Pith. sign in

REVIEW 3 major objections 5 minor 24 references

Dual-Attention U-Net++ with Class-Specific Ensembles and Bayesian Hyperparameter Optimization for Precise Wound and Scale Marker Segmentation

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read A dual-attention U-Net++ with an EfficientNet-B7 encoder, class-specific ensembles, and Bayesian hyperparameter tuning achieves a weighted F1 of 0.8640 on the NBC 2025 and PCBBE 2025 wound and scale-marker segmentation benchmark.

desk verdict A single unverified competition F1 is the whole load-bearing wall; the paper itself says official evaluation was still ongoing when it quotes the final number. read the letter →

arxiv 2507.05314 v1 pith:PKPRKDJY submitted 2025-07-07 eess.IV cs.AIcs.CVcs.LG

Daniel Cieślak , Miriam Reca , Olena Onyshchenko , Jacek Rumiński This is my paper
classification eess.IVcs.AIcs.CVcs.LG
keywords ImagesegmentationWoundassessmentDeeplearningMedicalimagingU-Net++AttentionmechanismBayesianoptimizationTesttimeaugmentation
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 automated wound and scale-marker segmentation can be made accurate and stable enough for clinical use by combining an attention-enhanced U-Net++ with a powerful encoder, per-class training, and careful hyperparameter search. The authors benchmark many architectures and encoders, find U-Net++ with EfficientNet-B7 best, then train separate models for wounds and scale markers with tailored augmentation and Bayesian optimization. They report an externally computed weighted F1 of 0.8640 on the NBC 2025 and PCBBE 2025 challenge test set. If correct, the result demonstrates that a relatively conventional CNN architecture, carefully tuned and ensembled, can handle severely imbalanced clinical segmentation tasks without transformers.

What carries the argument

The load-bearing object is the dual-attention U-Net++: a U-Net++ with nested skip connections, enhanced by a spatial attention module and an SCSE (simultaneous channel and spatial excitation) module in the decoding path, using EfficientNet-B7 as the encoder. Its role is to recalibrate feature maps so that small, irregular wounds and scale markers are emphasized despite severe class imbalance. Class-specific training and test-time augmentation then stabilize the predictions, and the Bayesian sweep tunes loss weights, augmentation strength, learning rate, and mixup alpha for each class-specific model.

What would settle it

Ask the competition organizers to release per-class F1 scores for the submitted predictions and recompute the 75/25 weighted combination; if those numbers do not reproduce 0.8640, the headline result is falsified.

Watch

Extended reading notes

Core claim

The central claim is that the proposed pipeline—dual-attention U-Net++ (spatial attention plus SCSE channel attention) with an ImageNet-pretrained EfficientNet-B7 encoder, independently trained wound and scale-marker models, Bayesian hyperparameter sweeps, and test-time augmentation—attains a weighted F1-score of 0.8640 on the NBC 2025 and PCBBE 2025 segmentation benchmark. The weighted F1 mixes wound segmentation (75%) and scale-marker segmentation (25%) and was computed by the competition organizers on undisclosed hardware. Internally, the paper reports validation Dice averages of 0.9718 for wounds and 0.9185 for scale markers across four runs. The paper notes that official evaluation was still ongoing when the manuscript was prepared, so the 0.8640 is presented as the final externally validated score that anchors the approach's claimed transfer beyond its own validation split.

Load-bearing premise

The load-bearing premise is that the single externally reported weighted F1 of 0.8640 was computed correctly from the described submission; the paper says official evaluation was still ongoing and offers no confidence interval, per-class external scores, or submission protocol.

Editorial extensions

If this is right

  • Following the same pipeline—dual-attention U-Net++, EfficientNet-B7, per-class training, and test-time augmentation—should reproduce the reported benchmark performance on similar wound-imaging data.
  • Training separate models for each class sidesteps the need for complex multi-class loss weighting, since each model sees only one foreground class against background.
  • The reported internal Dice averages of 0.9718 for wounds and 0.9185 for scale markers give future work concrete baselines to beat on this dataset.
  • The weighted F1 of 0.8640 sets a reference point for what a carefully tuned CNN ensemble can achieve on the NBC 2025 and PCBBE 2025 challenge.

