Pith. sign in

REVIEW 4 major objections 5 minor 34 references

Gaze-Assisted Human-Centric Domain Adaptation for Cardiac Ultrasound Image Segmentation

T0 review · 4 major / 5 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read The paper claims that gaze heatmaps from doctors, injected through a cross-attention feature alignment and a gaze-weighted loss, lift cardiac ultrasound segmentation from 70.14% to 76.14% Dice under domain adaptation, beating GAN- and…

desk verdict Interesting gaze-guided UDA idea, but the evaluation leaves the source of the gain uncontrolled and the reported SOTA claim doesn't hold as written. read the letter →

arxiv 2502.03781 v1 pith:5LBOXADO submitted 2025-02-06 cs.CV eess.IV

Ruiyi Li , Yuting He , Rongjun Ge , Chong Wang , Daoqiang Zhang , Yang Chen , Shuo Li This is my paper · ORCID
classification cs.CVeess.IV
keywords cardiacultrasounddomainadaptationimagesegmentationgazeheatmaphuman-centriclearningteacher-studentself-trainingcross-attentionpseudo-labels
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

Cardiac ultrasound segmentation models lose accuracy when they move between hospitals or machines, and existing fixes either generate distorted synthetic images (GAN-based adaptation) or trust unreliable guessed labels (self-training). This paper's proposal is to record where a doctor looks while reading the target-domain images and feed those gaze heatmaps into the adaptation loop. Gaze guides a cross-attention feature-alignment step (Gaze Augment Alignment, GAA) and a loss-reweighting step (Gaze Balance Loss, GBL) that concentrates training on structures the doctor actually scans. On the CAMUS-to-HMC-QU transfer, the method reports 76.14% Dice similarity coefficient versus 70.14% with no adaptation and 71.05% for the Fourier-domain FDA baseline, against a 90.23% fully supervised upper bound. The claim is that human gaze supplies label-free cross-domain guidance that narrows the gap to supervised training by roughly six Dice points.

What carries the argument

The load-bearing object is the doctor gaze heatmap: a 2D map of where the eye tracker sampled during reading, treated as a human-cognition prior that transfers across domains. An extractor converts it to $f_G$; GAA then uses cross-attention (query from the gaze feature, keys and values from the teacher's feature map) to assemble $f_{GA}$, a feature set shared between source and target, and pulls the student features toward it with an MSE loss. GBL regularizes the same heatmap into a weight mask $w$ and multiplies it into the student output before cross-entropy against pseudo-labels, so under-segmented gazed-at regions receive larger gradients and over-segmented distractors are suppressed. The teacher trained on the source domain initializes the student, and the pseudo-labels, the gaze heatmaps, and the two gaze losses together drive adaptation.

What would settle it

Repeat the CAMUS→HMC-QU experiment with target-image gaze heatmaps replaced by smooth random fixation maps matched in spatial extent, or by saliency maps from a generic model, keeping all other settings identical: if Dice stays near 76.14%, the reported gain comes from any spatially concentrated attention prior rather than from doctors' domain knowledge. Also rerun the comparison while holding gaze out of the exact test images and reporting train/test separation: if the margin over FDA shrinks, the baseline comparison was not fully unsupervised.

Watch

Extended reading notes

Core claim

The central claim is that a doctor's gaze heatmap, extracted from eye-tracking on target-domain ultrasound images, carries the cross-domain “human cognition general features” that standard unsupervised domain adaptation lacks, and that two gaze-driven modules turn that signal into better segmentation. GAA uses the gaze heatmap to produce a gaze feature $f_G$, then applies cross-attention with the teacher encoder's features $f_T$ as key and value and $f_G$ as query, forming $f_{GA} = \mathrm{Concat}(f_T, \mathrm{Softmax}(f_G f_T^T / \sqrt{d}) f_T)$; the student's features are aligned to $f_{GA}$ with an MSE loss. GBL turns the gaze heatmap into a weight mask $w$ and computes cross-entropy between the student's gaze-weighted output and the teacher's pseudo-labels, so the loss favors complete segmentation in areas the doctor actually inspected. With both modules, teacher–student training on CAMUS as source and HMC-QU as target reaches 76.14±3.2% DSC and 6.976±1.4 ASSD, which the paper reports as state-of-the-art for this task and as a 5.09-point improvement over FDA and a 30-point improvement over SIFA.

