Pith. sign in

REVIEW 4 major objections 5 minor 57 references

Enhancing Fairness in Skin Lesion Classification for Medical Diagnosis Using Prune Learning

T0 review · 4 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read A purely statistical skewness signal can identify and prune skin-tone bias in skin-lesion classifiers, improving fairness and shrinking the model.

desk verdict Skewness-based pruning for fairness is a plausible new idea with a useful ViT extension, but the empirical evidence is too thin to establish the central claim. read the letter →

arxiv 2509.00745 v1 pith:K3DC5G3S submitted 2025-08-31 cs.CV cs.AIcs.CYcs.LG

Kuniko Paxton , Koorosh Aslansefat , Dhavalkumar Thakker , Yiannis Papadopoulos , Tanaya Maslekar This is my paper · ORCID
classification cs.CVcs.AIcs.CYcs.LG
keywords skinlesionclassificationfairnesspruningskewnessVisionTransformerVGGmodelcompressiondermoscopicimages
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 proposes that internal model components whose activations are negatively skewed — spread widely across an image — are the ones encoding skin-tone information in skin-lesion classifiers, and that pruning these components makes the model fairer across skin tones. The authors test this on VGG11 and ViT-B16 using the ISIC2019 dataset, keeping only channels, patches, and attention heads whose median activation skewness is positive (localized, lesion-like). They report improved equalized-odds fairness metrics with little or no accuracy loss, and smaller models with lower FLOPs and memory. Because the criterion is purely statistical, the method needs no skin-tone labels, no group definitions, and no adversarial training.

What carries the argument

The load-bearing object is the skewness statistic computed on per-channel feature maps (CNNs) and per-head attention maps (ViTs). Negative skewness is read as widespread moderate activation over skin; positive skewness as localized high activation over a lesion. Algorithm 1 computes the median skewness of each filter/head over the validation set and discards components with non-positive median; Algorithm 2 physically removes the filters and adjusts the classifier input size; Algorithm 3 keeps ViT dimensions consistent by zero-padding pruned residual connections. This statistic carries the entire argument because it identifies 'skin-tone-related' components without attribute labels.

What would settle it

Train the same VGG and ViT models, then prune the same number of components chosen at random, or chosen with positive skewness. If fairness improves as much as with negative-skewness pruning, the skewness criterion is not the causal mechanism. Alternatively, use Grad-CAM or a probe classifier to show that the removed channels and heads were not preferentially selective for skin regions over lesion regions; if they were not, the pruning is removing something other than skin-tone encoding.

Watch

Extended reading notes

Core claim

The central discovery is that the skewness of feature-map and attention-map distributions is a usable, label-free signal for skin-tone bias in lesion classifiers. In dermoscopic images, skin occupies most of the frame and the lesion a small region, so a component that fires broadly (negative skewness) is plausibly tracking skin tone, while one that fires in a small hotspot (positive skewness) is tracking the lesion. The authors prune all components with negative median skewness and fine-tune the model. This improves EOpp1 and EOdd by about 1.5 percentage points on the VGG and 1.1 points in the best ViT configuration, with accuracy roughly unchanged and computation reduced. On the ViT, the be

Load-bearing premise

The load-bearing premise is that a negatively skewed activation distribution reliably marks a component as skin-tone-focused rather than as coding any other broad background structure or lesion-adjacent texture — an assumption the paper does not directly verify with attribution or visualization.

Editorial extensions

If this is right

  • Fairness can be improved without skin-tone labels, group definitions, or adversarial debiasing, removing a practical barrier for clinical deployment.
  • The pruning produces smaller and faster models, so fairer classifiers become lighter and cheaper to run, including on edge devices in low-resource settings.
  • The same skewness criterion applies to both convolutional and transformer architectures, suggesting a common mechanism for bias removal.
  • On the ViT, freezing the pruned patch-embedding weights during fine-tuning preserves the fairness gains, implying that retraining can undo bias removal if not constrained.

