Pith. sign in

REVIEW 3 major objections 4 minor 28 references

Pixel-wise Segmentation of Right Ventricle of Heart

T0 review · 3 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a plain U-Net trained with an adaptive foreground/background switching loss and majority-voting ensembling reaches state-of-the-art right ventricle segmentation in cardiac MRI without any post-processing.

desk verdict The abstract's SOTA claim is too broad — the 0.86/6.73 numbers are Test Set 2 only and the Dice is not even best there — but the switching loss and comparative study are real, reproducible-in-spirit work worth a serious referee. read the letter →

arxiv 1908.08004 v1 pith:RMSH42XL submitted 2019-08-21 eess.IV cs.CV

Yaman Dang , Deepak Anand , Amit Sethi This is my paper · ORCID
classification eess.IVcs.CV
keywords rightventriclesegmentationcardiacMRIU-NetswitchinglossclassimbalanceensemblevotingsemanticHausdorffdistance
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

The paper sets out to show that right ventricle segmentation in cardiac MRI does not need elaborate architectures or post-processing. Its proposal is a plain U-Net trained with a 'switching' loss—an adaptive blend of foreground and background Dice terms—and combined with majority-voting ensembling. The authors report state-of-the-art scores of 0.86 Dice coefficient and 6.73 mm Hausdorff distance on the RVSC-MICCAI 2012 dataset. If the claim holds, a simple, cheap pipeline could replace 15–20 minutes of manual contouring per patient and make right ventricle functional assessment far more practical in clinical settings.

What carries the argument

The load-bearing object is the switching loss, an adaptive cost function combining binary cross-entropy with Dice loss $L_D$ on the foreground and inverted Dice loss $L_I$ on the background. When the foreground pixel fraction $C_f/C_t$ is above a threshold $\tau$, the loss weights the foreground Dice term by $\lambda=0.75$ and the background term by $1-\lambda$; below the threshold, the weights are swapped. This lets the loss emphasize background Dice on apical slices where the ventricle is tiny, and foreground Dice elsewhere. The full pipeline is a depth-four U-Net with 3×3 convolutions and bilinear upsampling, trained with a cyclic learning-rate schedule, and combined across models by majority voting. The paper's claim is that this combination makes post-processing such as fully-connected CRF unnecessary.

What would settle it

Run the same training protocol on the same 12-patient training split while sweeping the switching threshold across its full range and changing the ensemble size; if the endocardium Dice on the 4-patient validation set varies by more than a trivial margin, the reported 0.86 Dice / 6.73 mm figure is tied to choices the paper does not specify.

Watch

Extended reading notes

Core claim

For right ventricle segmentation in short-axis cardiac MRI, the authors find that a depth-four U-Net trained with a switching loss and then combined by majority voting performs at or above the level of more elaborate networks such as dilated U-Net, global convolution networks, and GCN-UNet hybrids, and does not need CRF post-processing. On the RVSC-MICCAI 2012 dataset they report a Dice coefficient of 0.86 and a Hausdorff distance of 6.73 mm for the endocardial contour on test set 2, along with the best validation-set Dice among the compared methods. The switching loss, which adaptively weights the foreground Dice versus the inverted background Dice according to the fraction of foreground pixels, is the component that the paper credits for handling the extreme class imbalance of the small, crescent-shaped right ventricle.

Load-bearing premise

The load-bearing premise is that the exact choice of the switching threshold and of the number of models in the majority-vote ensemble does not change the trained model enough to affect the reported score.

Editorial extensions

If this is right

  • A modest U-Net, rather than a larger receptive-field model, is the better substrate for right ventricle segmentation once the loss is matched to the class imbalance.
  • The switching loss outperforms focal loss and fixed combinations of BCE, Dice, and inverted Dice on this dataset, giving a recipe for other small-foreground segmentation tasks.
  • Majority voting and average-probability ensembling give similar gains, and CRF post-processing adds nothing, so the pipeline can stay simple.
  • On RVSC test set 2, the method reports the best endocardium Dice (0.8652) and Hausdorff distance (6.73 mm) among the compared state-of-the-art methods, while on test set 1 it does not reach the best numbers.