Load-bearing premise

The load-bearing premise is that doctor gaze heatmaps are available on the target-domain images used for adaptation and that those eye movements reliably point at the structures to segment; the paper does not report the number of doctors, calibration procedure, gaze-to-heatmap construction, or train/test separation at gaze-collection time.

Editorial extensions

If this is right

  • With gaze heatmaps available, adaptation on CAMUS→HMC-QU reaches 76.14% DSC, a 5.28-point gain over no adaptation and a 5.09-point gain over FDA.
  • Adding GAA alone adds 1.53 points and GBL alone adds 3.39 points, so the two gaze modules are complementary rather than redundant.
  • The gaze-weighted loss is the larger single contributor, supporting the paper's claim that it counteracts over- and under-segmentation caused by incomplete pseudo-labels.
  • If replicated, the result suggests that a low-cost eye tracker can substitute for part of the annotation effort in clinical domain adaptation, since gaze is recorded passively while a doctor reads images.
  • The same teacher–student recipe with gaze guidance could apply to other cardiac views or to any imaging domain where expert reading behavior is available.

Reading between the lines

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

  • The reported comparison is only apples-to-apples if gaze heatmaps were recorded on target-domain training images and withheld from the test set; if gaze was collected on test images, part of the 6-point gain could be a form of target-side supervision that the unsupervised baselines do not get.
  • Because the eye tracker samples only five times per second, the construction of gaze heatmaps from sparse fixation coordinates (interpolation, smoothing, choice of doctor) is a design choice that can dominate results; ablating gaze quality, number of doctors, and calibration would clarify how much of the gain comes from gaze content versus its implementation.
  • A testable extension is to replace doctor gaze with saliency maps or smooth random heatmaps of similar shape: if DSC remains near 76.14%, the method is using a generic attention prior, whereas if performance drops, the domain-specific human guidance is doing the work.
  • If the mechanism transfers, the same GAA/GBL recipe could be applied to endoscopy, fundus imaging, or other expert-reviewed modalities where doctors already look at images during diagnosis, without requiring new annotations.
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 GAHCDA, a gaze-assisted human-centric domain adaptation framework for cardiac ultrasound image segmentation. It contains two modules: Gaze Augment Alignment (GAA), which uses cross-attention between teacher features and gaze heatmaps to align student features via MSE loss, and Gaze Balance Loss (GBL), which weights segmentation outputs by a gaze-derived mask before computing cross-entropy against pseudo-labels. The method is evaluated by transferring from CAMUS to HMC-QU, reporting DSC of 76.14±3.2% and ASSD of 6.976±1.4 mm, claimed as state-of-the-art. Ablations attribute +1.53 DSC to GAA and +3.39 DSC to GBL relative to no adaptation.

Significance. The idea of injecting clinician gaze as human cognitive guidance into medical-image domain adaptation is clinically motivated and potentially novel. If the method were shown to work without target-test leakage and with a fair comparison to unsupervised baselines, the work would be of interest to the medical imaging community. However, the current manuscript does not provide this evidence: the gaze recording protocol is almost entirely unspecified, the target-side gaze signal may be a privileged input that makes the comparison to unsupervised baselines unfair, Eq. (2) defines a mathematically invalid loss, and the ablation study lacks error bars or statistical tests. The paper ships no code, no machine-checked derivations, and no reproducible protocol, so the quantitative claims rest entirely on the incomplete experiment description.

