Pith. sign in

REVIEW 5 major objections 6 minor 16 references

Deep Distance Map Regression Network with Shape-aware Loss for Imbalanced Medical Image Segmentation

T0 review · 5 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read Distance-map regression branch lifts small-tumor segmentation Dice to 0.751.

desk verdict Incremental but plausible distance-map regression scheme for small tumor segmentation; the mechanism is under-tested, but the empirical gains justify a careful review. read the letter →

arxiv 2501.09116 v1 pith:WQ6FKSA6 submitted 2025-01-15 eess.IV cs.CV

Huiyu Li , Xiabi Liu , Said Boumaraf , Xiaopeng Gong , Donghai Liao , Xiaohong Ma This is my paper · ORCID
classification eess.IVcs.CV
keywords distancemapregressionmedicalimagesegmentationlivertumorclassimbalanceshape-awarelossDicelightweightnetworkCTimaging
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 distance maps—not just binary masks—should supervise segmentation of small anatomical objects such as liver tumors. It proposes a two-part framework: a lightweight regression network (LR-Net) that is pretrained on ground-truth masks to map a mask to a normalized inverse distance map and then frozen, plus a MapDice loss that uses the ground-truth distance map as a spatial penalty map. The authors argue that this injects geometric shape information that ordinary class-imbalance losses lack. On the LiTS validation set and a clinical tumor dataset, the combination raises per-case Dice from 0.647 with plain Dice to 0.751 with the full method, outperforming WCE, GDS, Tversky, Focal Tversky, and ExpLog baselines.

What carries the argument

The norm-inverse distance map (NI-DM) is the central representation: for each voxel it records the Euclidean distance to the nearest object boundary, then inverts and normalizes each connected component so boundary-adjacent voxels receive the highest values. The LR-Net is a lightweight U-shaped regression network, pretrained separately on ground-truth binary masks with a smooth L1 loss and then frozen, so that it can convert the main network's soft probability map into a predicted distance map in a differentiable way. The MapDice loss, $L_{\mathrm{MapDice}} = 1 - \sum_c \frac{2(p_c \varphi_c) + \varepsilon}{p_c + \varphi_c + \varepsilon}$, multiplies the predicted probability for each class by the ground-truth distance map $\varphi_c$, turning boundary proximity into a pixel-wise penalty. Together these pieces make shape and boundary information explicit in the training signal without changing the architecture of the existing segmentation network.

What would settle it

Train the same pipeline but replace the learned LR-Net with a direct, differentiable computation of the norm-inverse distance map from the soft probability map, or feed the LR-Net with soft probability maps sampled early in training; if Dice on the clinical dataset stays at 0.751 or the regression error on soft maps is uncorrelated with segmentation quality, the learned frozen mapping is not the actual source of the improvement.

Watch

Extended reading notes

Core claim

The central claim is that the rigorous mapping between a binary segmentation mask and its distance map can be learned once, by a small regression network, and then reused as a frozen geometric critic for any segmentation network. During training, the main network's soft probability map is fed into the frozen LR-Net, whose smooth-L1 regression loss pushes the whole system to produce masks whose implied distance maps match the true ones; at the same time, the MapDice loss reweights Dice by the boundary-emphasizing norm-inverse distance map. The paper reports that unfreezing the LR-Net degrades performance, which it reads as evidence that the pretrained, frozen mapping supplies a stable shape prior rather than a trainable module. The best configuration uses the norm-inverse distance map with ReLU activation, and the framework remains competitive on the public LiTS leaderboard while improving markedly over imbalance-focused losses on the clinical dataset.

Load-bearing premise

The load-bearing premise is that the LR-Net, pretrained on perfect binary masks, will still produce useful distance-map regression signals when it is fed the M-Net's soft probability maps during joint training; the paper's own ablation shows unfreezing the LR-Net hurts, so the method relies on this mismatch being benign.