Reading between the lines

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

  • One consequence the authors leave implicit is that their recipe is not fully reproducible as written: the switching threshold $\tau$ and the number of models in the majority-vote ensemble are never reported, so a follow-up should publish those values or tie $\tau$ to a validation-set sweep.
  • The comparative failure of focal loss on this tiny-foreground task suggests a testable hypothesis for other medical segmentation problems: adaptive foreground/background Dice weighting may beat focal loss when the foreground is small but spatially coherent, such as small lesions or vessels.
  • Since average-probability ensembling matches majority voting, a single model distilled to mimic the ensemble could retain most of the accuracy at lower inference cost; this is an experiment the paper does not run.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper proposes a UNet-based method for right ventricle segmentation in cardiac MRI, with a new 'switching loss' that adaptively weights foreground and background Dice terms, and evaluates it on the RVSC-MICCAI 2012 challenge data. The authors compare several loss functions, network architectures, and ensembling strategies, and report that their majority-voting ensemble without post-processing achieves Dice 0.86 and Hausdorff distance 6.73 mm on the challenge test data. The paper also presents the switching loss as superior to fixed-weight loss combinations and claims state-of-the-art performance.

Significance. If the claims were fully supported, the paper would be a useful practical contribution: it gives a systematic comparison of losses, architectures, and ensembling for a clinically relevant but challenging segmentation task, and the switching-loss idea is a reasonable approach to class imbalance. The extensive ablation is a strength, as is the explicit comparison of CRF post-processing versus ensembling. However, the headline 'state-of-the-art on RVSC-MICCAI 2012' claim is not supported by the paper's own tables, and key methodological details (the switching-loss threshold and the ensemble size) are missing, limiting reproducibility. The comparative study itself remains useful, but the central claim needs to be corrected and the missing details supplied.

major comments (3)
  1. [Abstract and Section IV-C-2, Tables IV and V] The abstract states that the method 'achieves the state-of-the-art performance of 0.86 Dice coefficient and 6.73 mm Hausdorff distance on RVSC-MICCAI 2012 dataset.' The paper's own Section IV-C-2 concedes that on Test Set 1 the results are not the best, and Table IV confirms this: the majority-voting ensemble scores 0.8170 Dice and 10.56 mm HD on Test Set 1, while Avendi et al. report 0.86 Dice and 7.80 mm HD. On Test Set 2, the reported 0.8652 Dice and 6.73 mm HD are better in Hausdorff distance than Guo et al.'s 6.93 mm, but Guo et al. achieve a higher Dice (0.87). Thus the abstract's unqualified claim of state-of-the-art performance on the RVSC dataset is not true for Dice on either official test set and is true for Hausdorff distance only on Test Set 2. The abstract and conclusion should be revised to state precisely which metric and which test set are meant, or to report pooled metrics across both test sets.
  2. [Section III-D-4, Eq. (Switching Loss)] The switching loss depends on two hyperparameters, λ and τ, and the equation is written with Cn/Ct while the text defines Cf/Ct as the foreground ratio. The paper reports λ = 0.75 but never reports the value of τ, nor does it state how the threshold was selected or how sensitive the results are to it. Because the 'switching' between foreground-weighted and background-weighted Dice terms is a central component of the proposed method, the missing τ value is a load-bearing reproducibility gap. Please report the value of τ, correct the Cn/Ct notation, and add an ablation or sensitivity statement for τ.
  3. [Section IV-C-2 and Tables IV and V] The majority-voting ensemble is presented as the best method, but neither the number of models in the ensemble nor the composition of the ensemble is specified anywhere in the manuscript. Without this information, the reported test-set scores cannot be reproduced or compared exactly with other methods. Please state explicitly how many models were combined and which input variations (e.g., augmentation runs, architectures, loss weights) were used for the ensemble.
minor comments (4)
  1. [Tables IV and V] Table IV and Table V both have captions beginning 'Comparison of Dice scores for Endocardium'; Table V is clearly for the Epicardium and should be labeled accordingly.
  2. [Section III-D-4] The switching-loss equation uses the symbol Cn, which is not defined; the text says the ratio is of foreground pixels Cf to total pixels Ct. Please align the notation.
  3. [Section IV-A, Table III] Table III compares validation results with published methods that may have used different validation splits or protocols; the paper should state clearly that these numbers are not directly comparable unless the same training/validation split is used.
  4. [General] There are minor grammatical issues (e.g., 'a train to test split ratio of 3:1' when the actual split is 12:4 patients) and inconsistent use of 'test' versus 'validation' in Section IV; please proofread for consistency.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the method's claims are empirical evaluations on held-out RVSC test sets, not derivations from fitted inputs.