major comments (4)
  1. [Section III-A and II-B/II-C] The gaze collection protocol is unspecified to a degree that undermines the central comparison. Section III-A says only 'A Tobii Eye-Tracker 5 record the doctor's gaze' with 5 Hz sampling. It does not state how many doctors participated, whether calibration was performed, how raw gaze coordinates were converted to the heatmaps used in Fig. 2 and Fig. 3, which dataset images were viewed, or--critically--whether the gaze recordings were made on the same target-domain images used for student training and pseudo-label generation. Since GAA (Eq. 1) and GBL (Eq. 2) both use these heatmaps on target-domain data, if gaze was recorded on the HMC-QU training set, the comparison against unsupervised baselines such as FDA is not apples-to-apples: GAHCDA uses a privileged target-side signal. If gaze was recorded on the test images, the reported 76.14±3.2 DSC may reflect leakage. Either way, the current text does not establish a controlled experimental setting, and the claim 'attains state-of-the-art performance' in Section III-B.1 is unsupported. The authors must disclose the full recording and heatmap generation protocol, state the train/test split of the gaze data, and, if gaze is available only on target data, reframe the method as privileged-information domain adaptation and compare against appropriate baselines.
  2. [Section II-C, Eq. (2)] Eq. (2) is not a valid cross-entropy loss. Writing L_GB = -(1/N) Σ [y_i log(w_i ŷ_i) + (1 - y_i) log(1 - w_i ŷ_i)] places the gaze weight w_i inside the logarithm. For w_i = 0, the term y_i log(0) is undefined when y_i = 1, and for w_i > 1 the argument w_i ŷ_i can exceed 1, giving a non-positive 'log-probability' and producing a per-sample loss with incorrect sign and curvature. Since the ablation in Section III-C.1 attributes +3.39 DSC to GBL on the basis of this term, the loss formulation is load-bearing for the central claim. The gaze weight should instead multiply the loss outside the logarithm, e.g., -w_i [y_i log ŷ_i + (1 - y_i) log(1 - ŷ_i)], or be incorporated through a different mechanism such as a spatial constraint. The authors should correct the formula and rerun the affected experiments.
  3. [Section III-C.1 and III-C.2] The ablation study is reported only as bar charts without numerical values, error bars, or significance tests. Given that the reported DSC standard deviations for the full method are 3.2 percentage points, the claimed improvements of +1.53 (GAA) and +3.39 (GBL) could be within the run-to-run noise; the paper provides no repeated-run statistics for the individual ablations. The hyperparameter analysis in Fig. 5(b) and 5(c) states that performance 'continuously improves' as λ_gaa and λ_gb increase, but no exact numbers or variance are given, and the range tested is unclear. These results directly support the claim that both modules are effective, so the manuscript needs error bars, the exact table of ablation values, and a statement of the number of independent runs.
  4. [Section III-B.1, Table I] The claim of state-of-the-art performance is not supported by the baseline selection. Table I compares GAHCDA with only four baselines: SIFA, CycleGAN, VAE, and FDA. Modern unsupervised domain adaptation methods for medical image segmentation, such as DaFormer (cited as [29] but not evaluated) and other recent self-training/feature-alignment methods, are missing. Without these comparisons, the phrase 'state-of-the-art' in Section III-B.1 overstates what the experimental evidence demonstrates. The authors should include recent UDA baselines and ideally also a gaze-free variant of their own architecture to isolate the effect of the gaze signal from the effect of the added attention module.
minor comments (5)
  1. [Section III-B.1] The text says the method achieves '30% DSC improvements' over SIFA, but the actual difference is 30 percentage points (76.14 vs. 46.14), which is an approximately 65% relative improvement; please correct the phrasing.
  2. [Section III-A] The sentence 'A Tobii Eye-Tracker 5 record the doctor's gaze' is grammatically incomplete and imprecise; specify the exact device model and the subject population.
  3. [Section II-A] The heading 'GAHCDA Framwork' contains a typo; it should be 'Framework'.
  4. [Figures 2 and 3] The embedded text in the figures appears as garbled glyphs in the PDF rendering, making the architecture diagrams difficult to interpret; please provide clean vector figures with readable labels.
  5. [Section II-C] The construction of the 'regularized gaze heatmap' is not described anywhere; at minimum define the regularization (e.g., Gaussian smoothing, normalization) used to produce the weight mask w.