Editorial extensions

If this is right

  • The frozen LR-Net is a plug-in: any classification-based segmentation backbone can receive distance-map supervision without architectural changes.
  • At inference time the LR-Net is removed, so the improved training signal costs nothing at deployment.
  • Distance-map variants matter: norm-inverse distance maps with ReLU activation outperform raw, inverse-only, sign, and sigmoid-mapped distance maps.
  • The MapDice loss and the regression loss are complementary; removing either drops Dice, and unfreezing the LR-Net drops Dice from 0.748 to 0.628 on the clinical dataset.
  • On the LiTS leaderboard submission the method reaches Dice per case 0.679 and Dice global 0.830, showing the gain transfers to a public benchmark.

Reading between the lines

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

  • Because the LR-Net is trained on binary ground-truth masks but evaluated on soft probability maps, its usefulness depends on that distribution shift being benign; a targeted test would be to train it on soft labels or to calibrate the main network's probabilities before regression.
  • The boundary-emphasizing weight map suggests MapDice could be combined with or compared against explicit boundary losses such as Hausdorff-style penalties; the paper does not test that combination.
  • The plug-in design implies the same frozen LR-Net could transfer to other small-object segmentation tasks such as pancreas or lymph-node lesions, but the paper only reports liver-tumor evidence.
  • A natural ablation would measure the LR-Net's regression error on the main network's soft maps at different training epochs to see whether the geometric signal is actually flowing through the frozen branch.
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

5 major / 6 minor

Summary. The manuscript proposes a segmentation framework that couples a main segmentation network (M-Net) with a lightweight distance-map regression network (LR-Net). The LR-Net is pre-trained on ground-truth binary masks to regress a normalized inverse distance map (NI-DM), then frozen; during joint training it receives soft probability maps from the M-Net and produces distance-map predictions, which are supervised by a smooth-L1 regression loss. In addition, the authors introduce a MapDice loss that weights a Dice-like term by the ground-truth distance map. Experiments on the MICCAI 2017 LiTS dataset and on a clinical CECT dataset report seven metrics and compare the method with several imbalance-oriented losses (WCE, GDS, Tversky, Focal Tversky, Exp-Log) as well as two earlier distance-map regression approaches. The best configuration reaches a Dice per case of 0.751 on the clinical dataset, outperforming the compared baselines, and the authors claim state-of-the-art or competitive performance.

Significance. If the mechanism is validated, the idea of pre-training a lightweight network to compute distance maps from masks and then freezing it as a differentiable shape-aware module is attractive: it can be plugged into any segmentation network and provides geometric supervision beyond binary masks. The paper also provides a useful comparison of several distance-map variants and imbalance losses on a clinical dataset. However, the support for the central claims is weakened by several definitional issues, an unreported balancing parameter, and the lack of any characterization of the LR-Net's behavior on soft probability inputs. The paper is not circular, and the empirical comparison is a genuine strength, but the current presentation leaves load-bearing points underspecified.

