . Radiology, v. 278, n. 2, p. 563--577, Feb 2016

    work page 2016

  7. [7]

    V ertebral body segmentation with G row C ut: I nitial experience, workflow and practical application

    Egger, Nimsky e Chen EGGER; NIMSKY; CHEN 2017 EGGER, J.; NIMSKY, C.; CHEN, X. V ertebral body segmentation with G row C ut: I nitial experience, workflow and practical application . SAGE Open Med, v. 5, p. 1--5, 2017

    work page 2017

  8. [8]

    JUNIOR et al

    Junior et al. JUNIOR et al. 2018 JUNIOR, J. R. F. et al. Radiomics-based features for pattern recognition of lung cancer histopathology and metastases. Computer Methods and Programs in Biomedicine, v. 159, p. 23 -- 30, 2018. ISSN 0169-2607

    work page 2018

  9. [9]

    CASTI et al

    Casti et al. CASTI et al. 2017 CASTI, P. et al. C ooperative strategy for a dynamic ensemble of classification models in clinical applications: the case of M R I vertebral compression fractures. International Journal of Computer Assisted Radiology and Surgery, v. 12, n. 11, p. 1971--1983, Nov 2017

    work page 2017

  10. [10]

    FRIGHETTO-PEREIRA et al

    Frighetto-Pereira et al. FRIGHETTO-PEREIRA et al. 2016 FRIGHETTO-PEREIRA, L. et al. Shape, texture and statistical features for classification of benign and malignant vertebral compression fractures in magnetic resonance images. Computers in Biology and Medicine, v. 73, p. 147 -- 156, 2016. ISSN 0010-4825

    work page 2016

  11. [11]

    CAZZOLATO et al

    Cazzolato et al. CAZZOLATO et al. 2019 CAZZOLATO, M. T. et al. dp-breath: Heat maps and probabilistic classification assisting the analysis of abnormal lung regions. Computer Methods and Programs in Biomedicine, v. 173, p. 27--34, 2019. ISSN 0169-2607

    work page 2019

  12. [12]

    XUE et al

    Xue et al. XUE et al. 2011 XUE, Z. et al. Spine X -ray image retrieval using partial vertebral boundaries. In: 2011 24th International Symposium on Computer-Based Medical Systems (CBMS). [S.l.: s.n.], 2011. p. 1--6. ISSN 1063-7125

    work page 2011

  13. [13]

    GURURAJAN et al

    Gururajan et al. GURURAJAN et al. 2011 GURURAJAN, A. et al. On the creation of a segmentation library for digitized cervical and lumbar spine radiographs. Computerized Medical Imaging and Graphics, v. 35, n. 4, p. 251 -- 265, 2011. ISSN 0895-6111

    work page 2011

  14. [14]

    KARIMI et al

    Karimi et al. KARIMI et al. 2018 KARIMI, D. et al. Prostate segmentation in MRI using a convolutional neural network architecture and training strategy based on statistical shape models. Int. J. Computer Assisted Radiology and Surgery, v. 13, n. 8, p. 1211--1219, 2018

    work page 2018

  15. [15]

    STEFAN et al

    Stefan et al. STEFAN et al. 2018 STEFAN, P. et al. A radiation-free mixed-reality training environment and assessment concept for C -arm-based surgery. International Journal of Computer Assisted Radiology and Surgery, v. 13, n. 9, p. 1335--1344, Sep 2018. ISSN 1861-6429

    work page 2018

  16. [16]

    BANERJEE et al

    Banerjee et al. BANERJEE et al. 2017 BANERJEE, P. et al. A semi-automated approach to improve the efficiency of medical imaging segmentation for haptic rendering. Journal of Digital Imaging, v. 30, n. 4, p. 519--527, Aug 2017. ISSN 1618-727X

    work page 2017

  17. [17]

    GrowCut - interactive multi-label N-D image segmentation by cellular automata

    Vezhnevets e Konouchine VEZHNEVETS; KONOUCHINE 2005 VEZHNEVETS, V.; KONOUCHINE, V. GrowCut - interactive multi-label N-D image segmentation by cellular automata. International Conference on Computer Graphics and Vision - GraphiCon , v. 1, Nov 2005

    work page 2005

  18. [18]