full rationale

The paper's central claim is an empirical performance comparison of a U-Net trained with a proposed switching loss and majority-voting ensembling against published results on the RVSC-MICCAI 2012 dataset. The loss hyperparameter lambda = 0.75 is selected using a 12:4 patient train/validation split, and the reported test-set numbers are evaluated on the official Test Set 1 and Test Set 2; this is standard model selection and independent testing, not a fitted input being renamed as a prediction. The switching loss equation is a weighted combination of BCE, Dice, and inverted Dice losses with a threshold tau; even though tau is not reported, choosing a weight empirically does not make the test-set Dice or Hausdorff distance equal to the loss by construction. No claim in the paper reduces to its own definition, and there are no load-bearing self-citations or imported uniqueness theorems. The only notable weakness is that the abstract's headline numbers (0.86 Dice, 6.73 mm HD) come from Test Set 2 only and, on Test Set 1, the method is not state-of-the-art; however, this is an accuracy and reporting concern, not a circularity concern. Under the stated hard rules, no circular step can be quoted and exhibited, so the appropriate score is 0.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central claim depends on the empirical soundness of the RVSC benchmark and on the chosen loss hyperparameters. The method introduces no new physical entities or mathematical axioms.

free parameters (2)
  • Switching loss weight λ = 0.75
    Tuned on the validation split; the paper states 'we found that λ = 0.75 works best for us' in Section III-D-4.
  • Switching loss threshold τ = not reported
    Defined in the switching loss equation in Section III-D-4 but its value is never given, so the exact model that produced the reported results is underspecified.
assumptions (4)
  • standard math Dice coefficient and Hausdorff distance as defined in Section III-G are valid and computed correctly.
    The evaluation metrics are standard, though the Dice formula omits the explicit intersection operator detail.
  • domain assumption The RVSC ground-truth contours (endocardium and epicardium) are accurate expert annotations.
    The paper treats the challenge labels as ground truth for training and evaluation, as described in Section III-A.
  • ad hoc to paper The switching loss improves class-imbalance handling relative to fixed-weight combinations.
    This is the paper's central methodological claim, supported only by its own validation experiments; no independent analysis or theory is provided.
  • domain assumption Reported prior-art numbers from other papers are directly comparable to this paper's validation and test results.
    The comparison tables take scores from previous publications without discussing differences in validation splits, preprocessing, or implementation details, which could bias the claimed improvements.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Pixel-wise Segmentation of Right Ventricle of Heart." pith.science (2026). https://pith.science/paper/RMSH42XL

@misc{pith2026190808004,
  author       = {Pith},
  title        = {Pith review of: Pixel-wise Segmentation of Right Ventricle of Heart},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RMSH42XL}},
  note         = {Machine review of arXiv:1908.08004}
}
read the original abstract

One of the first steps in the diagnosis of most cardiac diseases, such as pulmonary hypertension, coronary heart disease is the segmentation of ventricles from cardiac magnetic resonance (MRI) images. Manual segmentation of the right ventricle requires diligence and time, while its automated segmentation is challenging due to shape variations and illdefined borders. We propose a deep learning based method for the accurate segmentation of right ventricle, which does not require post-processing and yet it achieves the state-of-the-art performance of 0.86 Dice coefficient and 6.73 mm Hausdorff distance on RVSC-MICCAI 2012 dataset. We use a novel adaptive cost function to counter extreme class-imbalance in the dataset. We present a comprehensive comparative study of loss functions, architectures, and ensembling techniques to build a principled approach for biomedical segmentation tasks.

Figures

Figures reproduced from arXiv: 1908.08004 by the authors.

Figure 1
Figure 1. Typical image from dataset with endocardial and [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. UNet [10] architecture used as baseline model. [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Structure of GCN and the BR block The overall structure of the model uses a ResNet as feature extraction network and the fully connected layers as the segmentation networks. We have done various experiments with the size of kernels. 4) GCN with UNet feature map: This model has a structure very similar to the GCN model. The main difference is that the ResNet based feature extraction block is replaced by the baseline … view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

28 extracted references · 24 canonical work pages

  1. [1]

    Right ventricle segmentation from cardiac MRI: a collation study,

    C. Petitjean, M. A. Zuluaga, W. Bai, J.-N. Dacher, D. Grosgeorge, J. Caudron, S. Ruan, I. B. Ayed, M. J. Cardoso, H.-C. Chen et al., “Right ventricle segmentation from cardiac MRI: a collation study,” Medical image analysis, vol. 19, no. 1, pp. 187–202, 2015

    work page 2015

  2. [2]

    Local motion intensity clustering (lmic) model for segmentation of right ventricle in cardiac MRI images,