Reading between the lines

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

  • A random-pruning control with the same pruning budget would isolate whether the fairness gains come from removing skin-tone-specific components or from compression-induced regularization.
  • Applying the skewness criterion to other medical imaging tasks, where background structure differs, would show whether the positive/negative skewness separation is a general bias signal or specific to dermoscopy.
  • Evaluating with finer-grained skin tone groups rather than a binary light/dark split would show whether the method reduces bias across the full spectrum.
  • The same criterion could be tested on other vision backbones, such as ResNet or Swin Transformers, to see if the positive-skewness direction always aligns with lesion-focused attention.
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

4 major / 5 minor

Summary. The paper proposes a skewness-based pruning method ('SkewPrune') to improve fairness in skin lesion classification. For CNNs (VGG11), it computes the skewness of each channel's feature map after ReLU and max pooling, retaining channels with positive median skewness and pruning negatively skewed channels under the assumption that these encode skin-tone information. For ViT-B16, it applies analogous pruning to patch-embedding channels and attention heads, exploring six pruning/fine-tuning patterns. Fairness is measured by EOpp0 (TNR gap), EOpp1 (TPR gap), and EOdd; the authors report that pruning improves fairness on ISIC2019 while preserving accuracy and reducing computational cost. The strongest reported gains are small: VGG EOdd improves from 7.76% to 6.10%; ViT's best pattern improves EOdd from 8.22% to 7.11%.

Significance. If validated, the method would be an attractive fairness intervention because it avoids explicit skin-tone labels, is simple to implement, and simultaneously yields model compression. The paper also provides useful comparisons against SCP-FairPrune and reports FLOPs, parameter counts, and memory footprints. However, the current empirical support is not sufficient. The fairness gains are single-run point estimates with no error bars, no multiple seeds, and no significance testing; for ViT the authors select the best of six pruning patterns post hoc. The central assumption that negative skewness identifies skin-tone-specific components is asserted but never directly verified. These issues are load-bearing for the paper's main claims, although they are addressable with additional experiments. The concept is worth pursuing, but the manuscript in its present form does not convincingly establish the claimed improvements.

major comments (4)
  1. [Section V-B, Table III] The central fairness claim rests on single-run point estimates. No error bars, multiple seeds, or statistical significance tests are reported. The authors test six ViT pruning patterns and present Pattern 6 as the headline (EOdd 8.22→7.11, EOpp1 7.42→6.24), yet Pattern 3 worsens EOpp1 from 7.42% to 9.95%. With a single ISIC2019 train/validation/test split (Section IV-A), differences of 1–2 percentage points could easily be run-to-run noise. Please report per-seed results with confidence intervals, or a paired significance test, and either pre-specify the pruning pattern or correct for multiple comparisons.
  2. [Algorithm 1 and Section III-A] The method's load-bearing assumption is that a negatively skewed feature/attention distribution indicates a component that encodes skin-tone information, while a positively skewed distribution indicates lesion focus. This is asserted but never verified via feature visualization, attribution, or skin-tone perturbation. Negative skewness could also arise from broad background patterns, illumination, or lesion-adjacent texture. Please add a direct test: for example, show that pruned components respond more to skin-tone changes than to other image perturbations, or include a randomization/magnitude-pruning control that removes the same number of channels without the skewness criterion.
  3. [Section V-B, Table III] The claimed computational reduction for ViT is very small: GFLOPs 11.29→11.22 (0.6%), parameters 57.3M→56.97M (0.6%), memory footprint 327.3M→320.19M (2.2%). Describing these as 'significant' reductions (Section V-B) is not supported by the table. Please report relative reductions and temper the RQ3 claim for ViT. The VGG11 reductions (FLOPs 7.61→6.99, params 128.8M→107.91M, memory 491.33M→411.65M) are more substantial and should be distinguished from the ViT case.
  4. [Section IV/V] There is no control that prunes the same number of channels, patches, or heads using a neutral criterion, such as random pruning, magnitude pruning, or pruning based on activation mean. Such a control is necessary to attribute observed fairness improvements to the skewness criterion rather than to generic structural pruning and fine-tuning. The SCP-FairPrune comparisons in Table II are informative but do not isolate this mechanism, especially since SCP-FairPrune uses a different pruning objective and different prune/fine-tune schedule.
minor comments (5)
  1. [Section III-C] The metrics EOpp0 and EOpp1 are called 'Equal Opportunity', but EOpp0 is a TNR gap and EOpp1 is a TPR gap. Standard equal opportunity is usually defined via TPR only. Consider renaming them 'TNR gap' and 'TPR gap' to avoid confusion.
  2. [Section IV-B] Typo: 'VCG11' should be 'VGG11'. Also 'FLOPS' and 'FLOPs' are used inconsistently; choose one.
  3. [Introduction and Related Work] Typos: 'Unearning' should be 'Unlearning'; in Related Work, 'during interference from inputs' should be 'during inference from inputs'.
  4. [Table I] The table header contains 'Fine-Turning' (should be 'Fine-Tuning'), and the columns 'Prune1'/'Prune2' are not explained in the table caption; the text should define them explicitly.
  5. [Section III-B] The text says the input dimension is reduced 'from 786 to 336', but ViT-B16's patch embedding dimension is 768, not 786. Also, Algorithm 3's zero-padding procedure is underspecified: when d > |K| it pads, but the positions k_i are not defined, and the case d < |K| is not handled. Please clarify.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; pruning rule is fixed and fairness is measured independently.

