Pith. sign in

REVIEW 4 major objections 4 minor 22 references

Mask Mining for Improved Liver Lesion Segmentation

T0 review · 4 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 Retraining a converged U-Net on its own segmentation errors raises liver and lesion dice scores by up to 2 points across 2D, 3D, and combined cascaded architectures.

desk verdict Mask mining is a simple and genuinely different retraining trick with consistent but small dice gains; the paper's central causal claim is undercut by a missing continued-training control, but it deserves serious review with that ablation requested. read the letter →

arxiv 1908.05062 v4 pith:XLWCT334 submitted 2019-08-14 eess.IV cs.CV

Karsten Roth , Jürgen Hesser , Tomasz Konopczyński This is my paper · ORCID
classification eess.IVcs.CV
keywords liverlesionsegmentationCTU-NetmaskminingerrormasksLiTSdatasetdicescoreretraining
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 tries to establish that a simple post-training step can improve liver and lesion segmentation in CT images: take a U-Net that has already converged, generate its predictions on the training set, turn its mistakes into new four-class training masks, and retrain briefly on those masks. The central claim is that this 'mask mining' reliably raises dice scores across 2D, 3D, and combined cascaded U-Net architectures, by up to two points on the LiTS benchmark. If true, it matters because segmentation models could be improved without redesigning the network, adding hyperparameters, or rerunning the full training pipeline.

What carries the argument

The four-class error mask is the central object: each pixel is labelled true positive, false positive, false negative, or true negative by comparing the converged network's own prediction with ground truth. The paper appends four single-layer output channels to the existing U-Net output, keeping all pre-trained weights, and retrains on these error masks; because the masks force the network to separate correct from incorrect behaviour, it learns features that suppress its earlier errors. The final binary mask is recovered by grouping the error classes, and the choice of retraining loss (pixel-weighted cross-entropy versus dice) becomes a lever for favouring recall or precision.

What would settle it

Train a converged network for the same number of extra epochs with the original binary ground-truth masks and the original loss, and compare dice scores with the mask-mining retraining; if the gains match, the error masks are not the active ingredient.

Watch

Extended reading notes

Core claim

After a segmentation pipeline has converged, the paper computes segmentation masks for the training data in single forward passes and compares them with ground truth to build new target masks whose classes are true positives, false positives, false negatives, and true negatives. Four single-layer output channels are appended to the trained network, and retraining on these error masks with a dice-based loss (or, alternatively, a pixel-weighted cross-entropy loss) lets the network allocate its previous mistakes to separate channels and 'explain away' the features that caused them. The multiclass predictions are then collapsed back to binary segmentation, with the true-positive channel serving directly as the final mask under the dice loss. The paper reports consistent dice improvements over the initial models on validation and online test splits, a reduction in seed-dependent variation across three runs, and qualitative evidence that the choice of loss shifts the balance of false positives and false negatives.

Load-bearing premise

The improvement is attributed to the error-mask labels themselves, yet no control is trained for the same extra epochs on the original labels, so additional fine-tuning alone could in principle explain the gain.

Editorial extensions

If this is right

  • Add the mask-mining retraining step to any of the three tested pipelines and the volume-averaged dice for liver and lesion segmentation on LiTS improves, with the largest relative gains in the combined cascaded 2D setup.
  • Retraining on error masks shrinks seed-dependent variation: the standard deviations of dice scores over three training runs decrease in nearly all reported configurations.
  • Choosing a dice-based loss versus a pixel-weighted cross-entropy loss during mask mining shifts the false-positive/false-negative balance, letting a user steer the model toward higher recall or higher robustness after training without new hyperparameters.
  • Because the appended error channels are a small extension of a converged network and retraining converges quickly, the procedure does not require a costly re-run of the full training pipeline and is presented as applicable to other segmentation tasks.

Reading between the lines

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

  • A key untested alternative is that part or all of the gain comes simply from additional epochs on a fixed label set; a matched-duration fine-tuning control would settle this, and the paper's reported seed-variance reduction is not enough to rule it out.
  • If the mechanism generalises, mask mining should transfer to other dense prediction tasks such as organ or cell segmentation, and could be combined with hard-example mining methods by re-weighting the error classes.
  • The error-type control suggests a practical clinical workflow: choose a loss that favours recall when screening for lesions, then switch to a precision-favouring retraining for confirmation, all on the same base model.
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

4 major / 4 minor

Summary. The paper proposes a post-training "mask mining" procedure for U-Net-based liver and liver-lesion segmentation. After an initial model is trained, its predictions on the training set are compared against ground-truth masks to produce four-class error masks (true positive, true negative, false positive, false negative). Four auxiliary output channels are appended to the network, and the model is retrained with either a pixel-weighted cross-entropy loss or a multiclass dice-based loss on these mined masks. The final segmentation is then derived either by grouping the multiclass predictions or by reading the true-positive channel. The method is evaluated on the LiTS dataset across three architectures (cascaded 2D, cascaded 3D, and combined cascaded 2D), reporting consistent dice-score gains with improvements of up to 2 points claimed in the abstract. The paper also includes a qualitative analysis showing that the choice of loss shifts false-positive versus false-negative errors.