Circularity Check

0 steps flagged · score 0.0 of 10

No derivation-level circularity; the gaze protocol under-specification is an experimental fairness concern, not a circularity concern.

full rationale

This paper contains no mathematical derivation that reduces to its own inputs. GAA in Eq. (1) defines fGA as a cross-attention combination of teacher features fT and gaze-derived features fG, then aligns student features to fGA with MSE; this is a proposed mechanism, not a quantity fitted from the reported outcome. GBL in Eq. (2) is a weighted cross-entropy loss whose weights come from a regularized gaze heatmap; the gaze heatmap is an input prior, not a parameter fit to the reported DSC, so the reported 76.14% DSC value is an empirical result, not a renamed fit. The state-of-the-art claim rests on the Table I comparison against unsupervised baselines, and the description of the gaze data is genuinely under-specified: Section III-A says only 'A Tobii Eye-Tracker 5 record the doctor's gaze' and does not state whether gaze was recorded on the target training images, the test images, or both. That omission creates a legitimate apples-to-apples concern because GAHCDA receives target-side gaze guidance, but it is an experimental-design concern rather than a circular derivation. There is no load-bearing self-citation: the only overlapping-author citation, reference [2], supports the introductory statement that fully supervised methods require extra annotations and is not used to justify any central claim. No uniqueness theorem, ansatz-smuggling citation, or renaming of a known result is present. Therefore the appropriate circularity score is 0.

Assumptions & free parameters 5 free parameters · 4 assumptions · 1 invented entities

The framework rests on a private gaze dataset, a teacher-student pseudo-label assumption, and the assumption that gaze heatmaps encode segmentation-relevant human knowledge. Hyperparameters are tuned but only partially reported, and no code or data are released. The main invented construct is the 'human cognition general feature', which is not externally validated.

free parameters (5)
  • lambda_gaa = tuned; exact values not fully reported (Fig. 5b)
    Weight on the GAA MSE loss in Eq. (3), chosen on target-domain validation; affects the balance between feature alignment and segmentation losses.
  • lambda_gb = tuned; exact values not fully reported (Fig. 5c)
    Weight on the GBL loss in Eq. (3), chosen on target-domain validation; controls how strongly gaze reweights the pseudo-label cross-entropy.
  • lambda_dice = not stated
    Weight on the DICE loss in Eq. (3); the paper does not report its value.
  • lambda_ce = not stated
    Weight on the standard cross-entropy loss in Eq. (3); the paper does not report its value.
  • gaze heatmap smoothing and regularization = unspecified
    The construction of gaze heatmaps from 5 Hz eye-tracker samples is not described; any smoothing kernel or normalization affects Eq. (2) and the alignment signal.
assumptions (4)
  • domain assumption Doctor gaze trajectories contain transferable cross-domain recognition knowledge.
    Sections I and II-B assume gaze encodes 'human cognition general features' that transfer between source and target domains; no evidence is provided that gaze correlates with segmentation-relevant structure.
  • domain assumption Teacher pseudo-labels for target images are of sufficient quality for self-training.
    Section II-A uses teacher predictions as y_t in Eq. (2); if pseudo-labels are systematically wrong, GBL propagates the error into the student.
  • domain assumption Gaze heatmap regions correspond to segmentation under/over-segmentation areas.
    Section II-C states doctors focus on difficult-to-segment areas, but no analysis links gaze fixation patterns to actual segmentation errors.
  • domain assumption Cross-attention with gaze features aligns student and teacher features across domains.
    Section II-B and Eq. (1) define LGAA as MSE between f_GA and f_S, but no theoretical or empirical validation shows that this alignment transfers across domains.
invented entities (1)
  • Human cognition general features f_GA
    purpose: Cross-domain feature representation extracted from gaze heatmaps and teacher features, used to align the student via MSE in GAA.
    Postulated internal representation; the paper shows qualitative feature maps but no external validation that the feature captures cognition or generalizes beyond the reported dataset.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Gaze-Assisted Human-Centric Domain Adaptation for Cardiac Ultrasound Image Segmentation." pith.science (2026). https://pith.science/paper/5LBOXADO