full rationale

The paper's derivation chain is not circular. Algorithm 1 selects channels purely from the sign of the median skewness of feature maps/attention maps; the threshold (median > 0) is fixed and is never fitted to the fairness metrics EOpp0/EOpp1/EOdd. The fairness metrics are computed on a held-out ISIC2019 test split from [33], so the reported fairness values are independent of the pruning criterion. The conclusion that negative-skewness components are 'skin-color associated' is an untested interpretive premise—a validity/assumption risk, not a circularity, because the pruning could have failed empirically (and indeed Pattern 3 in Table III worsens EOdd from 8.22% to 10.95%). No load-bearing self-citations appear: the prior work cited for pruning and fairness ([31], [33], [40]–[42]) is external to the authors, and no uniqueness theorem is imported. The post-hoc selection of the best of six ViT patterns is a statistical-selection concern, but it does not make the central result equivalent to its inputs by construction: the chosen pattern's improvement over vanilla is still an empirical measurement on a fixed held-out split. Thus no step reduces to its own input.

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

The method introduces no new physical or mathematical entities. The main assumptions are domain-specific heuristics about what skewness means in skin lesion images, and the post hoc choice of pruning configuration. The free parameters listed are not fitted to data in the usual sense, but they are choices made during the study that affect the reported results.

free parameters (3)
  • Skewness threshold = 0
    The decision to prune components with median skewness less than or equal to zero is a hand-chosen threshold, not derived from theory or data. It is the key hyperparameter of the method.
  • Pruning pattern selection = Pattern 6 (Patch+Head with frozen embeddings)
    The paper tests six pruning configurations and reports the best performing one as the main result. This is a post hoc selection over configurations, effectively tuning the architecture on the test data.
  • Skin type grouping for evaluation = Fitzpatrick 1-3 vs 4-6
    The evaluation groups skin tones into Light and Dark categories, which is a modeling choice that affects all fairness metrics. The paper criticizes such groupings but uses them for evaluation.
assumptions (3)
  • domain assumption Skin lesions occupy a small fraction of the image, and skin occupies the majority, so positively skewed activations indicate lesion focus and negatively skewed activations indicate skin focus.
    This is the foundational assumption behind the pruning criterion, stated in Section III-A. It is not verified with lesion segmentation or attribution analysis.
  • domain assumption Channels, patches, or heads with negative median skewness are 'skin-tone-related' and removing them will reduce bias without affecting lesion-related features.
    The paper assumes that negative skewness is a proxy for skin-tone encoding. This is an empirical claim that is not directly tested; the paper only reports aggregate fairness metrics, not whether the pruned components are actually skin-tone-specific.
  • domain assumption Group fairness measured by EOpp0, EOpp1, and EOdd with a binary Light/Dark split is a meaningful evaluation of fairness.
    The paper adopts this framework from [31] despite earlier arguing that skin tone is continuous and that group definitions are problematic. The validity of the binary split is assumed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Enhancing Fairness in Skin Lesion Classification for Medical Diagnosis Using Prune Learning." pith.science (2026). https://pith.science/paper/K3DC5G3S