major comments (5)
  1. [§2.1, Eq. (1)] Equation (1) as written does not produce values in [0,1] as claimed. For a boundary voxel where D(x)=0, the formula gives φ(x) = (M+1)/M > 1, where M is the local maximum distance; the range is [1/M, 1+1/M]. If the intended normalization is instead (M−D(x))/M or (M+1−D(x))/(M+1), then the text and all downstream experiments need to be corrected, because the NI-DM is both the regression target for the LR-Net and the penalty map in MapDice. Please also clarify the notation: the variable x is used both for the evaluation point and as the index in the maximum over the connected component C.
  2. [§2.3, Eq. (2)] Equation (2) is not a well-defined scalar loss: it sums over classes but not over spatial positions, leaving the expression as a per-voxel tensor. Moreover, the term (p_c×φ_c+ε)/(p_c+φ_c+ε) evaluates to approximately 1 for background voxels where p_c=0 and φ_c=0, so unless the spatial sum is specified, the loss contains a large constant offset. Please provide the complete voxel-wise formulation and clarify how ε and the class sum are applied.
  3. [§2.3 and Table 1] The total loss includes α·L_smoothL1, but α is never defined, and its value is not reported anywhere. This is a free parameter tuned on the 13-case LiTS validation split, so the reported comparisons are not reproducible. The claim that α is 'verified to be effective for the performance improvement' is also unsupported by any ablation or sensitivity study. Please report the value of α and, ideally, a small sensitivity analysis.
  4. [§2.2 and Table 2, row 4] The training pipeline pre-trains the LR-Net on ground-truth binary masks and then freezes it, but during joint training the LR-Net receives soft probability maps from the M-Net. The paper asserts that the rigorous mapping trained on binary inputs 'can guarantee the correspondence' between M-Net outputs and LR-Net outputs, but no evidence is provided that the LR-Net generalizes from hard to soft inputs. The large drop when the LR-Net is unfrozen (DC 0.628 vs. 0.748, Table 2, rows 4 and 5) shows that the regression objective is sensitive to the training scheme, but it does not characterize the soft-input behavior. Please add a direct experiment on the LR-Net's soft-input predictions (e.g., distance error or visual comparison), or retrain the LR-Net with soft targets, to justify the frozen-binary design.
  5. [§3, Tables 1–2] All results are reported as single runs with no error bars or significance tests. The LiTS validation set contains only 13 cases, and the differences among the top configurations in Table 1 (e.g., 0.7463 vs. 0.7445) are small enough to be within noise. In addition, the clinical dataset section states that 'all the cases are used for testing' but does not describe how the model was trained or selected for this dataset; please clarify whether the LiTS-trained checkpoint was applied directly, and report standard deviations or confidence intervals over at least a few runs.
minor comments (6)
  1. [Abstract and §3.2] The abstract and conclusion claim to outperform 'other existing state-of-the-arts,' but the experimental comparison is limited to five loss functions and two earlier regression-based methods. No recent segmentation networks are compared, so the SOTA claim is broader than the evidence.
  2. [Table 1] The text cites baseline results of MNet+LDice (0.6581) and MNet+MapDice (0.6856), but these rows are not present in Table 1, which makes the comparison difficult to follow.
  3. [§3, Dataset] The clinical dataset description says all 137 cases are used for testing but does not state whether any clinical data were used for training or validation; please specify the training/evaluation protocol for this dataset.
  4. [§2.2 and Figure 3] There are several typos: 'During reference' should be 'During inference,' 'indicates indicates' is repeated, 'reslults' should be 'results,' 'Tvserky' should be 'Tversky,' 'I-DM is derived from the I-DM' should presumably read 'I-DM is derived from the O-DM,' and the Figure 3 caption repeats 'Fig. 3.'
  5. [§2.1] In Eq. (1), the connected component C is not formally defined in the notation; please define it as a set of voxels belonging to one object component in the binary mask.
  6. [§3.2] The LiTS leaderboard submission is reported without a date or version; since leaderboards change over time, this result should be described as a single submission rather than a stable state-of-the-art comparison.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the reported segmentation gains are empirical predictions tested on held-out data and against independent baselines.

full rationale

The paper's central empirical claim—that a frozen distance-map regression network coupled with a distance-weighted Dice loss improves small-object segmentation—is not a disguised restatement of its inputs. The LR-Net is pre-trained on deterministic distance maps computed from ground-truth binary masks via Eq. (1), and the regression target is the same ground-truth distance map; however, the M-Net's segmentation outputs are evaluated on a held-out clinical dataset and the public LiTS testing set, against baseline losses (WCE, GDS, Tversky, Focal Tversky, ExpLog) that do not share the proposed loss. The improvement is therefore an empirically tested prediction rather than a tautology. The MapDice loss in Eq. (2) is a modified Dice term with a distance-map penalty φ_c; it is fit to ground truth, but that is a training loss, not a claimed prediction. The only fitted scalar α is tuned on validation data (mentioned in Sec. 3.1), which is a mild reporting omission but does not reduce the result to an input by construction. No load-bearing self-citation appears: [12], the closest prior work on distance-transform guidance, is by different authors, and no 'uniqueness' theorem is invoked. The claim that the frozen LR-Net 'can therefore guarantee the correspondence between the outputs of the M-Net and the LR-Net' (Sec. 2.2) is an unverified distribution-shift assumption—LR-Net is trained on binary masks but applied to soft probability maps—but that is a correctness/robustness concern, not circularity, because the outcome could in principle falsify the method (indeed Table 2, row 4, shows unfreezing the LR-Net degrades performance from 0.748 to 0.628 DC).