Reading between the lines

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

  • The authors leave implicit that the gap between the high internal Dice scores and the weighted F1 of 0.8640 could signal test-time distribution shift; per-class external scores would resolve this.
  • A natural extension would be to ablate the auxiliary pretraining dataset to measure how much of the gain comes from extra data rather than the dual-attention architecture and tuning.
  • The single externally computed score would carry more weight if paired with a confidence interval across multiple runs, since the paper does not report one.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The manuscript reports a deep-learning pipeline for segmenting wounds and scale markers in clinical images, developed for the NBC 2025 & PCBBE 2025 challenge. The method combines a U-Net++ decoder with a dual-attention module (SCSE and spatial attention), an EfficientNet-B7 encoder, separate models per class, Bayesian hyperparameter optimization (WandB sweeps), test-time augmentation, and optional CRF post-processing. Internal validation Dice are 0.9718 (wounds) and 0.9185 (scale markers), and the paper's headline result is an externally computed weighted F1-score of 0.8640. The manuscript also reports 5-fold benchmarking of multiple architectures and encoders and details the preprocessing and training protocol.

Significance. The paper's potential significance rests on the credibility and generalizability of the external 0.8640 F1-score. If this number is confirmed by the competition organizers and correctly attributed to the described ensemble, the work provides a competent challenge solution and a useful reference configuration for imbalanced medical image segmentation. The external computation by organizers is a genuine strength, as is the systematic 5-fold architecture/encoder benchmarking and the per-class model strategy. However, the current manuscript does not substantiate several load-bearing attributions: the dual-attention module's contribution is never isolated by an ablation, the encoder selection narrative is self-contradictory and omits EfficientNet-B7 from the reported benchmark, and the single externally reported F1 lacks confidence intervals, per-class external scores, and a verification protocol. The scientific novelty is modest because the components are established techniques; the main value is as an engineering benchmark report pending confirmation.

major comments (3)
  1. [Section 4 (Results)] The manuscript states "At the time of manuscript preparation, our models' official evaluation by competition organizers is ongoing; preliminary internal tests suggest high reliability and robustness" and then, immediately after, asserts "Our final externally validated weighted F1-score was 0.8640." These two statements are mutually incompatible: a final externally validated score cannot exist while official evaluation is still ongoing. Because the abstract and conclusion both rest on this single number, the central claim is currently unverifiable. Please provide the organizers' score report, submission identifiers, evaluation date, and the exact protocol used to compute the weighted F1, including per-class external scores; without this, the 0.8640 must be treated as preliminary, not as a validated result.
  2. [Sections 3.3, 3.7, and 5 (Discussion)] There is no ablation isolating the dual-attention module. Section 3.3 introduces the dual-attention enhancement of U-Net++, and the Discussion claims it "provided significant improvements," but Table 2 reports only the final model's Dice. To support this attribution, the paper must compare U-Net++ with and without the SCSE+Spatial module under identical training, hyperparameters, data splits, and ensemble settings, and ideally also ablate TTA, CRF, and the class-specific training strategy. Without such comparisons, the architectural novelty claimed in the title and abstract is not demonstrated.
  3. [Sections 3.3, 5, and Abstract] The encoder selection is described inconsistently. Section 3.3 says the benchmarking "identified U-Net++ with EfficientNet-B7 encoder as superior," and the abstract repeats that EfficientNet-B7 was "identified as the optimal encoder backbone." However, Section 3.3 lists the benchmarked encoders as DenseNet121, EfficientNet-B0, ResNet34, VGG16, MiT-B0, and MobileNetV2, and Section 5 explicitly states that "initial benchmarking results (Fig. 1, Fig. 2) demonstrated EfficientNet-B0 as the top-performing encoder" and that B7 was chosen "due to its architectural similarity yet greater representational capacity." These claims cannot all be true; B7 was apparently never benchmarked in the reported experiments and was therefore not empirically identified as optimal. Please clarify whether B7 was benchmarked, provide the corresponding results, or revise the abstract and Section 3.3 to present B7 as a design choice rather than an empirical finding.
minor comments (5)
  1. [Section 3.1 (Dataset)] Please clarify how the 163 training images without a second label were used for the scale-marker model: were they excluded, treated as background, or ignored in the loss for that class? This detail affects the reproducibility and the interpretation of the scale-marker Dice.
  2. [Section 3.4 and Table 2] The text says "Final performance was averaged across five runs," but Table 2 reports four runs. Please reconcile this discrepancy and report the standard deviation or inter-run variability for each class.
  3. [Appendix A, Algorithm 1] The ensemble weights w1 and w2, the number of CRF iterations, and whether CRF was used in the final submission are not specified. Please report these values and, if CRF was used, include an ablation showing its effect on the weighted F1.
  4. [Appendix B (Hyperparameter Optimization)] The Bayesian sweep description is not reproducible: it gives no sweep ranges for the loss ratios, mixup alpha, augmentation strength, or learning rate, no number of trials, and no objective criterion for the sweep. The statement that configurations are "available in the project's WandB repository" lacks a URL or access instructions.
  5. [Figures 1 and 2] Figures 1 and 2 need axis labels, metric names, and error bars, and the text should state which validation metric (Dice or IoU) is shown. Currently the reader cannot assess the reported benchmarking ranking from the figures.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the reported 0.8640 weighted F1 is an externally computed benchmark score, and no derivation step reduces to its own input.