@misc{pith2026250900745,
  author       = {Pith},
  title        = {Pith review of: Enhancing Fairness in Skin Lesion Classification for Medical Diagnosis Using Prune Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/K3DC5G3S}},
  note         = {Machine review of arXiv:2509.00745}
}
read the original abstract

Recent advances in deep learning have significantly improved the accuracy of skin lesion classification models, supporting medical diagnoses and promoting equitable healthcare. However, concerns remain about potential biases related to skin color, which can impact diagnostic outcomes. Ensuring fairness is challenging due to difficulties in classifying skin tones, high computational demands, and the complexity of objectively verifying fairness. To address these challenges, we propose a fairness algorithm for skin lesion classification that overcomes the challenges associated with achieving diagnostic fairness across varying skin tones. By calculating the skewness of the feature map in the convolution layer of the VGG (Visual Geometry Group) network and the patches and the heads of the Vision Transformer, our method reduces unnecessary channels related to skin tone, focusing instead on the lesion area. This approach lowers computational costs and mitigates bias without relying on conventional statistical methods. It potentially reduces model size while maintaining fairness, making it more practical for real-world applications.

Figures

Figures reproduced from arXiv: 2509.00745 by the authors.

Figure 1
Figure 1. Pruning Process and Target for VGG max pooling layer, which identifies and retains only the most informative channels, those contributing meaningfully to lesion detection. Algorithm 1 CNN: Select Keep Channels Output: Channels to keep K Load model f(x) with trained weights Initialize Skew to store all filter skewness scores of all images for Each input x in Validation set X do Initialize feature map list M Compute f… view at source ↗
Figure 2
Figure 2. Pruning Process and Target for ViT-B16 pruning. In contrast, [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

57 extracted references · 50 canonical work pages

  1. [1]

    ViT-B16 processes input images by dividing them into 16×16 patches

    Patch Pruning: Our first pruning target is the convolution channels in the initial layers of ViT-B16. ViT-B16 processes input images by dividing them into 16×16 patches. Let the input image be x ∈ RH·W ·C, where H, W , and C denote height, width, and channels, respectively. Let P denote patch resolution, and N = H·W P 2 be the number of patches. Then, the...

    work page

  2. [2]

    The pruning target is the convolution layer immediately preceding the max pooling layer. When multiple convolution layers exist upstream of the max pooling layer, filters with matching indices across these layers are pruned simultane- ously, enabling a consistent structural reduction. Finally, since pruning alters the number of output channels in the feat...

    work page

  3. [3]

    Skew Prune (Patch) Y Y N Full

    work page

  4. [4]

    Head Pruning: While Patch Pruning removes input patches with strong skin color dependency, attention heads within the self-attention mechanism may still reflect such bias. This is because patches and attention heads have distinct roles: the former influences input representations, while the latter governs how relationships are encoded in the latent space....

    work page

  5. [5]

    Skew Prune (Patch) Y N N Full

    work page

  6. [6]

    • However, EOpp1 and EOdd showed significant group dif- ferences of approximately 7.4% and 8.2%, respectively, reflecting substantial fairness concerns

    Skew Prune (Patch+Head) Y N Y Partial TABLE I VIT: P RUNING PATTERNS Metrics 1.Vanilla 2.SCP-Fair Prune 3.SCP-Fair Prune (n=229) 4.SCP-Fair Prune Removed 5.SkewPrune Performance Accuracy 0.78 0.80 0.76 0.79 0.79 F1-score 0.66 0.67 0.65 0.66 0.65 Fairness EOpp0 0.0110 0.0051 0.0106 0.0068 0.0099 EOpp1 0.0666 0.0691 0.0530 0.0937 0.0511 EOdd 0.0776 0.0742 0...

    work page

  7. [7]

    Skew Prune (Head) N N Y Full

    work page

  8. [8]

    Skew Prune (Patch+Head) Y N Y Full

    work page

Show all 57 references
  1. [9]

    A differentiable distance approximation for fairer image classification,

    N. E Rosa, T. Drummond, and M. Harandi, “A differentiable distance approximation for fairer image classification,” in Proceedings of the Asian Conference on Computer Vision , 2022, pp. 212–228