Significance. If the reported effect is real and causally due to the mined error masks, the method would be a simple, architecture-agnostic post-processing step applicable to a wide range of medical segmentation pipelines. The strengths of the paper are its use of a public benchmark, evaluation on three structurally distinct U-Net variants, detailed implementation settings, and a qualitative demonstration of error-type control. However, the central causal claim is currently not established: the retraining step changes several variables at once, there is no control for continued training with the original loss and labels, and the reported standard deviations overlap for several online-test metrics without any significance testing. The magnitude of the claimed improvement is also overstated in the abstract relative to the online test results in Table 1.

major comments (4)
  1. [§3.1, §3.3, §3.4, Table 1] The central claim that the mined error masks cause the observed improvement is not established, because the retraining procedure changes multiple variables simultaneously: it adds extra optimization epochs, appends four output channels, switches from the original loss to a multiclass dice-based loss, and replaces the binary ground-truth masks with mined error masks. Section 3.1 states that each pipeline is "trained to convergence before applying our extension to ensure that we do not just prolong the training process," but no convergence criterion or plateau check is reported, and Section 3.3 simply states that training is performed for 70 epochs. A control experiment that continues training the original architecture with the original binary labels and the original loss for the same number of additional epochs or updates is required. Without such a control, the before/after comparison in Table 1 cannot support the attribution in Section 3.4 to "the inclusions of mined trained masks."
  2. [§4, Table 1] The claim of a consistent gain is not statistically supported. For several online-test metrics the standard deviations overlap between the baseline and the mask-mined model, for example 2D liver 95.3±0.2 vs 95.5±0.3, 2D lesion 62.9±0.3 vs 63.5±0.2, and 3D lesion 55.5±0.9 vs 56.5±0.2. Only three runs are averaged, and no significance test, paired comparison, or confidence interval is provided. Since the paper reports volume-averaged dice scores, per-volume paired statistics across the test set would be straightforward to compute and should be reported for the main online-test metrics.
  3. [Abstract, Table 1] The abstract claim of "an increase in dice score of up to 2 points" is not supported by the held-out online test results. In Table 1, the largest online test gain is about 1.1 points (combined setup, lesion: 61.9±0.2 to 63.0±0.1), while the 2-point gain appears in the Training Dice column (3D liver: 92.2±1.4 to 94.2±0.3). The abstract should either cite the online test gain or be qualified to specify the training/validation subset to which the 2-point figure refers.
  4. [§2.2, Eq. (1)] The definition of the retraining targets is ambiguous and needs to be specified precisely. For the dice-based loss, the text says "the true positive error mask class is replaced with the ground truth segmentation mask," so the four output channels are not purely "segmentation errors" as defined in Section 2.1. The exact target tensor for each of the four channels under each loss is not given, and it is also not stated which output channel is used for the final binary mask at inference for the dice-based variant (Eq. (1) applies explicitly only to the pwce grouping). Without this specification the method is not fully reproducible.
minor comments (4)
  1. [§3.2] The sentence "The dataset is publicly (Creative Commons License) and was collected..." is missing a verb; it should read "publicly available under a Creative Commons license."
  2. [§3.3] The hardware is listed as "NVIDIA GeForce 1080Ti"; the standard product name is "NVIDIA GeForce GTX 1080 Ti."
  3. [§2.2] The abbreviation "pwce" is used before being defined; please expand it to "pixel-weighted cross-entropy" at first use.
  4. [§3.4] The sentence "This is arguably due to the simultaneous boost in liver and lesion segmentation performance" is a placeholder explanation rather than a finding; it should be supported by evidence or removed.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: mask-mining targets are self-generated, but the claimed dice improvements are evaluated against external ground truth and held-out test data.

