Pith. sign in

REVIEW 3 major objections 5 minor 55 references

How Far from Clinical Deployment? Evaluating the Complete Unsupervised Domain Adaptation Pipeline in Medical Imaging

T0 review · 3 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read This paper claims that the bottleneck for clinically deployable unsupervised domain adaptation is not adaptation itself but label-free model selection: a capable checkpoint usually exists, yet 13 evaluated validators leave a large…

desk verdict A solid, large medical-imaging benchmark for UDA model selection, but the 6.1-point 'structural' gap is inflated by winner's curse and needs a bias-corrected oracle. read the letter →

arxiv 2608.12035 v1 pith:7PWDXU2B submitted 2026-08-12 cs.CV cs.AI

Yiheng Xiong , Luisa Gallée , Daniel Santak Wolf , Heiko Hillenhagen , Michael Götz This is my paper · ORCID
classification cs.CVcs.AI
keywords unsuperviseddomainadaptationlabel-freemodelselectionmedicalimagingshiftclinicaldeploymentcheckpointbenchmarkevaluation
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

Unsupervised domain adaptation (UDA) trains models on a labeled source domain and an unlabeled target domain, but a clinical deployment also has to choose which of the many trained checkpoints to ship, without target labels to test them on. This paper evaluates that complete pipeline together across eleven clinically relevant medical imaging transfer scenarios, ten UDA algorithms, thirteen label-free selection methods, and over 80,000 checkpoints. The central finding is that a capable adapted model usually exists in the pool, but every evaluated label-free selector leaves a large gap to the best available model: on average 6.1 percentage points, up to 10.5. The paper concludes that the remaining distance to clinical deployment lies mostly in the selection step, not in adaptation, and that ensembling or a small target-labeling budget narrows but does not close the gap.

What carries the argument

The central object is the 'validator,' a label-free scoring function $V(\theta) \mapsto s \in \mathbb{R}$ that assigns a scalar score to each checkpoint $\theta$ without using target labels; the deployed model is $\theta^\star = \arg\max_{\theta\in\Theta} V(\theta)$, with $\Theta$ the pool of checkpoints across algorithms, hyperparameters, and training iterations. Comparing a validator's choice with the oracle choice defines the selection gap. To test whether the gap is a bad choice among validators or a structural limitation, the paper uses the Spearman rank correlation between validation scores and true target accuracy, measured both within single algorithms and on an across-algorithm pool formed by taking each algorithm's top-10 checkpoints by validator score. The combination of the selection gap and the Spearman diagnostic is what turns the observation that 'validators sometimes fail' into a claim about structural unreliability.

What would settle it

On the same checkpoint pools, compute across-algorithm Spearman correlations pooling all checkpoints rather than the top-10 per algorithm, or test a new label-free score on all 11 scenarios: if any validator shows consistently positive correlation (say $\rho>0.5$) across most scenarios, or selects a checkpoint within 2 percentage points of the oracle across most scenarios, the structural-gap claim is refuted.

Watch

Extended reading notes

Core claim

Under oracle selection (choosing with target labels), the across-algorithm checkpoint pool beats source-only in every scenario and approaches target-supervised training, so adaptation itself usually works. The failure is label-free model selection: under the best of the 13 validators, selected models average 6.1 percentage points below the oracle-best model (maximum 10.5), and the same gap appears across four different backbone architectures. The gap is structural: for within-algorithm selection a validator such as BNM (V) can be highly reliable on one algorithm ($\rho=0.88$) and reversed on another ($\rho=-0.42$), while for across-algorithm selection most validator correlations are weak or negative. Ensembling validator-selected checkpoints, and spending a small labeled target budget (about 20 samples for brain MRI, 4–5% for other modalities), both reduce the gap, but both remain below the oracle. The conclusions are drawn for binary classification measured by balanced accuracy.

Load-bearing premise

The structural-gap claim assumes the 13 label-free selection methods are representative of such methods and that the top-10-per-algorithm pooling protocol is a fair test of ranking; if an untested selector is consistently reliable, or the pooling rule artificially lowers correlations, the gap would not be structural, and the finding is also scoped to binary classification with balanced accuracy.