    2022

  2. [10]

    Furthermore, the authors extend their heartfelt gratitude to Dr

    Skew Prune (P+H) Performance Accuracy 0.82 0.82 0.83 0.83 0.83 0.83 F1-score 0.73 0.72 0.73 0.75 0.74 0.73 Fairness EOpp0 0.0080 0.0110 0.0099 0.0073 0.0096 0.0086 EOpp1 0.0742 0.0678 0.0995 0.0645 0.0714 0.0624 Eodd 0.0822 0.0788 0.1095 0.0718 0.0810 0.0711 Computing Cost GFL...

  3. [11]

    Dermatologist-level classification of skin cancer with deep neural networks,

    A. Esteva, B. Kuprel, R. A. Novoa, J. Ko, S. M. Swetter, H. M. Blau, and S. Thrun, “Dermatologist-level classification of skin cancer with deep neural networks,” nature, vol. 542, no. 7639, pp. 115–118, 2017

    2017

  4. [12]

    Deep neural networks are superior to dermatologists in melanoma image classification,

    T. J. Brinker, A. Hekler, A. H. Enk, C. Berking, S. Haferkamp, A. Hauschild, M. Weichenthal, J. Klode, D. Schadendorf, T. Holland- Letz et al. , “Deep neural networks are superior to dermatologists in melanoma image classification,” European Journal of Cancer , vol. 119, pp. 1...

    2019

  5. [13]

    Ai-driven healthcare: Fairness in ai healthcare: A survey,

    S. V . Chinta, Z. Wang, A. Palikhe, X. Zhang, A. Kashif, M. A. Smith, J. Liu, and W. Zhang, “Ai-driven healthcare: Fairness in ai healthcare: A survey,” PLOS Digital Health , vol. 4, no. 5, p. e0000864, 2025

    2025

  6. [14]

    Fairness of artifi- cial intelligence in healthcare: review and recommendations,

    D. Ueda, T. Kakinuma, S. Fujita, K. Kamagata, Y . Fushimi, R. Ito, Y . Matsui, T. Nozaki, T. Nakaura, N. Fujima et al., “Fairness of artifi- cial intelligence in healthcare: review and recommendations,” Japanese Journal of Radiology , vol. 42, no. 1, pp. 3–15, 2024

    2024

  7. [15]

    Estimating skin tone and effects on classification performance in dermatology datasets,

    N. M. Kinyanjui, T. Odonga, C. Cintas, N. C. Codella, R. Panda, P. Sattigeri, and K. R. Varshney, “Estimating skin tone and effects on classification performance in dermatology datasets,” arXiv preprint arXiv:1910.13268, 2019

    1910 arXiv

  8. [16]

    Skin type diversity in skin lesion datasets: A review,

    N. Alipour, T. Burke, and J. Courtney, “Skin type diversity in skin lesion datasets: A review,” Current Dermatology Reports , vol. 13, no. 3, pp. 198–210, 2024

    2024

  9. [17]

    What about applied fairness?

    J. Sylvester and E. Raff, “What about applied fairness?” arXiv preprint arXiv:1806.05250, 2018

    2018 arXiv

  10. [18]

    Biasing & debiasing based approach towards fair knowledge transfer for equitable skin analysis,

    A. Pundhir, B. Raman, and P. Singh, “Biasing & debiasing based approach towards fair knowledge transfer for equitable skin analysis,” arXiv preprint arXiv:2405.10256 , 2024

    2024 arXiv

  11. [19]

    Algorithmic fairness in lesion classification by mitigating class imbalance and skin tone bias,

    F. Ansari, T. Chakraborti, and S. Das, “Algorithmic fairness in lesion classification by mitigating class imbalance and skin tone bias,” in International Conference on Medical Image Computing and Computer- Assisted Intervention. Springer, 2024, pp. 373–382

    2024

  12. [20]

    Fairdisco: Fairer ai in dermatology via disentanglement contrastive learning,

    S. Du, B. Hers, N. Bayasi, G. Hamarneh, and R. Garbi, “Fairdisco: Fairer ai in dermatology via disentanglement contrastive learning,” in European Conference on Computer Vision . Springer, 2022, pp. 185– 202