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The method relies on the standard distance transform, but its two key design choices, frozen LR-Net and MapDice weighting, are heuristic and unvalidated beyond the reported experiments.

free parameters (2)
  • alpha (balancing weight for loss terms) = not reported
    Balances the classification loss and the smooth L1 regression loss; the paper states it is effective but does not give its value or selection procedure (Section 3.1).
  • epsilon (MapDice denominator constant) = not reported
    Small constant in the MapDice loss to avoid division by zero (Eq. 2); value not specified.
assumptions (3)
  • standard math Euclidean distance transform provides a rigorous, deterministic mapping between a binary mask and a distance map.
    Used to compute ground truth distance maps (Eq. 1, Section 2.1).
  • ad hoc to paper A lightweight network trained on ground-truth masks will generalize to soft probability maps when frozen.
    Section 2.2 states the LR-Net is trained independently on ground-truth masks and frozen; the paper does not justify this distribution shift beyond empirical results.
  • ad hoc to paper Weighting the Dice loss by distance map values improves shape awareness.
    Section 2.3 introduces MapDice as a heuristic; no derivation is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Deep Distance Map Regression Network with Shape-aware Loss for Imbalanced Medical Image Segmentation." pith.science (2026). https://pith.science/paper/WQ6FKSA6

@misc{pith2026250109116,
  author       = {Pith},
  title        = {Pith review of: Deep Distance Map Regression Network with Shape-aware Loss for Imbalanced Medical Image Segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WQ6FKSA6}},
  note         = {Machine review of arXiv:2501.09116}
}
read the original abstract

Small object segmentation, like tumor segmentation, is a difficult and critical task in the field of medical image analysis. Although deep learning based methods have achieved promising performance, they are restricted to the use of binary segmentation mask. Inspired by the rigorous mapping between binary segmentation mask and distance map, we adopt distance map as a novel ground truth and employ a network to fulfill the computation of distance map. Specially, we propose a new segmentation framework that incorporates the existing binary segmentation network and a light weight regression network (dubbed as LR-Net). Thus, the LR-Net can convert the distance map computation into a regression task and leverage the rich information of distance maps. Additionally, we derive a shape-aware loss by employing distance maps as penalty map to infer the complete shape of an object. We evaluated our approach on MICCAI 2017 Liver Tumor Segmentation (LiTS) Challenge dataset and a clinical dataset. Experimental results show that our approach outperforms the classification-based methods as well as other existing state-of-the-arts.

Figures

Figures reproduced from arXiv: 2501.09116 by the authors.

Figure 1
Figure 1. Illustration of the main segmentation network with the LR-Net. capturing the distinct geometric properties of the small object by distance map regression. An illustration of the overall segmentation system is shown in [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. An illustration of different kinds of distance maps. 3.1 Comparison of Various Distance Maps Regression A simple illustration of distance maps is shown in [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Liver tumor segmentation results of different methods on the LiTS vali [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

16 extracted references · 11 canonical work pages

  1. [1]

    In: 2019 IEEE 16th International Symposium on Biomedical Imaging (ISBI 2019)

    Abraham, N., Khan, N.M.: A novel focal tversky loss function with improved atten- tion u-net for lesion segmentation. In: 2019 IEEE 16th International Symposium on Biomedical Imaging (ISBI 2019). pp. 683–687. IEEE (2019)

    work page 2019

  2. [2]

    Computer Vision and Image Understanding 189, 102809 (2019)

    Audebert, N., Boulch, A., Le Saux, B., Lef` evre, S.: Distance transform regres- sion for spatially-aware deep semantic segmentation. Computer Vision and Image Understanding 189, 102809 (2019)

    work page 2019

  3. [3]

    arXiv preprint arXiv:1901.04056 (2019)

    Bilic, P., Christ, P.F., Vorontsov, E., Chlebus, G., Chen, H., Dou, Q., Fu, C.W., Han, X., Heng, P.A., Hesser, J., et al.: The liver tumor segmentation benchmark (lits). arXiv preprint arXiv:1901.04056 (2019)

    arXiv 2019

  4. [4]

    IEEE transactions on medical imaging 25(11), 1451–1461 (2006)

    Crum, W.R., Camara, O., Hill, D.L.: Generalized overlap measures for evaluation and validation in medical image analysis. IEEE transactions on medical imaging 25(11), 1451–1461 (2006)

    work page 2006

  5. [5]