Editorial extensions

If this is right

  • Deployable medical UDA is gated by model selection: even when a strong adapted checkpoint exists, a practitioner using any of the 13 evaluated validators will, on average, leave 6.1 points of balanced accuracy unrealized.
  • No single validator can be pre-chosen as a default for clinical use; reliability flips with algorithm and scenario, so the same validator that ranks one algorithm's checkpoints well can rank another's in reverse.
  • Across-algorithm selection is harder than within-algorithm selection: pooling checkpoints from all algorithms makes most validator correlations weak or negative, so choosing which UDA algorithm to deploy is an unresolved sub-problem.
  • Ensembling the checkpoints selected by all validators across algorithms recovers much of the gap in several scenarios and can even match the best hindsight validator, but it is computationally costly and can hurt when poor checkpoints are naively averaged.
  • A small target-labeling budget (about 20 samples for brain MRI, 4–5% for other scenarios) selects models more accurately than the best validator or ensembling, but still leaves a gap to the oracle.

Reading between the lines

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

  • Extension: the across-algorithm collapse of validator correlations suggests that validators should be designed for cross-algorithm comparability rather than tuned within one algorithm; absolute score calibration across heterogeneous checkpoints may be the missing ingredient.
  • Extension: the budget results imply a hybrid deployment recipe the paper does not push: use a label-free validator to shortlist checkpoints, then spend a small number of target labels to choose among the shortlist, which combines the two strategies and could reach the oracle with fewer than 25 labels.
  • Extension: if the structural gap persists for segmentation or detection, where prediction structure is richer, label-free selection difficulty could grow; the paper itself lists this as future work, and the likely direction is that the gap widens.
  • Extension: a testable consequence is that validator disagreements could be used as an uncertainty signal: checkpoints that many validators rank highly are more likely to be near the oracle, which would motivate studying consensus across validator scores rather than any single score.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper evaluates the complete unsupervised domain adaptation (UDA) pipeline for medical imaging classification, treating both adaptation and label-free model selection as parts of one deployment problem. Across eleven cross-domain scenarios (brain MRI, chest X-ray, retina), ten UDA algorithms, thirteen label-free validators, and over 80,000 trained checkpoints, it reports that a capable adapted model usually exists under oracle selection, but the models selected by label-free validators fall substantially short of the best available model, with an average accuracy gap of 6.1 points and a maximum of 10.5 points. The authors attribute this gap to a structural limitation of current validators, supported by Spearman correlations showing no validator is consistently reliable within or across algorithms. They further show that ensembling validator-selected checkpoints or spending a small target-labeling budget narrows but does not close the gap.

Significance. The paper addresses a practically important and previously understudied problem: model selection in the UDA deployment pipeline. It provides a large-scale, multi-dataset benchmark that includes medical-specific UDA algorithms and a broad set of validators, and it is one of the first studies to evaluate selection across algorithms rather than only within a single algorithm. The paper ships open-source code and reports both mean±std and median with 95% confidence intervals, which strengthens the reliability of its descriptive statistics. If the central claim survives the methodological concerns below, the paper would be a valuable reference for the medical imaging and UDA communities, highlighting that adaptation accuracy does not translate into deployable performance unless reliable label-free selection is solved.

major comments (3)
  1. [§4.2, Figure 3] The headline selection gap of 6.1 points is computed by selecting the Oracle checkpoint as the argmax over roughly 1,500 checkpoints on the same small target validation set (around 100 samples for brain MRI, per §4.4) and then reporting that same validation accuracy as the Oracle's performance. Because validator-selected models are not chosen using target labels, their reported accuracies are unbiased estimates, while the Oracle accuracy is the maximum of many noisy estimates and is therefore positively biased by winner's curse. The gap thus mixes genuine validator deficiency with the cost of estimating the best model from a finite sample, so the 'large and structural' claim in §4.2 and the abstract is not a sound estimate of the shortfall attributable to label-free selection. I recommend re-estimating the gap with a nested oracle (e.g., select on one half of the target validation set and evaluate on the other half) or a bias-corrected oracle, and reporting both the raw and corrected gaps.
  2. [§4.3, Figure 6(b) and Figure 7] The across-algorithm Spearman correlations are computed after restricting each algorithm's checkpoints to the validator's top-10 scores before pooling. This range restriction attenuates correlation coefficients, so the observed collapse to near zero may be a procedural artifact rather than evidence that validators are structurally unreliable for across-algorithm selection. The paper's claim that 'the gap therefore appears largely structural' relies on these correlations. I request a re-analysis using the full checkpoint pool (or a random subsample), or a formal justification for why top-10 pooling is the correct protocol for the deployment setting. Without this, the 'structural' conclusion is not established for the across-algorithm case.
  3. [§4.3, Figures 5 and 7] The conclusion that no evaluated validator is consistently reliable is based on only 13 validators, all from prior general-vision or shallow-adaptation literature. The paper states this gap is 'largely structural,' which is a strong claim about the impossibility of reliable label-free selection given the current validator paradigm, but a sample of 13 validators is limited evidence for such structurality. I recommend softening the language to 'no validator among those evaluated' or adding an explicit representative-coverage argument for the validator set.
minor comments (5)
  1. [§4.3, Figure 6(b)] The p-value is printed as 'p = 0.0.678' in the figure caption; this should be 'p = 0.678'.
  2. [Tables 5–37] The supplementary tables are not referenced individually in the main text; adding a sentence in §6.1 pointing to the corresponding main-text figures or claims would improve navigability.
  3. [Figure 5] The heatmap colorbar in Figure 5 ranges only from 0 to 1.0, but several cells contain negative correlations (e.g., -0.42). The colorbar should include the full range of the displayed values, or the text should clarify how negative values are mapped.
  4. [§4.4, Table 4] For the small labeling budget, the paper does not specify explicitly whether the checkpoint selected on the labeled subset is evaluated on the full target validation set or on the same labeled subset. If the latter, the reported gains are inflated; please clarify.
  5. [§3, Implementation Details] The adaptation strength is varied over λ∈{0.1,0.5,1.0}, but there is no discussion of whether this hyperparameter range is sufficient for the algorithms to show their typical behavior; a sentence justifying the range would be helpful.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper measures an empirical selection gap against an explicitly label-dependent oracle and never presents a derived quantity as predicted from its own inputs.