    2022

  13. [21]

    Skin deep: Investigating subjectivity in skin tone annotations for computer vision benchmark datasets,

    T. Barrett, Q. Chen, and A. Zhang, “Skin deep: Investigating subjectivity in skin tone annotations for computer vision benchmark datasets,” in Proceedings of the 2023 ACM Conference on Fairness, Accountability, and Transparency, 2023, pp. 1757–1771

    2023

  14. [22]

    What would the outputs be if the skin tone were lighter?

    can increase dataset diversity artificially, but they intro- duce their own biases and raise concerns about the representa- tiveness of augmented and generated data. These methods also increase implementation costs through longer training times and greater resource demands. Co...

  15. [23]

    Which skin tone measures are the most inclusive? an investigation of skin tone measures for artificial intelligence,

    C. M. Heldreth, E. P. Monk, A. T. Clark, C. Schumann, X. Eyee, and S. Ricco, “Which skin tone measures are the most inclusive? an investigation of skin tone measures for artificial intelligence,” ACM Journal on Responsible Computing , vol. 1, no. 1, pp. 1–21, 2024

    2024

  16. [24]

    The validity and practicality of sun-reactive skin types i through vi,

    T. B. Fitzpatrick, “The validity and practicality of sun-reactive skin types i through vi,” Archives of dermatology , vol. 124, no. 6, pp. 869–871, 1988

    1988

  17. [25]

    Developing the monk skin tone scale,

    M. Ellis, “Developing the monk skin tone scale,” 2022. [Online]. Available: https://skintone.google/

    2022

  18. [26]

    Expert in skin and hair types around the world,

    L’Or ´eal, “Expert in skin and hair types around the world,” Jul 2020. [Online]. Available: https://www.loreal.com/en/articles/ science-and-technology/expert-inskin/

    2020

  19. [27]

    Towards transparency in dermatology image datasets with skin tone annotations by experts, crowds, and an algorithm,

    M. Groh, C. Harris, R. Daneshjou, O. Badri, and A. Koochek, “Towards transparency in dermatology image datasets with skin tone annotations by experts, crowds, and an algorithm,” Proceedings of the ACM on Human-Computer Interaction, vol. 6, no. CSCW2, pp. 1–26, 2022

    2022

  20. [28]

    Understanding racial and ethnic differences in health in late life: A research agenda,

    R. A. Bulatao, N. B. Anderson et al., “Understanding racial and ethnic differences in health in late life: A research agenda,” 2004

    2004

  21. [29]

    A web-based mpox skin lesion detection system using state-of-the-art deep learning models considering racial diversity,

    S. N. Ali, M. T. Ahmed, T. Jahan, J. Paul, S. S. Sani, N. Noor, A. N. Asma, and T. Hasan, “A web-based mpox skin lesion detection system using state-of-the-art deep learning models considering racial diversity,” Biomedical Signal Processing and Control , vol. 98, p. 106742, 2024

    2024

  22. [30]

    Edgemixup: improving fairness for skin dis- ease classification and segmentation,

    H. Yuan, A. Hadzic, W. Paul, D. V . de Flores, P. Mathew, J. Aucott, Y . Cao, and P. Burlina, “Edgemixup: improving fairness for skin dis- ease classification and segmentation,” arXiv preprint arXiv:2202.13883, 2022

    2022 arXiv

  23. [31]

    Assessing bias in skin lesion classifiers with contemporary deep learning and post-hoc explainability techniques,

    A. Corbin and O. Marques, “Assessing bias in skin lesion classifiers with contemporary deep learning and post-hoc explainability techniques,” IEEE Access, 2023

    2023

  24. [32]

    Circle: Color invariant representation learning for unbiased classification of skin lesions,

    A. Pakzad, K. Abhishek, and G. Hamarneh, “Circle: Color invariant representation learning for unbiased classification of skin lesions,” in European Conference on Computer Vision . Springer, 2022, pp. 203– 219

    2022

  25. [33]

    Evaluating and mitigating bias in image classifiers: A causal perspective using coun- terfactuals,