    Medical physics 46(12), 5637–5651 (2019)

    Dangi, S., Linte, C.A., Yaniv, Z.: A distance map regularized cnn for cardiac cine mr image segmentation. Medical physics 46(12), 5637–5651 (2019)

    work page 2019

  6. [6]

    In: International Conference on Medical Image Computing and Computer-Assisted Intervention

    Gao, Y., Huang, R., Chen, M., Wang, Z., Deng, J., Chen, Y., Yang, Y., Zhang, J., Tao, C., Li, H.: Focusnet: Imbalanced large and small organ segmentation with an end-to-end deep neural network for head and neck ct images. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. pp. 829–838. Springer (2019)

    work page 2019

  7. [7]

    Medical image analysis 35, 18–31 (2017)

    Havaei, M., Davy, A., Warde-Farley, D., Biard, A., Courville, A., Bengio, Y., Pal, C., Jodoin, P.M., Larochelle, H.: Brain tumor segmentation with deep neural net- works. Medical image analysis 35, 18–31 (2017)

    2017

  8. [8]

    In: Proceedings of the IEEE interna- tional conference on computer vision

    He, K., Zhang, X., Ren, S., Sun, J.: Delving deep into rectifiers: Surpassing human- level performance on imagenet classification. In: Proceedings of the IEEE interna- tional conference on computer vision. pp. 1026–1034 (2015)

    2015

Show all 16 references
  1. [9]

    Expert Systems with Applications 130, 124–131 (2019)

    Huang, Y., Chen, Y., Zhu, H., Li, W., Ge, Y., Huang, X., He, J.: A liver fibrosis staging method using cross-contrast network. Expert Systems with Applications 130, 124–131 (2019)

    2019

  2. [10]

    IEEE Access 7, 24898–24909 (2019)

    Jiang, H., Shi, T., Bai, Z., Huang, L.: Ahcnet: An application of attention mech- anism and hybrid connection for liver tumor segmentation in ct volumes. IEEE Access 7, 24898–24909 (2019)

    2019

  3. [11]

    In: Proceedings of the IEEE international conference on computer vision

    Lin, T.Y., Goyal, P., Girshick, R., He, K., Doll´ ar, P.: Focal loss for dense object detection. In: Proceedings of the IEEE international conference on computer vision. pp. 2980–2988 (2017)

    2017

  4. [12]

    In: Medical Imaging with Deep Learning (2020)

    Ma, J., Wei, Z., Zhang, Y., Wang, Y., Lv, R., Zhu, C., Chen, G., Liu, J., Peng, C., Wang, L., et al.: How distance transform maps boost segmentation cnns: An empirical study. In: Medical Imaging with Deep Learning (2020)

    2020

  5. [13]

    In: International Conference on Medical image computing and computer-assisted intervention

    Ronneberger, O., Fischer, P., Brox, T.: U-net: Convolutional networks for biomedi- cal image segmentation. In: International Conference on Medical image computing and computer-assisted intervention. pp. 234–241. Springer (2015)

    2015

  6. [14]

    In: International Workshop on Machine Learning in Medical Imaging

    Salehi, S.S.M., Erdogmus, D., Gholipour, A.: Tversky loss function for image seg- mentation using 3d fully convolutional deep networks. In: International Workshop on Machine Learning in Medical Imaging. pp. 379–387. Springer (2017) 10 Huiyu Li et al

    2017

  7. [15]

    NeuroImage 155, 159–168 (2017)

    Valverde, S., Cabezas, M., Roura, E., Gonz´ alez-Vill` a, S., Pareto, D., Vilanova, J.C., Rami´ o-Torrent` a, L., Rovira,`A., Oliver, A., Llad´ o, X.: Improving automated multi- ple sclerosis lesion segmentation with a cascaded 3d convolutional neural network approach. NeuroIm...

    2017

  8. [16]

    In: International Conference on Medical Image Computing and Computer-Assisted Intervention

    Wong, K.C., Moradi, M., Tang, H., Syeda-Mahmood, T.: 3d segmentation with exponential logarithmic loss for highly unbalanced object sizes. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. pp. 612–619. Springer (2018)

    2018

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.