full rationale

The paper is an empirical evaluation, not a derivation, and its central claims are measurements over external prior-work methods. The Oracle is explicitly defined as the argmax of target validation accuracy (Section 4, 'Oracle selects using target labels and is the actual best model'), so reporting its accuracy as an upper bound is a definitional reference point, not a fitted input renamed as a prediction. The selection gap is then a comparison between this labeled oracle and label-free validators, and the paper repeatedly notes that Best Validator and Best Pair are themselves identified using target labels and 'cannot be chosen in advance,' which undercuts any hidden equivalence. The Spearman-based reliability analysis uses held-out target labels as ground truth for ranking quality, which is a measurement rather than a circular construction. No load-bearing self-citation or imported uniqueness theorem appears: the validators and algorithms are cited from prior literature, and no central premise is justified solely by the present authors' earlier work. The acknowledged limitations (binary classification, balanced accuracy, small labeling-budget annotation cost) are scope statements, not circular steps. The winner's-curse concern about the oracle upper bound is a statistical-bias critique of the benchmark design, not an identity between the paper's inputs and outputs, and would need separate bias-correction rather than a circularity finding.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

The paper is an empirical benchmark. It introduces no new entities and fits no constants. Its claims rest on the representativeness of the chosen scenarios, algorithms, validators, and the balanced-accuracy metric, plus the hand-chosen hyperparameter grid that defines the model pool.

free parameters (1)
  • Hyperparameter grid (lambda in {0.1, 0.5, 1.0}; 50 checkpoints per run; 10k or 30k iterations)
    Chosen by hand to define the candidate checkpoint pool. The existence of a capable adapted model and the measured selection gap are conditional on this grid; no target labels were used to set these values.
assumptions (4)
  • domain assumption Balanced accuracy on the target validation set is the deployment-relevant performance metric.
    Used throughout Sections 3 and 4 to measure target performance; other clinically relevant metrics, such as sensitivity and specificity, are not evaluated, as acknowledged in Section 4.5.
  • domain assumption The 11 scenarios and 10 algorithms are representative of clinical UDA shifts.
    Covers three organs and within-modality and across-modality shifts, but the generality of the structural conclusion rests on this representativeness.
  • domain assumption The 13 validators are implemented faithfully and capture the space of label-free selection methods.
    The conclusion that no validator is consistently reliable is based on these 13; better validators could exist outside the evaluated set.
  • standard math Oracle selection with target labels is a valid upper bound for deployment selection.
    By definition, the oracle maximizes target accuracy; it is used as the reference point for defining the selection gap.

how reviews work

0 comments
Cite this review

Pith. "Pith review of How Far from Clinical Deployment? Evaluating the Complete Unsupervised Domain Adaptation Pipeline in Medical Imaging." pith.science (2026). https://pith.science/paper/7PWDXU2B

@misc{pith2026260812035,
  author       = {Pith},
  title        = {Pith review of: How Far from Clinical Deployment? Evaluating the Complete Unsupervised Domain Adaptation Pipeline in Medical Imaging},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7PWDXU2B}},
  note         = {Machine review of arXiv:2608.12035}
}
read the original abstract