    Z. Guo, W. Tan, L. Wang, L. Xu, X. Wang, B. Yang, and Y . Yao, “Local motion intensity clustering (lmic) model for segmentation of right ventricle in cardiac MRI images,”IEEE journal of biomedical and health informatics, vol. 23, no. 2, pp. 723–730, 2018

    work page 2018

  3. [3]

    Right ventric- ular segmentation in cardiac MRI with moving mesh correspondences,

    K. Punithakumar, M. Noga, I. B. Ayed, and P. Boulanger, “Right ventric- ular segmentation in cardiac MRI with moving mesh correspondences,” Computerized Medical Imaging and Graphics, vol. 43, pp. 15–25, 2015

    work page 2015

  4. [4]

    Right ventricle segmentation using a 3d cylindrical shape model,

    O. Moolan-Feroze, M. Mirmehdi, and M. Hamilton, “Right ventricle segmentation using a 3d cylindrical shape model,” in 2016 IEEE 13th International Symposium on Biomedical Imaging (ISBI) . IEEE, 2016, pp. 44–48

    work page 2016

  5. [5]

    Fast, accurate, and fully automatic segmentation of the right ventricle in short-axis cardiac MRI,

    J. Ringenberg, M. Deo, V . Devabhaktuni, O. Berenfeld, P. Boyers, and J. Gold, “Fast, accurate, and fully automatic segmentation of the right ventricle in short-axis cardiac MRI,” Computerized Medical Imaging and Graphics, vol. 38, no. 3, pp. 190–201, 2014

    work page 2014

  6. [6]

    Right ventricle segmentation in cardiac MR images using u-net with partly dilated convolution,

    G. Borodin and O. Senyukova, “Right ventricle segmentation in cardiac MR images using u-net with partly dilated convolution,” in International Conference on Artificial Neural Networks. Springer, 2018, pp. 179–185

    work page 2018

  7. [7]

    ν-net: deep learning for generalized biventricular mass and function parameters using multicenter cardiac MRI data,

    H. B. Winther, C. Hundt, B. Schmidt, C. Czerner, J. Bauersachs, F. Wacker, and J. V ogel-Claussen, “ν-net: deep learning for generalized biventricular mass and function parameters using multicenter cardiac MRI data,” JACC: Cardiovascular Imaging, p. 2479, 2018

    work page 2018

  8. [8]

    A deep learning network for right ventricle segmentation in short-axis MRI,

    G. Luo, R. An, K. Wang, S. Dong, and H. Zhang, “A deep learning network for right ventricle segmentation in short-axis MRI,” in 2016 Computing in Cardiology Conference (CinC) . IEEE, 2016, pp. 485– 488

    work page 2016

Show all 28 references
  1. [9]

    A fully convolutional neural network for cardiac segmenta- tion in short-axis MRI,

    P. V . Tran, “A fully convolutional neural network for cardiac segmenta- tion in short-axis MRI,” arXiv preprint arXiv:1604.00494 , 2016

    2016 arXiv

  2. [10]

    U-net: Convolutional networks for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” in International Conference on Medical image computing and computer-assisted intervention. Springer, 2015, pp. 234–241

    2015

  3. [11]

    Large kernel matters– improve semantic segmentation by global convolutional network,

    C. Peng, X. Zhang, G. Yu, G. Luo, and J. Sun, “Large kernel matters– improve semantic segmentation by global convolutional network,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 4353–4361

    2017

  4. [12]

    Cardiac MRI segmentation,

    C.-H. Yee, “Cardiac MRI segmentation,” https://chuckyee.github.io/ cardiac-segmentation/

  5. [13]

    A review of segmentation methods in short axis cardiac MR images,

    C. Petitjean and J.-N. Dacher, “A review of segmentation methods in short axis cardiac MR images,” Medical image analysis , vol. 15, no. 2, pp. 169–184, 2011. TABLE IV: Comparison of Dice scores for Endocardium with state-of-the-art methods for TestSet1 and TestSet2. T represe...