full rationale

The paper's derivation chain is: train a U-Net baseline to convergence; generate error masks by comparing the model's own predictions on training data with ground truth; retrain on a multiclass dice-based loss using those masks; then evaluate on the LiTS online test set. The only self-referential element is that the error masks are derived from the model's own predictions, but this is the method itself rather than a hidden input to the claimed conclusion. No test-set labels are used to construct the masks, no parameters are fitted to the reported dice scores, and the final evaluation metric is an external ground-truth dice score. The self-citations to prior work by the same authors, [6] and [8], are used only as related-work and motivation; removing them would not change the experimental comparison. The absence of a continued-training control with the original loss is a legitimate experimental-control concern about attributing the gain to the error masks versus extra epochs, but it is not a circularity of the derivation. No equation or fitted value reduces by construction to the target result, so no circular step can be exhibited and the appropriate score is 0.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

The method introduces no newly fitted numerical constants; standard training hyperparameters (learning rate, batch size, crop size, epochs) are fixed across runs and not tuned to produce the reported gains. The categorical choice of dice over cross-entropy loss is a model selection, not a fitted value. The main unproved premise is that the self-generated error masks, rather than the additional fine-tuning, drive the improvement.

assumptions (3)
  • domain assumption CT intensity windowing and normalization to [-100, 600] HU is a valid preprocessing step.
    Used in Section 3.2 for all data; standard for CT but an unproved domain choice.
  • domain assumption Largest connected component post-processing for liver segmentation defines the evaluation protocol.
    Mentioned in Section 3.2; applied to both baseline and Inc, so it does not directly confound the relative improvement, but it is part of the assumed evaluation pipeline.
  • ad hoc to paper The error masks generated by the pretrained model provide a valid training signal for the auxiliary channels.
    This is the core mechanism of the paper. If false, the observed gains could be due to additional fine-tuning epochs rather than the error masks themselves. The paper does not test this premise with a control baseline.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Mask Mining for Improved Liver Lesion Segmentation." pith.science (2026). https://pith.science/paper/XLWCT334

@misc{pith2026190805062,
  author       = {Pith},
  title        = {Pith review of: Mask Mining for Improved Liver Lesion Segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XLWCT334}},
  note         = {Machine review of arXiv:1908.05062}
}
read the original abstract

We propose a novel procedure to improve liver and lesion segmentation from CT scans for U-Net based models. Our method extends standard segmentation pipelines to focus on higher target recall or reduction of noisy false-positive predictions, boosting overall segmentation performance. To achieve this, we include segmentation errors into a new learning process appended to the main training setup, allowing the model to find features which explain away previous errors. We evaluate this on semantically distinct architectures: cascaded two- and three-dimensional as well as combined learning setups for multitask segmentation. Liver and lesion segmentation data are provided by the Liver Tumor Segmentation challenge (LiTS), with an increase in dice score of up to 2 points.

Figures

Figures reproduced from arXiv: 1908.05062 by the authors.

Figure 1
Figure 1. fig. 1. To do so, the distribution of segmentation error types [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

22 extracted references · 14 canonical work pages

  1. [1]

    Mask Mining for Improved Liver Lesion Segmentation

    INTRODUCTION Liver imaging nowadays is mostly done via Computed To- mography (CT) [1]. Providing fully-automatic segmentation of liver and lesion tissue from CT data can hence be a useful tool to help with diagnosis and treatment planning. Common approaches utilize U-Nets [2], e.g. [3, 1, 4]. However, training of neural networks can be a difficult endeavou...

    work page Pith review arXiv 1908

  2. [2]

    2) is the gener- ation of new training masks to alter the current network per- formance and allow the network to learn from its own errors

    METHOD Fundamental for our proposed extension (fig. 2) is the gener- ation of new training masks to alter the current network per- formance and allow the network to learn from its own errors. 2.1. Basic Setup A segmentation pipeline of choice is trained until conver- gence following any training procedure. Now, segmentation masks over the training data are...

    work page

  3. [3]

    APPLICA TION TO LIVER AND LESION SEGMENTA TION 3.1. Network Architectures We investigate the performance of our method on liver and lesion segmentation by evaluating dice score performance on distinct architectures: Cascaded 2D [11], which trains a 2D segmentation network for liver and lesion segmenta- tion separately, Cascaded 3D, which does the same for...

    work page 2017

  4. [4]

    CONTROL OF SEGMENTA TION ERRORS We also qualitatively study the usage of our method to con- trol the produced segmentation error types, with examples in Fig. 4. Qualitative examination of the control capability of our pipeline extension. We evaluate false positive/false negative/true positive pixel count change for a fixed validation set on lesion segmenta...

    work page

  5. [5]

    CONCLUSION We introduced a novel extension to standard liver and lesion segmentation pipelines on the basis of the Liver Tumor Seg- mentation (LiTS) dataset. By helping the network learn and thereby explain away previously made errors using automat- ically generated training labels, we boost segmentation per- formance on different and distinct architectur...

    work page

  6. [6]