Deploying unsupervised domain adaptation (UDA) in clinical practice requires choosing which algorithm to use and which of its trained models to ship. However, the deployment (target) domain is unlabeled, so models cannot be evaluated directly on it, leaving it unclear which to select. We address this by evaluating the complete UDA pipeline, considering both adaptation and label-free selection together. Our study covers eleven clinically relevant cross-domain scenarios from nine medical imaging datasets, with ten UDA algorithms and 13 label-free selection methods (validators), evaluating over 80,000 trained models in total. By this, we find that a capable adapted model usually exists, but identifying it without target labels is difficult: the validator-selected models leave a large and structural target performance gap to the best available one, with no evaluated validator consistently reliable. Towards closing it, we explore two strategies, ensembling and a small target-labeling budget; both narrow this gap but do not close it entirely. Overall, deployable UDA depends on the complete pipeline; addressing the less explored selection step could bring much of current UDA closer to clinical use.

Figures

Figures reproduced from arXiv: 2608.12035 by the authors.

Figure 1
Figure 1. The complete UDA pipeline at deployment. Adapta [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Target accuracy under Oracle selection across UDA scenarios. Bars show SourceOnly (no adaptation), the per-algorithm av￾erage (Avg.), Across-Algo (selection over all algorithms’ pooled checkpoints), and TargetOnly (labeled-target). Across-Algo exceeds SourceOnly in all scenarios and approaches TargetOnly, indicating a capable adapted model usually exists. Error bars: standard deviation. ADNI-1 2 ADNI-1 3 ADNI-2 1 AD… view at source ↗
Figure 3
Figure 3. Target accuracy across UDA scenarios under three selections (%). Bars show the accuracy of the model selected by the best [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Within-algorithm Spearman correlation (ρ) between val￾idation score and target accuracy. Each point is a checkpoint. The same validator can flip from reliable to reversed with a change of algorithm (a, b) or scenario (c, d) alone. in the designed direction (ρ = 0.88, p…
Figure 5
Figure 5. Figure 5: Within-algorithm Spearman correlation (ρ) between validation scores and target accuracy. Blue denotes correlation in the intended direction (darker is stronger), warm red the opposite. No validator is consistently reliable across all scenarios or algorithms, and the mo…
Figure 6
Figure 6. Figure 6: Spearman correlation (ρ) between Source-Risk scores and target accuracy on ADNI-1→ADNI-2. Each point is a check￾point. (a) Within BNM algorithm, the correlation is moderate. (b) Across all algorithms, it collapses to near zero. 4.5. Discussion Our results show that the…
Figure 8
Figure 8. Figure 8: Target accuracy under Oracle selection across UDA scenarios, reported as the median with 95% confidence intervals. Bars show SourceOnly (no adaptation), the per-algorithm average (Avg.), Across-Algo (selection over all algorithms’ pooled checkpoints), and TargetOnly (l…
Figure 9
Figure 9. Figure 9: Target accuracy across UDA scenarios under three selections (%), reported as the median with 95% confidence intervals. Bars [PITH_FULL_IMAGE:figures/full_fig_p012_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

55 extracted references · 49 canonical work pages

  1. [1]

    Covid-19 radiography database.https : / / www . kaggle . com / datasets / tawsifurrahman / covid19-radiography-database. 3

    work page

  2. [2]

    Lungs disease dataset (4 types).https : / / www . kaggle . com / datasets / omkarmanohardalvi / lungs-disease-dataset-4-types. 3

    work page

  3. [3]

    Sireesha Chamarthi, Katharina Fogelberg, Titus J Brinker, et al. Mitigating the influence of domain shift in skin lesion classification: A benchmark study of unsupervised domain adaptation methods.Informatics in Medicine Unlocked, 44: 101430, 2024. 1, 2

    work page 2024

  4. [4]

    Reusing the task-specific classifier as a discriminator: Discriminator-free adversarial domain adaptation

    Lin Chen, Huaian Chen, Zhixiang Wei, et al. Reusing the task-specific classifier as a discriminator: Discriminator-free adversarial domain adaptation. InCVPR, pages 7181–7190,

    work page

  5. [5]

    Advancing medical imaging informatics by deep learning-based domain adaptation.Yearbook of medical informatics, 29(01):129– 138, 2020