    S. Dash, V . N. Balasubramanian, and A. Sharma, “Evaluating and mitigating bias in image classifiers: A causal perspective using coun- terfactuals,” in Proceedings of the IEEE/CVF winter conference on applications of computer vision , 2022, pp. 915–924

    2022

  26. [34]

    Detecting melanoma fairly: Skin tone detection and debiasing for skin lesion classification,

    P. J. Bevan and A. Atapour-Abarghouei, “Detecting melanoma fairly: Skin tone detection and debiasing for skin lesion classification,” in MICCAI Workshop on Domain Adaptation and Representation Transfer. Springer, 2022, pp. 1–11

    2022

  27. [35]

    Turning a blind eye: Explicit removal of biases and variation from deep neural network embeddings,

    M. Alvi, A. Zisserman, and C. Nell ˚aker, “Turning a blind eye: Explicit removal of biases and variation from deep neural network embeddings,” in Proceedings of the European conference on computer vision (ECCV) workshops, 2018, pp. 0–0

    2018

  28. [36]

    Learning not to learn: Training deep neural networks with biased data,

    B. Kim, H. Kim, K. Kim, S. Kim, and J. Kim, “Learning not to learn: Training deep neural networks with biased data,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 9012–9020

    2019

  29. [37]

    Estimating and improving fairness with adversarial learning,

    X. Li, Z. Cui, Y . Wu, L. Gu, and T. Harada, “Estimating and improving fairness with adversarial learning,” arXiv preprint arXiv:2103.04243 , 2021

    2021 arXiv

  30. [38]

    Exploit- ing transferable knowledge for fairness-aware image classification,

    S. Hwang, S. Park, P. Lee, S. Jeon, D. Kim, and H. Byun, “Exploit- ing transferable knowledge for fairness-aware image classification,” in Proceedings of the Asian Conference on Computer Vision , 2020

    2020

  31. [39]

    Achieving fairness in dermatological disease diagnosis through automatic weight adjusting federated learning and personalization,

    G. Xu, Y . Wu, J. Hu, and Y . Shi, “Achieving fairness in dermatological disease diagnosis through automatic weight adjusting federated learning and personalization,” arXiv preprint arXiv:2208.11187 , 2022

    2022 arXiv

  32. [40]

    Achieving flexible fairness metrics in federated medical imaging,

    H. Xing, R. Sun, J. Ren, J. Wei, C.-M. Feng, X. Ding, Z. Guo, Y . Wang, Y . Hu, W. Wei et al. , “Achieving flexible fairness metrics in federated medical imaging,” Nature Communications, vol. 16, no. 1, p. 3342, 2025

    2025

  33. [41]

    Fairprune: Achieving fairness through pruning for dermatological disease diagnosis,

    Y . Wu, D. Zeng, X. Xu, Y . Shi, and J. Hu, “Fairprune: Achieving fairness through pruning for dermatological disease diagnosis,” in International Conference on Medical Image Computing and Computer-Assisted Inter- vention. Springer, 2022, pp. 743–753. 10 IEEE TRANSACTIONS AND ...

    2022

  34. [42]

    Toward fairness through fair multi-exit framework for dermatological disease diagnosis,

    C.-H. Chiu, H.-W. Chung, Y .-J. Chen, Y . Shi, and T.-Y . Ho, “Toward fairness through fair multi-exit framework for dermatological disease diagnosis,” in International Conference on Medical Image Computing and Computer-Assisted Intervention . Springer, 2023, pp. 97–107

    2023

  35. [43]

    Achieving fairness through channel pruning for dermatological disease diagnosis,

    Q. Kong, C.-H. Chiu, D. Zeng, Y .-J. Chen, T.-Y . Ho, J. Hu, and Y . Shi, “Achieving fairness through channel pruning for dermatological disease diagnosis,” in International Conference on Medical Image Computing and Computer-Assisted Intervention . Springer, 2024, pp. 24–34

    2024

  36. [44]

    Fairquantize: Achieving fairness through weight quantization for dermatological disease diagnosis,