@misc{pith2026250203781,
  author       = {Pith},
  title        = {Pith review of: Gaze-Assisted Human-Centric Domain Adaptation for Cardiac Ultrasound Image Segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5LBOXADO}},
  note         = {Machine review of arXiv:2502.03781}
}
read the original abstract

Domain adaptation (DA) for cardiac ultrasound image segmentation is clinically significant and valuable. However, previous domain adaptation methods are prone to be affected by the incomplete pseudo-label and low-quality target to source images. Human-centric domain adaptation has great advantages of human cognitive guidance to help model adapt to target domain and reduce reliance on labels. Doctor gaze trajectories contains a large amount of cross-domain human guidance. To leverage gaze information and human cognition for guiding domain adaptation, we propose gaze-assisted human-centric domain adaptation (GAHCDA), which reliably guides the domain adaptation of cardiac ultrasound images. GAHCDA includes following modules: (1) Gaze Augment Alignment (GAA): GAA enables the model to obtain human cognition general features to recognize segmentation target in different domain of cardiac ultrasound images like humans. (2) Gaze Balance Loss (GBL): GBL fused gaze heatmap with outputs which makes the segmentation result structurally closer to the target domain. The experimental results illustrate that our proposed framework is able to segment cardiac ultrasound images more effectively in the target domain than GAN-based methods and other self-train based methods, showing great potential in clinical application.

Figures

Figures reproduced from arXiv: 2502.03781 by the authors.

Figure 1
Figure 1. (a) Domain adaptation based on GAN is prone to mode col [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. The Gaze-Assisted Human-Centric domain adaptation [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The GAA helps the student model leverage general feat [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Visual comparison of outputs. The results show that o [PITH_FULL_IMAGE:figures/full_fig_p003_4.png]
Figure 5
Figure 5. Figure 5: The ablation study demonstrated that GAA and GBL effe [PITH_FULL_IMAGE:figures/full_fig_p004_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 31 canonical work pages

  1. [29]

    Daformer: Improving n etwork architectures and training strategies for domain-adaptiv e semantic seg- mentation,

    L. Hoyer, D. Dai, and L. V an Gool, “Daformer: Improving n etwork architectures and training strategies for domain-adaptiv e semantic seg- mentation,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2022, pp. 9924–9935

    work page 2022

  2. [1]

    Domain generalization for prostate segmentation in transrectal u ltrasound images: A multi-center study,

    S. V esal, I. Gayo, I. Bhattacharya, S. Natarajan, L. S. Ma rks, D. C. Barratt, R. E. Fan, Y . Hu, G. A. Sonn, and M. Rusu, “Domain generalization for prostate segmentation in transrectal u ltrasound images: A multi-center study,” Medical image analysis, vol. 82, p. 102620, 2022

    work page 2022

  3. [2]

    Imagi ng foundation model for universal enhancement of non-ideal me asurement ct,

    Y . Liu, R. Ge, Y . He, Z. Wu, C. Y ou, S. Li, and Y . Chen, “Imagi ng foundation model for universal enhancement of non-ideal me asurement ct,” arXiv preprint arXiv:2410.01591 , 2024

    arXiv 2024

  4. [3]

    Deep semi-supervised ultrasound image segmentat ion by using a shadow aware network with boundary refinement,