    Anirudh Choudhary, Li Tong, Yuanda Zhu, et al. Advancing medical imaging informatics by deep learning-based domain adaptation.Yearbook of medical informatics, 29(01):129– 138, 2020. 1, 2

    work page 2020

  6. [6]

    Towards discriminability and diversity: Batch nuclear-norm maxi- mization under label insufficient situations

    Shuhao Cui, Shuhui Wang, Junbao Zhuo, et al. Towards discriminability and diversity: Batch nuclear-norm maxi- mization under label insufficient situations. InCVPR, pages 3941–3950, 2020. 3

    work page 2020

  7. [7]

    Imagenet: A large-scale hierarchical image database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale hierarchical image database. In2009 IEEE conference on computer vision and pattern recognition, pages 248–255. Ieee, 2009. 3

    2009

  8. [8]

    Cross- moda 2021 challenge: Benchmark of cross-modality do- main adaptation techniques for vestibular schwannoma and cochlea segmentation.Medical Image Analysis, 83:102628,

    Reuben Dorent, Aaron Kujawa, Marina Ivory, Spyridon Bakas, Nicola Rieke, Samuel Joutard, Ben Glocker, Jorge Cardoso, Marc Modat, Kayhan Batmanghelich, et al. Cross- moda 2021 challenge: Benchmark of cross-modality do- main adaptation techniques for vestibular schwannoma and cochlea segmentation.Medical Image Analysis, 83:102628,

    work page 2021

Show all 55 references
  1. [9]

    Kathryn A Ellis, Ashley I Bush, David Darby, et al. The aus- tralian imaging, biomarkers and lifestyle (aibl) study of ag- ing: methodology and baseline characteristics of 1112 indi- viduals recruited for a longitudinal study of alzheimer’s dis- ease.International psychogeriat...

    2009

  2. [10]

    Better prac- tices for domain adaptation

    Linus Ericsson, Da Li, and Timothy Hospedales. Better prac- tices for domain adaptation. InInternational Conference on Automated Machine Learning, pages 4–1. PMLR, 2023. 2

    2023

  3. [11]

    Contrastive domain adaptation with consistency match for automated pneumonia diagnosis.Medical Image Analysis, 83:102664, 2023

    Yangqin Feng, Zizhou Wang, Xinxing Xu, Yan Wang, Huazhu Fu, Shaohua Li, Liangli Zhen, Xiaofeng Lei, Ying- nan Cui, Jordan Sim Zheng Ting, et al. Contrastive domain adaptation with consistency match for automated pneumonia diagnosis.Medical Image Analysis, 83:102664, 2023. 1, 3

    2023

  4. [12]

    Unsupervised domain adaptation by backpropagation

    Yaroslav Ganin and Victor Lempitsky. Unsupervised domain adaptation by backpropagation. InICML, pages 1180–1189. PMLR, 2015. 2, 3

    2015

  5. [13]

    Domain-adversarial training of neural networks.Journal of machine learning research, 17(59):1–35, 2016

    Yaroslav Ganin, Evgeniya Ustinova, Hana Ajakan, et al. Domain-adversarial training of neural networks.Journal of machine learning research, 17(59):1–35, 2016. 3

    2016

  6. [14]

    Domain adaptation for medical image analysis: a survey.IEEE Transactions on Biomedical Engineering, 69(3):1173–1185, 2021

    Hao Guan and Mingxia Liu. Domain adaptation for medical image analysis: a survey.IEEE Transactions on Biomedical Engineering, 69(3):1173–1185, 2021. 1, 2

    2021

  7. [15]

    Multi-site mri har- monization via attention-guided deep domain adaptation for brain disorder identification.Medical image analysis, 71: 102076, 2021

    Hao Guan, Yunbi Liu, Erkun Yang, et al. Multi-site mri har- monization via attention-guided deep domain adaptation for brain disorder identification.Medical image analysis, 71: 102076, 2021. 1, 3

    2021

  8. [16]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, et al. Deep residual learning for image recognition. InCVPR, pages 770–778, 2016. 3

    2016

  9. [17]

    Mixed samples as probes for un- supervised model selection in domain adaptation.Advances in Neural Information Processing Systems, 36:37923–37941,

    Dapeng Hu, Jian Liang, Jun Hao Liew, Chuhui Xue, Song Bai, and Xinchao Wang. Mixed samples as probes for un- supervised model selection in domain adaptation.Advances in Neural Information Processing Systems, 36:37923–37941,

  10. [18]

    Towards reli- able model selection for unsupervised domain adaptation: An empirical study and a certified baseline.NeurIPS, 37: 135883–135903, 2024