    Y . Guo, Z. Jia, J. Hu, and Y . Shi, “Fairquantize: Achieving fairness through weight quantization for dermatological disease diagnosis,” in International Conference on Medical Image Computing and Computer- Assisted Intervention. Springer, 2024, pp. 329–338

    2024

  37. [45]

    A fair loss function for network pruning,

    R. Meyer and A. Wong, “A fair loss function for network pruning,” arXiv preprint arXiv:2211.10285 , 2022

    2022 arXiv

  38. [46]

    Fairgrape: Fairness-aware gradient pruning method for face attribute classification,

    X. Lin, S. Kim, and J. Joo, “Fairgrape: Fairness-aware gradient pruning method for face attribute classification,” in European Conference on Computer Vision. Springer, 2022, pp. 414–432

    2022

  39. [47]

    Imagenet: A large-scale hierarchical image database,

    J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei, “Imagenet: A large-scale hierarchical image database,” in 2009 IEEE conference on computer vision and pattern recognition . Ieee, 2009, pp. 248–255

    2009

  40. [48]

    Domain generalization for medical image analysis: A review,

    J. S. Yoon, K. Oh, Y . Shin, M. A. Mazurowski, and H.-I. Suk, “Domain generalization for medical image analysis: A review,” Proceedings of the IEEE, 2024

    2024

  41. [49]

    Domain adaptation for medical image analysis: a survey,

    H. Guan and M. Liu, “Domain adaptation for medical image analysis: a survey,” IEEE Transactions on Biomedical Engineering , vol. 69, no. 3, pp. 1173–1185, 2021

    2021

  42. [50]

    Not all patches are what you need: Expediting vision transformers via token reorganizations,

    Y . Liang, C. Ge, Z. Tong, Y . Song, J. Wang, and P. Xie, “Not all patches are what you need: Expediting vision transformers via token reorganizations,” arXiv preprint arXiv:2202.07800 , 2022

    2022 arXiv

  43. [51]

    Intriguing properties of vision transformers,

    M. M. Naseer, K. Ranasinghe, S. H. Khan, M. Hayat, F. Shahbaz Khan, and M.-H. Yang, “Intriguing properties of vision transformers,” Ad- vances in Neural Information Processing Systems , vol. 34, pp. 23 296– 23 308, 2021

    2021

  44. [52]

    Analyzing multi-head self-attention: Specialized heads do the heavy lifting, the rest can be pruned,

    E. V oita, D. Talbot, F. Moiseev, R. Sennrich, and I. Titov, “Analyzing multi-head self-attention: Specialized heads do the heavy lifting, the rest can be pruned,” arXiv preprint arXiv:1905.09418 , 2019

    1905 arXiv

  45. [53]

    An image is worth 16x16 words: Transformers for image recognition at scale,

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly et al., “An image is worth 16x16 words: Transformers for image recognition at scale,” arXiv preprint arXiv:2010.11929 , 2020

    2010 arXiv

  46. [54]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,” Advances in neural information processing systems , vol. 30, 2017

    2017

  47. [55]

    The ham10000 dataset, a large collection of multi-source dermatoscopic images of common pigmented skin lesions,

    P. Tschandl, C. Rosendahl, and H. Kittler, “The ham10000 dataset, a large collection of multi-source dermatoscopic images of common pigmented skin lesions,” Scientific data, vol. 5, no. 1, pp. 1–9, 2018

    2018

  48. [56]

    N. C. Codella, D. Gutman, M. E. Celebi, B. Helba, M. A. Marchetti, S. W. Dusza, A. Kalloo, K. Liopyris, N. Mishra, H. Kittler et al. , “Skin lesion analysis toward melanoma detection: A challenge at the 2017 international symposium on biomedical imaging (isbi), hosted by the i...

    2017

  49. [57]

    Bcn20000: Dermoscopic lesions in the wild,

    C. Hern ´andez-P´erez, M. Combalia, S. Podlipnik, N. C. Codella, V . Rotemberg, A. C. Halpern, O. Reiter, C. Carrera, A. Barreiro, B. Helba et al., “Bcn20000: Dermoscopic lesions in the wild,” Scientific data, vol. 11, no. 1, p. 641, 2024. Kuniko Paxton is a PhD candidate in C...

    2024

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.