    The liver tumor segmentation benchmark (lits),

    Patrick Bilic et al., “The liver tumor segmentation benchmark (lits),” CoRR, vol. abs/1901.04056, 2019

    arXiv 1901

  7. [7]

    U-net: Convolutional net- works for biomedical image segmentation,

    Olaf Ronneberger et al., “U-net: Convolutional net- works for biomedical image segmentation,” in Inter- national Conference on Medical image computing and computer-assisted intervention, 2015, pp. 234–241

    work page 2015

  8. [8]

    nnu-net: Self-adapting frame- work for u-net-based medical image segmentation,

    Fabian Isensee et al., “nnu-net: Self-adapting frame- work for u-net-based medical image segmentation,” CoRR, vol. abs/1809.10486, 2018

    arXiv 2018

Show all 22 references
  1. [9]

    The importance of skip connec- tions in biomedical image segmentation,

    Michal Drozdzal et al., “The importance of skip connec- tions in biomedical image segmentation,” CoRR, vol. abs/1608.04117, 2016

    2016 arXiv

  2. [10]

    Tversky loss function for image segmentation using 3d fully convolu- tional deep networks,

    Seyed Sadegh Mohseni Salehi et al., “Tversky loss function for image segmentation using 3d fully convolu- tional deep networks,” in Machine Learning in Medical Imaging, 2017, pp. 379–387

    2017

  3. [11]

    Liver lesion segmentation with slice-wise 2d tiramisu and tversky loss function,

    Karsten Roth, Tomasz K. Konopczynski, and Jürgen Hesser, “Liver lesion segmentation with slice-wise 2d tiramisu and tversky loss function,” CoRR, vol. abs/1905.03639, 2019

    1905 arXiv

  4. [12]

    Conditional generative refinement adversarial networks for unbalanced medical image se- mantic segmentation,

    Mina Rezaei et al., “Conditional generative refinement adversarial networks for unbalanced medical image se- mantic segmentation,” CoRR, vol. abs/1810.03871, 2018

    2018 arXiv

  5. [13]

    Mic: Mining interclass characteristics for improved metric learning,

    Karsten Roth, Biagio Brattoli, and Bjorn Ommer, “Mic: Mining interclass characteristics for improved metric learning,” in Proceedings of the IEEE International Conference on Computer Vision, 2019, pp. 8000–8009

    2019

  6. [14]

    3d u-net: Learning dense volumet- ric segmentation from sparse annotation,

    Özgün Çiçek et al., “3d u-net: Learning dense volumet- ric segmentation from sparse annotation,” CoRR, vol. abs/1606.06650, 2016

    2016 arXiv

  7. [15]

    Liver lesion segmentation informed by joint liver segmentation,

    E. V orontsov et al., “Liver lesion segmentation informed by joint liver segmentation,” in2018 IEEE 15th Interna- tional Symposium on Biomedical Imaging (ISBI 2018) , April 2018, pp. 1332–1335

    2018

  8. [16]

    Automatic liver and lesion segmentation in CT using cascaded fully con- volutional neural networks and 3d conditional random fields,

    Patrick Ferdinand Christ et al., “Automatic liver and lesion segmentation in CT using cascaded fully con- volutional neural networks and 3d conditional random fields,” CoRR, vol. abs/1610.02177, 2016

    2016 arXiv

  9. [17]

    Automatic liver lesion segmentation using A deep convolutional neural network method,

    Xiao Han, “Automatic liver lesion segmentation using A deep convolutional neural network method,” CoRR, vol. abs/1704.07239, 2017

    2017 arXiv

  10. [18]

    Batch normaliza- tion: Accelerating deep network training by reducing internal covariate shift,

    Sergey Ioffe and Christian Szegedy, “Batch normaliza- tion: Accelerating deep network training by reducing internal covariate shift,” CoRR, vol. abs/1502.03167, 2015

    2015 arXiv

  11. [19]

    Deep residual learning for image recogni- tion,

    K. He et al., “Deep residual learning for image recogni- tion,” in2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2016, pp. 770–778

    2016

  12. [20]

    Squeeze-and-excitation networks,

    Jie Hu et al., “Squeeze-and-excitation networks,”CoRR, vol. abs/1709.01507, 2017

    2017 arXiv

  13. [21]

    Automatic differentiation in py- torch,

    Adam Paszke et al., “Automatic differentiation in py- torch,” 2017

    2017

  14. [22]

    Adam: A method for stochastic optimization,

    Diederik P. Kingma et al., “Adam: A method for stochastic optimization,” CoRR, vol. abs/1412.6980, 2014

    2014 arXiv

Pith tools

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