    Dapeng Hu, Romy Luo, Jian Liang, et al. Towards reli- able model selection for unsupervised domain adaptation: An empirical study and a certified baseline.NeurIPS, 37: 135883–135903, 2024. 2, 6

    2024

  11. [19]

    Clifford R Jack Jr, Matt A Bernstein, Nick C Fox, et al. The alzheimer’s disease neuroimaging initiative (adni): Mri methods.Journal of Magnetic Resonance Imaging: An Offi- cial Journal of the International Society for Magnetic Reso- nance in Medicine, 27(4):685–691, 2008. 3

    2008

  12. [20]

    Minimum class confusion for versatile domain adaptation

    Ying Jin, Ximei Wang, Mingsheng Long, et al. Minimum class confusion for versatile domain adaptation. InECCV, pages 464–480. Springer, 2020. 3

    2020

  13. [21]

    Consensus-driven active model selection

    Justin Kay, Grant Van Horn, Subhransu Maji, et al. Consensus-driven active model selection. InICCV, pages 4594–4604, 2025. 8

    2025

  14. [22]

    Identifying medical diagnoses and treatable diseases by image-based deep learning.cell, 172(5):1122–1131, 2018

    Daniel S Kermany, Michael Goldbaum, Wenjia Cai, et al. Identifying medical diagnoses and treatable diseases by image-based deep learning.cell, 172(5):1122–1131, 2018. 3

    2018

  15. [23]

    Deep learning for un- supervised domain adaptation in medical imaging: Recent advancements and future perspectives.Computers in Biol- ogy and Medicine, 170:107912, 2024

    Suruchi Kumari and Pravendra Singh. Deep learning for un- supervised domain adaptation in medical imaging: Recent advancements and future perspectives.Computers in Biol- ogy and Medicine, 170:107912, 2024. 1, 2

    2024

  16. [24]

    Skada-bench: Benchmarking unsupervised domain adapta- tion methods with realistic validation on diverse modalities

    Yanis Lalou, Th ´eo Gnassounou, Antoine Collas, et al. Skada-bench: Benchmarking unsupervised domain adapta- tion methods with realistic validation on diverse modalities. arXiv preprint arXiv:2407.11676, 2024. 2

    2024

  17. [25]

    Medical image segmenta- tion with domain adaptation: A survey.arXiv preprint arXiv:2311.01702, 2023

    Yuemeng Li and Yong Fan. Medical image segmenta- tion with domain adaptation: A survey.arXiv preprint arXiv:2311.01702, 2023. 1, 2

    2023 arXiv

  18. [26]

    Domain adaptation with auxiliary target domain-oriented classifier

    Jian Liang, Dapeng Hu, and Jiashi Feng. Domain adaptation with auxiliary target domain-oriented classifier. InCVPR, pages 16632–16642, 2021. 3

    2021

  19. [27]

    Hasd: hierarchical adaption for pathol- ogy slide-level domain-shift

    Jingsong Liu, Han Li, Chen Yang, Michael Deutges, Ario Sadafi, Xin You, Katharina Breininger, Nassir Navab, and Peter J Sch ¨uffler. Hasd: hierarchical adaption for pathol- ogy slide-level domain-shift. InInternational Conference on Medical Image Computing and Computer-Assiste...

    2025

  20. [28]

    Attention-guided partial domain adaptation for automated pneumonia diagnosis from chest x-ray images.IEEE Journal of Biomedical and Health Informatics, 27(12):5848–5859,

    Wentao Liu, Zhiwei Ni, Qian Chen, and Liping Ni. Attention-guided partial domain adaptation for automated pneumonia diagnosis from chest x-ray images.IEEE Journal of Biomedical and Health Informatics, 27(12):5848–5859,

  21. [29]

    A convnet for the 2020s.CVPR, 2022

    Zhuang Liu, Hanzi Mao, Chao-Yuan Wu, et al. A convnet for the 2020s.CVPR, 2022. 5

    2022

  22. [30]

    Learn- ing transferable features with deep adaptation networks

    Mingsheng Long, Yue Cao, Jianmin Wang, et al. Learn- ing transferable features with deep adaptation networks. In ICML, pages 97–105. PMLR, 2015. 3

    2015

  23. [31]

    Con- ditional adversarial domain adaptation.NeurIPS, 31, 2018

    Mingsheng Long, Zhangjie Cao, Jianmin Wang, et al. Con- ditional adversarial domain adaptation.NeurIPS, 31, 2018. 3