    2011

  6. [14]

    T 0.85 (0.11) 8.63 (6.20) 0.88 (0.08) 8.03 (5.57) 0.81 (0.14) 9.46 (6.93) 0.87 (0.08) 6.93 (4.33) 0.88 (0.08) 6.94 (5.05) 0.85 (0.08) 6.92 (2.94) K Punithakumar, et al [3] T 0.83 (0.13) 7.72 (3.97) 0.77 (0.16) 9.64 (4.15) 0.85 (0.15) 6.49 (4.44) J Ringenberg, et al

  7. [15]

    T 0.83 (0.16) 9.05 (6.98) 0.88 (0.11) 7.69 (6.03) 0.77 (0.18) 10.71 (7.69) 0.83 (0.18) 8.73 (7.62) Chuck-Hou Yee

  8. [16]

    Avendi, et al

    D 0.84 (0.21) 0.84 (0.21) M R. Avendi, et al

  9. [17]

    T represents traditional methods, D represents deep learning based methods, ED stands for End diastolic and ES stands for End systolic cycle

    D 0.86 (0.11) 7.80 (4.26) 0.79 (0.16) 7.51 (3.66) 0.86 (0.10) 7.85 (4.56) 0.76 (0.20) 8.27 (4.23) Booz, et al [9] D 0.84 (0.21) 8.86 (11.27) 0.84 (0.21) 8.86 (11.27) Majority voting (without CRF) ours D 0.8170 (0.22) 10.56 (11.53) 0.8578 (0.18) 10.12 (10.78) 0.7671 (0.25) 11.0...

  10. [18]

    T K Punithakumar, et al [3] T 0.87 (0.08) 8.08 (3.80) 0.82 (0.10) 9.99 (3.85) 0.88 (0.10) 6.95 (3.98) J Ringenberg, et al

  11. [19]

    T 0.86 (0.11) 9.60 (7.01) 0.90 (0.08) 8.02 (5.96) 0.82 (0.13) 11.52 (7.70) 0.86 (0.14) 9.00 (7.46) Chuck-Hou Yee

  12. [20]

    Avendi, et al

    D 0.88 (0.18) 0.88 (0.18) MR. Avendi, et al

  13. [21]

    D Booz, et al [9] D 0.86 (0.20) 9.33 (10.79) 0.86 (0.20) 9.33 (10.79) Majority V oting (without CRF) ours D 0.8389 (0.20) 11.86 (13.27) 0.8648 (0.18) 11.57 (13.35) 0.8073 (0.23) 12.22 (13.21) 0.8936 (0.13) 7.11 (6.04) 0.9164 (0.10) 6.50 (5.72) 0.8652 (0.15) 7.87 (6.37) Majorit...

  14. [22]

    A hybrid graph-based approach for right ventricle segmentation in cardiac MRI by long axis information transition,

    M. G. Oghli, A. Mohammadzadeh, R. Kafieh, and S. Kermani, “A hybrid graph-based approach for right ventricle segmentation in cardiac MRI by long axis information transition,” Physica Medica, vol. 54, pp. 103–116, 2018

    2018

  15. [23]

    Automatic segmen- tation of the right ventricle from cardiac MRI using a learning-based approach,

    M. R. Avendi, A. Kheradvar, and H. Jafarkhani, “Automatic segmen- tation of the right ventricle from cardiac MRI using a learning-based approach,” Magnetic resonance in medicine , vol. 78, no. 6, pp. 2439– 2448, 2017

    2017

  16. [24]

    Efficient inference in fully connected crfs with gaussian edge potentials,

    P. Kr ¨ahenb¨uhl and V . Koltun, “Efficient inference in fully connected crfs with gaussian edge potentials,” in Advances in neural information processing systems, 2011, pp. 109–117

    2011

  17. [25]

    Semi- supervised automatic segmentation of layer and fluid region in retinal optical coherence tomography images using adversarial learning,

    X. Liu, J. Cao, T. Fu, Z. Pan, W. Hu, K. Zhang, and J. Liu, “Semi- supervised automatic segmentation of layer and fluid region in retinal optical coherence tomography images using adversarial learning,” IEEE Access, vol. 7, pp. 3046–3061, 2018

    2018

  18. [26]

    Focal loss for dense object detection,

    T.-Y . Lin, P. Goyal, R. Girshick, K. He, and P. Doll ´ar, “Focal loss for dense object detection,” in Proceedings of the IEEE international conference on computer vision , 2017, pp. 2980–2988

    2017

  19. [27]

    Automatic differentiation in PyTorch,

    A. Paszke, S. Gross, S. Chintala, G. Chanan, E. Yang, Z. DeVito, Z. Lin, A. Desmaison, L. Antiga, and A. Lerer, “Automatic differentiation in PyTorch,” in NIPS Autodiff Workshop, 2017

    2017

  20. [28]

    SGDR: Stochastic gradient descent with warm restarts,

    I. Loshchilov and F. Hutter, “SGDR: Stochastic gradient descent with warm restarts,” arXiv preprint arXiv:1608.03983 , 2016

    2016 arXiv

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.