    F. Chen, L. Chen, W. Kong, W. Zhang, P . Zheng, L. Sun, D. Zha ng, and H. Liao, “Deep semi-supervised ultrasound image segmentat ion by using a shadow aware network with boundary refinement,” IEEE Transactions on Medical Imaging , 2023

    work page 2023

  5. [4]

    Shadow-consistent semi-supervised learning for prostat e ultrasound segmentation,

    X. Xu, T. Sanford, B. Turkbey, S. Xu, B. J. Wood, and P . Y an, “Shadow-consistent semi-supervised learning for prostat e ultrasound segmentation,” IEEE Transactions on Medical Imaging , vol. 41, no. 6, pp. 1331–1345, 2021

    work page 2021

  6. [5]

    Contrastive rendering with semi-supervised learning for ovary and follicle segmentation from 3d ultras ound,

    X. Y ang, H. Li, Y . Wang, X. Liang, C. Chen, X. Zhou, F. Zeng, J. Fang, A. Frangi, Z. Chen et al. , “Contrastive rendering with semi-supervised learning for ovary and follicle segmentation from 3d ultras ound,” Med- ical Image Analysis , vol. 73, p. 102134, 2021

    work page 2021

  7. [6]

    Contrastive semi-supervised learning for domai n adaptive segmentation across similar anatomical structures,

    R. Gu, J. Zhang, G. Wang, W. Lei, T. Song, X. Zhang, K. Li, an d S. Zhang, “Contrastive semi-supervised learning for domai n adaptive segmentation across similar anatomical structures,” IEEE Transactions on Medical Imaging , vol. 42, no. 1, pp. 245–256, 2023

    work page 2023

  8. [7]

    Synergis tic im- age and feature adaptation: Towards cross-modality domain adaptation for medical image segmentation,

    C. Chen, Q. Dou, H. Chen, J. Qin, and P .-A. Heng, “Synergis tic im- age and feature adaptation: Towards cross-modality domain adaptation for medical image segmentation,” in Proceedings of The Thirty-Third Conference on Artificial Intelligence (AAAI) , 2019, pp. 865–872

    work page 2019

Show all 34 references
  1. [8]

    Unsupervi sed bidirectional cross-modality adaptation via deeply syner gistic image and feature alignment for medical image segmentation,

    C. Chen, Q. Dou, H. Chen, J. Qin, and P . A. Heng, “Unsupervi sed bidirectional cross-modality adaptation via deeply syner gistic image and feature alignment for medical image segmentation,” arXiv preprint arXiv:2002.02255, 2020

    2002 arXiv

  2. [9]

    Unpaired im age-to-image translation using cycle-consistent adversarial networks ,

    J.-Y . Zhu, T. Park, P . Isola, and A. A. Efros, “Unpaired im age-to-image translation using cycle-consistent adversarial networks ,” in Computer Vision (ICCV), 2017 IEEE International Conference on , 2017

    2017

  3. [10]

    Self-ensembling with gan-b ased data augmentation for domain adaptation in semantic segmentati on,

    J. Choi, T. Kim, and C. Kim, “Self-ensembling with gan-b ased data augmentation for domain adaptation in semantic segmentati on,” in Proceedings of the IEEE/CVF international conference on co mputer vision, 2019, pp. 6830–6840

    2019

  4. [11]

    The domain shift problem of medical image segmentation and vendor- adaptation by unet-gan,

    W. Y an, Y . Wang, S. Gu, L. Huang, F. Y an, L. Xia, and Q. Tao, “The domain shift problem of medical image segmentation and vendor- adaptation by unet-gan,” in Medical Image Computing and Computer Assisted Intervention–MICCAI 2019: 22nd International Co nference, Shenzhen, Ch...