    2018

  24. [32]

    Active model selection: A variance minimization approach

    Mitsuru Matsuura and Satoshi Hara. Active model selection: A variance minimization approach. InNeurIPS 2023 Work- shop on Adaptive Experimental Design and Active Learning in the Real World, 2023. 8

    2023

  25. [33]

    Minimal-entropy correlation alignment for unsupervised deep domain adaptation.arXiv preprint arXiv:1711.10288,

    Pietro Morerio, Jacopo Cavazza, and Vittorio Murino. Minimal-entropy correlation alignment for unsupervised deep domain adaptation.arXiv preprint arXiv:1711.10288,

    arXiv

  26. [34]

    Unsu- pervised domain adaptation: A reality check.arXiv preprint arXiv:2111.15672, 2021

    Kevin Musgrave, Serge Belongie, and Ser-Nam Lim. Unsu- pervised domain adaptation: A reality check.arXiv preprint arXiv:2111.15672, 2021. 1, 2, 3

    2021 arXiv

  27. [35]

    Three new validators and a large-scale benchmark rank- ing for unsupervised domain adaptation.arXiv preprint arXiv:2208.07360, 2022

    Kevin Musgrave, Serge Belongie, and Ser-Nam Lim. Three new validators and a large-scale benchmark rank- ing for unsupervised domain adaptation.arXiv preprint arXiv:2208.07360, 2022. 2, 3

    2022 arXiv

  28. [36]

    M3-uda: A new benchmark for unsupervised domain adaptive fetal cardiac structure detection

    Bin Pu, Liwen Wang, Jiewen Yang, et al. M3-uda: A new benchmark for unsupervised domain adaptive fetal cardiac structure detection. InCVPR, pages 11621–11630, 2024. 1, 2

    2024

  29. [37]

    Maximum classifier discrepancy for unsupervised domain adaptation

    Kuniaki Saito, Kohei Watanabe, Yoshitaka Ushiku, et al. Maximum classifier discrepancy for unsupervised domain adaptation. InCVPR, pages 3723–3732, 2018. 3

    2018

  30. [38]

    Tune it the right way: Unsupervised validation of domain adaptation via soft neighborhood density

    Kuniaki Saito, Donghyun Kim, Piotr Teterwak, et al. Tune it the right way: Unsupervised validation of domain adaptation via soft neighborhood density. InICCV, pages 9184–9193,

  31. [39]

    Domain adaptation and generalization on functional medical images: A systematic survey.arXiv preprint arXiv:2212.03176, 2022

    Gita Sarafraz, Armin Behnamnia, Mehran Hosseinzadeh, et al. Domain adaptation and generalization on functional medical images: A systematic survey.arXiv preprint arXiv:2212.03176, 2022. 1, 2

    2022 arXiv

  32. [40]

    Active comparison of prediction models.Advances in neural information processing systems, 25, 2012

    Christoph Sawade, Niels Landwehr, and Tobias Scheffer. Active comparison of prediction models.Advances in neural information processing systems, 25, 2012. 8

    2012

  33. [41]

    M3da: Benchmark for unsupervised domain adap- tation in 3d medical image segmentation.arXiv preprint arXiv:2502.17029, 2025

    Boris Shirokikh, Anvar Kurmukov, Mariia Donskova, et al. M3da: Benchmark for unsupervised domain adap- tation in 3d medical image segmentation.arXiv preprint arXiv:2502.17029, 2025. 1, 2

    2025 arXiv

  34. [42]

    Navigating dis- tribution shifts in medical image analysis: A survey.arXiv preprint arXiv:2411.05824, 2024

    Zixian Su, Jingwei Guo, Xi Yang, Qiufeng Wang, Frans Co- enen, Amir Hussain, and Kaizhu Huang. Navigating dis- tribution shifts in medical image analysis: A survey.arXiv preprint arXiv:2411.05824, 2024. 1, 2

    2024 arXiv

  35. [43]

    Covariate shift adaptation by importance weighted cross validation.JMLR, 8(5), 2007

    Masashi Sugiyama, Matthias Krauledat, and Klaus-Robert M¨uller. Covariate shift adaptation by importance weighted cross validation.JMLR, 8(5), 2007. 1, 2, 3

    2007

  36. [44]

    Domain adap- tation for skin lesion: Evaluating real-world generalisation

    Nurjahan Sultana, Wenqi Lu, Xinqi Fan, et al. Domain adap- tation for skin lesion: Evaluating real-world generalisation. InCVPRW, pages 3433–3443, 2025. 1, 2

    2025

  37. [45]

    Fairdomain: Achieving fairness in cross-domain medical image segmentation and classifica- tion

    Yu Tian, Congcong Wen, Min Shi, Muhammad Muneeb Afzal, Hao Huang, Muhammad Osama Khan, Yan Luo, Yi Fang, and Mengyu Wang. Fairdomain: Achieving fairness in cross-domain medical image segmentation and classifica- tion. InEuropean Conference on Computer Vision, pages 251–271. Sp...