full rationale

The paper's claimed chain is: benchmark architectures and encoders via 5-fold cross-validation, select a pipeline, train class-specific dual-attention U-Net++ models with Bayesian hyperparameter tuning, ensemble with test-time augmentation, and report an externally computed weighted F1-score. None of these steps defines a predicted quantity in terms of a fitted parameter that produced it. The internal validation Dice scores in Table 2 and the external 0.8640 weighted F1 are separate measurements, and the paper explicitly states the external score was calculated by competition organizers on undisclosed hardware. There is no self-citation chain used to justify the central result, no imported uniqueness theorem, and no ansatz smuggled in through prior work; the architecture is a standard composition of published components evaluated de novo. The only notable issues are provenance and consistency concerns: Section 4 says official evaluation was still ongoing and then immediately reports a final externally validated F1, and the Discussion admits benchmarking ranked EfficientNet-B0 first while the abstract says EfficientNet-B7 was identified as optimal. These are verification and reporting problems, not circularity, because neither makes the output equal to its input by construction. The derivation is therefore self-contained relative to the external benchmark, and the appropriate circularity score is 0.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

These four premises are load-bearing for the pipeline. The free parameters are numerical choices tuned to validation data; no invented entities are introduced. Together they show the report is an engineering configuration dependent on measured validation performance and transfer assumptions, not a first-principles derivation.

free parameters (6)
  • Mixup alpha = 0.3604
    Tuned with Weights and Biases Bayesian sweeps; reported in Section 3.5 and Appendix B.
  • Augmentation strength = 1.137
    Tuned via Bayesian sweeps; controls brightness, contrast, and geometric distortion intensity in Section 3.2.
  • Loss ratios (BCE, Dice, Focal) = Not reported
    Optimized via WandB sweeps per Appendix B, but the final values are omitted, preventing exact reproduction.
  • Learning rate = 0.001
    Listed as AdamW learning rate in Section 3.5 and described as part of the optimized hyperparameter set.
  • Ensemble weights w1, w2 = Not reported
    Algorithm 1 uses w1 and w2 to combine two scale-marker models, but the final weights are not given.
  • CRF iterations = Not reported
    Algorithm 1 requires CRF iterations, but the value is unspecified and CRF is described as optional.
assumptions (4)
  • domain assumption ImageNet-pretrained EfficientNet features transfer to wound and scale marker images.
    Section 3.5 specifies an ImageNet-pretrained EfficientNet-B7 encoder and ImageNet normalization; the paper does not test training from scratch or domain-specific pretraining.
  • domain assumption Internal validation Dice is a reliable proxy for the externally computed weighted F1 during model selection.
    Architectures, encoders, loss ratios, augmentation strength, mixup alpha, and ensemble composition were chosen from 5-fold cross-validation and validation Dice in Sections 3.3 and 3.4, while the final metric weights wounds 75% and scale markers 25%.
  • domain assumption The challenge annotations are correct and sufficiently consistent for training.
    Section 3.1 describes the dataset but reports no annotation quality control, inter-rater agreement, or handling of the 163 images without a second label.
  • ad hoc to paper A larger-capacity EfficientNet-B7 generalizes better than the empirically preferred EfficientNet-B0.
    Section 5, Clarification on EfficientNet Encoder Selection, states initial benchmarking showed B0 as top, but B7 was chosen because of representational capacity and presumed generalization; this premise contradicts the reported measurements.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Dual-Attention U-Net++ with Class-Specific Ensembles and Bayesian Hyperparameter Optimization for Precise Wound and Scale Marker Segmentation." pith.science (2026). https://pith.science/paper/PKPRKDJY