    2019

  5. [12]

    Structure preserving cycle-gan for unsupervised medical image domain adaptation,

    P . Iacono and N. Khan, “Structure preserving cycle-gan for unsupervised medical image domain adaptation,” arXiv preprint arXiv:2304.09164 , 2023

    2023 arXiv

  6. [13]

    Mi 2 ga n: gen- erative adversarial network for medical image domain adapt ation using mutual information constraint,

    X. Xie, J. Chen, Y . Li, L. Shen, K. Ma, and Y . Zheng, “Mi 2 ga n: gen- erative adversarial network for medical image domain adapt ation using mutual information constraint,” in International Conference on Medical Image Computing and Computer-Assisted Intervention . Springer,...

    2020

  7. [14]

    A novel 3d unsupervised domain adaptation framework for cros s-modality medical image segmentation,

    K. Y ao, Z. Su, K. Huang, X. Y ang, J. Sun, A. Hussain, and F. Coenen, “A novel 3d unsupervised domain adaptation framework for cros s-modality medical image segmentation,” IEEE Journal of Biomedical and Health Informatics, vol. 26, no. 10, pp. 4976–4986, 2022

    2022

  8. [15]

    What can be transferred: Unsupervised domain adaptation for endoscop ic lesions segmentation,

    J. Dong, Y . Cong, G. Sun, B. Zhong, and X. Xu, “What can be transferred: Unsupervised domain adaptation for endoscop ic lesions segmentation,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2020, pp. 4023–4032

    2020

  9. [16]

    Self-attentive spatial adaptive normalization for cross -modality domain adaptation,

    D. Tomar, M. Lortkipanidze, G. Vray, B. Bozorgtabar, an d J.-P . Thiran, “Self-attentive spatial adaptive normalization for cross -modality domain adaptation,” IEEE transactions on medical imaging , vol. 40, no. 10, pp. 2926–2938, 2021

    2021

  10. [17]

    Unsup ervised intra- domain adaptation for semantic segmentation through self- supervision,

    F. Pan, I. Shin, F. Rameau, S. Lee, and I. S. Kweon, “Unsup ervised intra- domain adaptation for semantic segmentation through self- supervision,” in Proceedings of the IEEE/CVF conference on computer vision a nd pattern recognition, 2020, pp. 3764–3773

    2020

  11. [18]

    Seeing what a gan cannot generate,

    D. Bau, J.-Y . Zhu, J. Wulff, W. Peebles, H. Strobelt, B. Z hou, and A. Torralba, “Seeing what a gan cannot generate,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2019, pp. 4502–4511

    2019

  12. [19]

    Ultrasound domain adaptation using frequency domain analysis,

    M. Sharifzadeh, A. K. Tehrani, H. Benali, and H. Rivaz, “ Ultrasound domain adaptation using frequency domain analysis,” in 2021 IEEE International Ultrasonics Symposium (IUS) . IEEE, 2021, pp. 1–4

    2021

  13. [20]

    Asc: Appearance and str ucture consistency for unsupervised domain adaptation in fetal br ain mri seg- mentation,

    Z. Xu, H. Gong, X. Wan, and H. Li, “Asc: Appearance and str ucture consistency for unsupervised domain adaptation in fetal br ain mri seg- mentation,” in International Conference on Medical Image Computing and Computer-Assisted Intervention . Springer, 2023, pp. 325–335

    2023

  14. [21]

    Fda: Fourier domain adaptation f or semantic segmentation,

    Y . Y ang and S. Soatto, “Fda: Fourier domain adaptation f or semantic segmentation,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2020, pp. 4085–4095

    2020

  15. [22]

    Source free domain adaptation for medical image segmentation with fourier style mining,

    C. Y ang, X. Guo, Z. Chen, and Y . Y uan, “Source free domain adaptation for medical image segmentation with fourier style mining,” Medical Image Analysis , vol. 79, p. 102457, 2022

    2022

  16. [23]

    Unsupervised domain adaptation wi th variational approximation for cardiac segmentation,