    ZHU et al

    Zhu et al. ZHU et al. 2014 ZHU, L. et al. An effective interactive medical image segmentation method using Fast GrowCut . In: . [S.l.: s.n.], 2014. v. 17

    work page 2014

  19. [19]

    KOREZ et al

    Korez et al. KOREZ et al. 2016 KOREZ, R. et al. Model-based segmentation of vertebral bodies from MR images with 3D CNNs . In: OURSELIN, S. et al. (Ed.). Medical Image Computing and Computer-Assisted Intervention -- MICCAI 2016. Cham: Springer Int. Publishing, 2016. p. 433--441. ISBN 978-3-319-46723-8

    work page 2016

  20. [20]

    GAONKAR et al

    Gaonkar et al. GAONKAR et al. 2017 GAONKAR, B. et al. Multi-parameter ensemble learning for automated vertebral body segmentation in heterogeneously acquired clinical MR images. J. of Translational Engineering in Health and Medicine, v. 5, p. 1--12, 2017. ISSN 2168-2372

    work page 2017

  21. [21]

    HILLE et al

    Hille et al. HILLE et al. 2018 HILLE, G. et al. Vertebral body segmentation in wide range clinical routine spine MRI data. Computer Methods and Programs in Biomedicine, v. 155, p. 93 -- 99, 2018. ISSN 0169-2607

    work page 2018

  22. [22]

    RAMOS et al

    Ramos et al. RAMOS et al. 2019 RAMOS, J. S. et al. BG rowth: an efficient approach for the segmentation of vertebral compression fractures in magnetic resonance imaging. Symposium on Applied Computing, p. 1--8, April 2019

    work page 2019

  23. [23]

    VANNESCHI et al

    Vanneschi et al. VANNESCHI et al. 2004 VANNESCHI, L. et al. Fitness clouds and problem hardness in genetic programming. In: SPRINGER. Genetic and Evolutionary Computation Conference. [S.l.], 2004. p. 690--701

    work page 2004

  24. [24]

    The distribution of the flora in the alpine zone

    Jaccard JACCARD 1912 JACCARD, P. The distribution of the flora in the alpine zone. New Phytologist, v. 11, n. 2, p. 37--50, fev. 1912

    work page 1912

  25. [25]

    A Method of Establishing Groups of Equal Amplitude in Plant Sociology Based on Similarity of Species Content and Its Application to Analyses of the Vegetation on Danish Commons

    S rensen S RENSEN 1948 S RENSEN, T. A Method of Establishing Groups of Equal Amplitude in Plant Sociology Based on Similarity of Species Content and Its Application to Analyses of the Vegetation on Danish Commons. [S.l.]: I kommission hos E. Munksgaard, 1948. (Biologiske skrifter)

    work page 1948

  26. [26]

    BARBIERI et al

    Barbieri et al. BARBIERI et al. 2015 BARBIERI, P. D. et al. Vertebral body segmentation of spine MR images using Superpixels . In: JUNIOR, C. T. et al. (Ed.). 28th IEEE International Symposium on Computer-Based Medical Systems. S\ ao Carlos and Ribeir\ ao Preto, Brazil: Conference Publishing Services (CPS), 2015. p. 44--49. ISSN 1063-7125

    work page 2015

  27. [27]

    Massey MASSEY 1951 MASSEY, F. J. The K olmogorov- S mirnov test for goodness of fit. Journal of the American Statistical Association, American Statistical Association, v. 46, n. 253, p. 68--78, 1951

    work page 1951

  28. [28]

    Critical values and probability levels for the Wilcoxon rank sum test and the Wilcoxon signed rank test

    Wilcoxon, Katti e Wilcox WILCOXON; KATTI; WILCOX 1970 WILCOXON, F.; KATTI, S.; WILCOX, R. Critical values and probability levels for the Wilcoxon rank sum test and the Wilcoxon signed rank test. Selected Tables in Mathematical Statistics, v. 1, p. 171--259, 1970

    work page 1970