@misc{pith2026250705314,
  author       = {Pith},
  title        = {Pith review of: Dual-Attention U-Net++ with Class-Specific Ensembles and Bayesian Hyperparameter Optimization for Precise Wound and Scale Marker Segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PKPRKDJY}},
  note         = {Machine review of arXiv:2507.05314}
}
read the original abstract

Accurate segmentation of wounds and scale markers in clinical images remainsa significant challenge, crucial for effective wound management and automatedassessment. In this study, we propose a novel dual-attention U-Net++ archi-tecture, integrating channel-wise (SCSE) and spatial attention mechanisms toaddress severe class imbalance and variability in medical images effectively.Initially, extensive benchmarking across diverse architectures and encoders via 5-fold cross-validation identified EfficientNet-B7 as the optimal encoder backbone.Subsequently, we independently trained two class-specific models with tailoredpreprocessing, extensive data augmentation, and Bayesian hyperparameter tun-ing (WandB sweeps). The final model ensemble utilized Test Time Augmentationto further enhance prediction reliability. Our approach was evaluated on a bench-mark dataset from the NBC 2025 & PCBBE 2025 competition. Segmentationperformance was quantified using a weighted F1-score (75% wounds, 25% scalemarkers), calculated externally by competition organizers on undisclosed hard-ware. The proposed approach achieved an F1-score of 0.8640, underscoring itseffectiveness for complex medical segmentation tasks.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

24 extracted references · 9 canonical work pages

  1. [1]

    write newline

    " write newline " cite write " FUNCTION editor.postfix editor num.names #1 > "( )" "( )" if FUNCTION editor.trans.postfix editor num.names #1 > "( )" "( )" if FUNCTION trans.postfix translator num.names #1 > "( )" "( )" if FUNCTION authors.editors.reflist.apa5 'field := 'dot := field num.names 'numnames := numnames 'format.num.names := format.num.names na...

  2. [2]

    sn-aps.bst

    FUNCTION identify.aps.version "sn-aps.bst" " [2024/07/19 v1.1 APS bibliography style]" * top ENTRY address author booktitle chapter doi edition editor eid howpublished institution journal key keywords month note number organization pages publisher school series title type url volume year eprint archive archivePrefix primaryClass adsurl adsnote version lab...

    2024

  3. [3]

    write newline

    " write newline "" before.all 'output.state := FUNCTION if.digit duplicate "0" = swap duplicate "1" = swap duplicate "2" = swap duplicate "3" = swap duplicate "4" = swap duplicate "5" = swap duplicate "6" = swap duplicate "7" = swap duplicate "8" = swap "9" = or or or or or or or or or FUNCTION n.separate 't := "" #0 'numnames := t empty not t #-1 #1 subs...

  4. [4]

    , Baqersad , J

    barticle Yazdi , S.J.M. , Baqersad , J. : Mechanical modeling and characterization of human skin: A review . Journal of Biomechanics 130 , 110864 ( 2022 ) 10.1016/j.jbiomech.2021.110864 barticle

    work page arXiv 2022

  5. [5]

    , Ladyzynski , P

    barticle Foltynski , P. , Ladyzynski , P. : Internet service for wound area measurement using digital planimetry with adaptive calibration and image segmentation with deep convolutional neural networks . Biocybernetics and Biomedical Engineering 43 ( 1 ), 17 -- 29 ( 2023 ) 10.1016/j.bbe.2022.11.004 barticle

    work page doi:10.1016/j.bbe.2022.11.004 2023

  6. [6]

    , Wang , C

    barticle Li , F. , Wang , C. , Liu , X. , Peng , Y. , Jin , S. : A composite model of wound segmentation based on traditional methods and deep neural networks . Computational and Mathematical Methods in Medicine ( 2018 ) 10.1155/2018/4149103 barticle

    work page doi:10.1155/2018/4149103 2018

  7. [7]

    , Fischer , P

    botherref Ronneberger , O. , Fischer , P. , Brox , T. : U-Net: Convolutional Networks for Biomedical Image Segmentation (2015). https://arxiv.org/abs/1505.04597 botherref

    arXiv 2015

  8. [8]

    , Zahia , S

    barticle Niri , R. , Zahia , S. , Stefanelli , A.e.a. : Wound segmentation with u-net using a dual attention mechanism and transfer learning . Journal of Digital Imaging ( 2025 ) 10.1007/s10278-025-01386-w barticle

    work page doi:10.1007/s10278-025-01386-w 2025

Show all 24 references
  1. [9]