    F. Wu and X. Zhuang, “Unsupervised domain adaptation wi th variational approximation for cardiac segmentation,” IEEE Transactions on Medical Imaging, vol. 40, no. 12, pp. 3555–3567, 2021

    2021

  17. [24]

    Metacorrection: Dom ain-aware meta loss correction for unsupervised domain adaptation in semantic segmentation,

    X. Guo, C. Y ang, B. Li, and Y . Y uan, “Metacorrection: Dom ain-aware meta loss correction for unsupervised domain adaptation in semantic segmentation,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2021, pp. 3927–3936

    2021

  18. [25]

    Unsupervised domai n adaptation for semantic segmentation via class-balanced self-traini ng,

    Y . Zou, Z. Y u, B. Kumar, and J. Wang, “Unsupervised domai n adaptation for semantic segmentation via class-balanced self-traini ng,” in Proceed- ings of the European conference on computer vision (ECCV) , 2018, pp. 289–305

    2018

  19. [26]

    Unsupervised domain adap tation to improve image segmentation quality both in the source and ta rget domain,

    J.-A. Bolte, M. Kamp, A. Breuer, S. Homoceanu, P . Schlic ht, F. Huger, D. Lipinski, and T. Fingscheidt, “Unsupervised domain adap tation to improve image segmentation quality both in the source and ta rget domain,” in proceedings of the IEEE/CVF conference on computer vision ...

    2019

  20. [27]

    Unsupervised domain adaptation in brain lesion segmentat ion with adversarial networks,

    K. Kamnitsas, C. Baumgartner, C. Ledig, V . Newcombe, J. Simpson, A. Kane, D. Menon, A. Nori, A. Criminisi, D. Rueckert et al. , “Unsupervised domain adaptation in brain lesion segmentat ion with adversarial networks,” in Information Processing in Medical Imaging: 25th Interna...

    2017

  21. [28]

    Category anchor- guided unsupervised domain adaptation for semantic segmentation ,

    Q. Zhang, J. Zhang, W. Liu, and D. Tao, “Category anchor- guided unsupervised domain adaptation for semantic segmentation ,” Advances in neural information processing systems , vol. 32, 2019

    2019

  22. [30]

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

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

    2010 arXiv

  23. [31]

    Unsup ervised domain adaptation for medical image segmentation by select ive entropy constraints and adaptive semantic alignment,

    W. Feng, L. Ju, L. Wang, K. Song, X. Zhao, and Z. Ge, “Unsup ervised domain adaptation for medical image segmentation by select ive entropy constraints and adaptive semantic alignment,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 37, no. 1, 2023, pp. 623–631

    2023

  24. [32]

    Deep learning for segmentation using an open large-scale d ataset in 2d echocardiography,

    S. Leclerc, E. Smistad, J. Pedrosa, A. Østvik, F. Cerven ansky, F. Es- pinosa, T. Espeland, E. A. R. Berg, P .-M. Jodoin, T. Grenier et al. , “Deep learning for segmentation using an open large-scale d ataset in 2d echocardiography,” IEEE transactions on medical imaging , vol....

    2019

  25. [33]

    Early myocardial infarction detection over multi-view ec hocardiogra- phy,

    A. Degerli, S. Kiranyaz, T. Hamid, R. Mazhar, and M. Gabb ouj, “Early myocardial infarction detection over multi-view ec hocardiogra- phy,” Biomedical Signal Processing and Control , vol. 87, p. 105448, 2024

    2024

  26. [34]

    Unsupervised domain adaptation through shape model ing for medical image segmentation,

    Y . Y ao, F. Liu, Z. Zhou, Y . Wang, W. Shen, A. Y uille, and Y . Lu, “Unsupervised domain adaptation through shape model ing for medical image segmentation,” in Proceedings of The 5th International Conference on Medical Imaging with Deep Lear ning, ser. Proceedings of Machine...

    2022 arXiv

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.