    2024

  38. [46]

    Training data-efficient image transformers & distillation through at- tention

    Hugo Touvron, Matthieu Cord, Matthijs Douze, Francisco Massa, Alexandre Sablayrolles, and Herv ´e J´egou. Training data-efficient image transformers & distillation through at- tention. InInternational conference on machine learning, pages 10347–10357. PMLR, 2021. 5

    2021

  39. [47]

    Resmlp: Feedforward networks for image classification with data-efficient training.IEEE TPAMI, 45(4):5314–5321,

    Hugo Touvron, Piotr Bojanowski, Mathilde Caron, et al. Resmlp: Feedforward networks for image classification with data-efficient training.IEEE TPAMI, 45(4):5314–5321,

  40. [48]

    Assessing model out-of-distribution generalization with softmax pre- diction probability baselines and a correlation method

    Weijie Tu, Weijian Deng, Tom Gedeon, et al. Assessing model out-of-distribution generalization with softmax pre- diction probability baselines and a correlation method. 2, 3

  41. [49]

    Chestx-ray8: Hospital-scale chest x-ray database and benchmarks on weakly-supervised classification and localization of common thorax diseases

    Xiaosong Wang, Yifan Peng, Le Lu, et al. Chestx-ray8: Hospital-scale chest x-ray database and benchmarks on weakly-supervised classification and localization of common thorax diseases. InCVPR, pages 2097–2106, 2017. 3

    2017

  42. [50]

    Unsu- pervised domain adaptation for medical image segmentation by disentanglement learning and self-training.IEEE Trans- actions on Medical Imaging, 43(1):4–14, 2022

    Qingsong Xie, Yuexiang Li, Nanjun He, Munan Ning, Kai Ma, Guoxing Wang, Yong Lian, and Yefeng Zheng. Unsu- pervised domain adaptation for medical image segmentation by disentanglement learning and self-training.IEEE Trans- actions on Medical Imaging, 43(1):4–14, 2022. 1

    2022

  43. [51]

    A survey on unsupervised domain adaptation in medical imaging: Meth- ods, dataset, and future outlook.Applied Soft Computing, page 115314, 2026

    Hancun Yang, Henjun Guo, Guanlin Liu, Linkuan Zhou, Junlin Xu, Leyi Wei, Ran Su, and Qiangguo Jin. A survey on unsupervised domain adaptation in medical imaging: Meth- ods, dataset, and future outlook.Applied Soft Computing, page 115314, 2026. 1, 2

    2026

  44. [52]

    Can we evaluate domain adaptation models without target-domain labels? InInternational Conference on Learn- ing Representations, pages 35061–35081, 2024

    Jianfei Yang, Hanjie Qian, Yuecong Xu, Kai Wang, and Li- hua Xie. Can we evaluate domain adaptation models without target-domain labels? InInternational Conference on Learn- ing Representations, pages 35061–35081, 2024. 2, 3

    2024

  45. [53]

    Towards accurate model selection in deep unsupervised domain adap- tation

    Kaichao You, Ximei Wang, Mingsheng Long, et al. Towards accurate model selection in deep unsupervised domain adap- tation. InICML, pages 7124–7133. PMLR, 2019. 1, 2, 3

    2019

  46. [54]

    Collaborative unsupervised domain adaptation for medical image diagno- sis.IEEE TIP, 29:7834–7844, 2020

    Yifan Zhang, Ying Wei, Qingyao Wu, et al. Collaborative unsupervised domain adaptation for medical image diagno- sis.IEEE TIP, 29:7834–7844, 2020. 1, 3

    2020

  47. [55]

    In addition, we provide the full within-algorithm validator reliability analyses, reported per algorithm and per scenario, in Section 6.2

    Supplementary Material In this supplementary material, we provide the full results of our complete unsupervised domain adaptation (UDA) pipeline for each evaluated clinically relevant cross-domain scenario in Section 6.1. In addition, we provide the full within-algorithm valid...

    1956

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.