    , Wang , W

    botherref Xie , E. , Wang , W. , Yu , Z. , Anandkumar , A. , Alvarez , J.M. , Luo , P. : SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers (2021) botherref

    2021

  2. [10]

    , Rahman Siddiquee , M.M

    bchapter Zhou , Z. , Rahman Siddiquee , M.M. , Tajbakhsh , N. , Liang , J. : Unet++: A nested u-net architecture for medical image segmentation . In: Stoyanov , D. , Taylor , Z. , Carneiro , G. , Syeda-Mahmood , T. , Martel , A. , Maier-Hein , L. , Tavares , J.M.R.S. , Bradley...

    2018

  3. [11]

    , Zhu , Y

    bchapter Chen , L.-C. , Zhu , Y. , Papandreou , G. , Schroff , F. , Adam , H. : Encoder-decoder with atrous separable convolution for semantic image segmentation . In: Proceedings of the European Conference on Computer Vision (ECCV) , pp. 801 -- 818 ( 2018 ). 10.1007/978-3-030...

    2018 doi

  4. [12]

    , Li , D

    botherref Cui , L. , Li , D. , Yang , X. , Liu , C. : Towards reliable healthcare imaging: conditional contrastive generative adversarial network for handling class imbalance in mr images. Journal of Biomedical Informatics (2023) botherref

    2023

  5. [13]

    , Sieberth , T

    barticle Zimmermann , N. , Sieberth , T. , Dobay , A. : Automated wound segmentation and classification of seven common injuries in forensic medicine . Forensic Science International ( 2023 ) 10.1007/s12024-023-00668-5 barticle

    2023 doi

  6. [14]

    , Mahbod , A

    botherref Wang , C. , Mahbod , A. , Ellinger , I. , Galdran , A. , Gopalakrishnan , S. , Niezgoda , J. , Yu , Z. : Fuseg: The foot ulcer segmentation challenge. Information 15(3) (2024) 10.3390/info15030140 botherref

    2024 doi

  7. [15]

    sn-basic.bst

    FUNCTION identify.basic.version "sn-basic.bst" " [2024/07/19 v1.1 bibliography style]" * top ENTRY address archive author booktitle chapter doi edition editor eid eprint howpublished institution journal key keywords month note number organization pages publisher school series ...

    2024

  8. [16]

    write newline

    " write newline "" before.all 'output.state := FUNCTION add.period duplicate empty 'skip "." * add.blank if FUNCTION if.digit duplicate "0" = swap duplicate "1" = swap duplicate "2" = swap duplicate "3" = swap duplicate "4" = swap duplicate "5" = swap duplicate "6" = swap dupl...

  9. [17]

    write newline

    " write newline "" before.all 'output.state := FUNCTION output.doi doi empty skip "doi:" doi * "" * output if FUNCTION format.archive archivePrefix empty "" archivePrefix ":" * if FUNCTION format.primaryClass primaryClass empty "" " [" primaryClass * "] " * if FUNCTION format....

  10. [18]

    write newline

    " write newline "" before.all 'output.state := FUNCTION string.to.integer 't := t text.length 'k := #1 'char.num := t char.num #1 substring 's := s is.num s "." = or char.num k = not and char.num #1 + 'char.num := while char.num #1 - 'char.num := t #1 char.num substring FUNCTI...

  11. [19]

    write newline

    " write newline "" before.all 'output.state := FUNCTION string.to.integer 't := t text.length 'k := #1 'char.num := t char.num #1 substring 's := s is.num s "." = or char.num k = not and char.num #1 + 'char.num := while char.num #1 - 'char.num := t #1 char.num substring FUNCTI...

  12. [20]

    sn-nature.bst

    FUNCTION identify.nature.version "sn-nature.bst" " [2024/07/19 v1.1 bibliography style]" * top ENTRY address archive author booktitle chapter edition editor eprint howpublished institution journal key keywords month note number organization pages publisher school series title ...

    2024

  13. [21]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

  14. [22]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

  15. [23]

    sn-vancouver-num.bst

    FUNCTION identify.vancouver.version "sn-vancouver-num.bst" " [2024/07/19 v1.1 Vancouver bibliography style]" * top ENTRY address assignee author booktitle chapter cartographer day edition editor howpublished institution inventor journal key keywords month note number organizat...

    2024

  16. [24]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.