Pith. sign in

REVIEW 4 major objections 6 minor 68 references

ACE: Anatomically Consistent Embeddings in Composition and Decomposition

T0 review · 4 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read ACE pretrains medical-image encoders to respect part-whole anatomy, and the resulting embeddings transfer to downstream classification, segmentation, and few-shot tasks.

desk verdict A solid, well-motivated medical SSL method with a genuinely new grid-based composition/decomposition loss; the evaluation is broad, but the missing Adam-v2 baseline and unspecified local-branch augmentations need fixing before the central correspondence claim is verifiable. read the letter →

arxiv 2501.10131 v1 pith:JGEZZGQQ submitted 2025-01-17 cs.CV

Ziyu Zhou , Haozhe Luo , Mohammad Reza Hosseinzadeh Taher , Jiaxuan Pang , Xiaowei Ding , Michael Gotway , Jianming Liang This is my paper · ORCID
classification cs.CV
keywords self-supervisedlearningmedicalimageanalysiscompositionalitydecompositionalityanatomicallyconsistentembeddingschestX-rayfew-shotstudent-teachermodel
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 argues that medical images carry a composable and decomposable anatomical structure that existing self-supervised learning methods ignore. It introduces ACE, a self-supervised pretraining method that learns global consistency between overlapping crops and local consistency through composition and decomposition losses. If the claim holds, ACE gives medical imaging models a representation that transfers better to classification, segmentation, few-shot, and landmark tasks, which matters when labeled data is scarce. The paper reports this across six datasets and two backbones, with an ablation showing each loss component contributes.

What carries the argument

The load-bearing mechanism is grid-wise image cropping: an input image is divided into a 32x32 patch grid, and two random crops of 14x14 and 28x28 patches are extracted so that every large-crop patch overlaps four small-crop patches in exact spatial correspondence. A composer head merges 2x2 student patch embeddings into one, a decomposer head splits one patch embedding into four, and CLIP-style cross-correlation matrices with Gaussian-smoothed targets supervise the matching. A student-teacher EMA setup ties the global and local branches together. This replaces approximate nearest-neighbor patch matching with precise, anatomy-aligned matching.

What would settle it

Pretrain ACE with flips or random-resized crops that break grid alignment in the local branch while keeping everything else fixed; if decompositionality accuracy (the f(X) - f(X_excised) ≈ f(C) test) drops sharply or downstream transfer degrades, the precise-patch-matching premise is load-bearing. Alternatively, inspect the learned matching matrix M_comp: if high correlation values appear far from the Gaussian-smoothed diagonal, the model is not actually honoring spatial alignment.

Watch

Extended reading notes

Core claim

ACE's central claim is that training a student-teacher network to respect part-whole relationships in standardized medical images produces anatomically consistent embeddings, and that this consistency transfers to downstream tasks. Concretely, the paper trains the embedding of a whole patch to be consistent with the aggregated embeddings of its sub-patches (composition) and trains sub-patch embeddings to be consistent with the decomposition of the whole-patch embedding (decomposition), using precisely matched grid-aligned crops. The paper reports that ACE outperforms a diverse set of SSL baselines on ChestX-ray14, Shenzhen, RSNA, JSRT, ChestX-Det, and SIIM, and that it shows emergent cross-patient landmark correspondence and useful few-shot performance, including over 91% of full-data heart-segmentation Dice from two labeled samples.

Load-bearing premise

The local consistency branch assumes that, after resizing and random augmentation, a grid position in one crop still corresponds to the same anatomical grid position in the other crop, so the Gaussian-smoothed target matrix pairs the right patches.

Editorial extensions

If this is right

  • Pretrained ACE weights can be fine-tuned for thoracic disease classification and segmentation with substantially less labeled data; the paper reports 2-shot heart segmentation reaching over 91% of full-data performance.
  • The embedding space supports unsupervised cross-patient anatomy correspondence, meaning the same model can retrieve or match anatomical landmarks across different patients without fine-tuning.
  • Because the method only assumes standardized imaging protocols, it should transfer to other modalities with consistent anatomy; the paper demonstrates this on fundus photography.
  • The composition and decomposition losses provide a training signal that bridges large pathologies and small tissue anomalies, which should improve weakly supervised localization of small findings such as nodules.

Reading between the lines

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

  • The paper tests ACE only on 2D projection images; if the spatial-alignment premise holds, the same grid-wise composition/decomposition losses could extend to 3D volumes such as CT and MRI, where sub-volume part-whole structure is even more explicit.
  • The Gaussian-smoothed target matrix is a soft spatial prior; replacing it with a hard one-hot target or a learned soft assignment would isolate how much of ACE's gain comes from the spatial prior versus the composition/decomposition objective itself.
  • The measured decompositionality property, where removing a region from an image yields an embedding difference close to the embedding of the excised region, suggests ACE features could support localized image editing or disease attribution, a use the paper only hints at through Grad-CAM visualizations.
  • The paper's local branch assumes augmentation preserves grid alignment; a testable consequence is that strong flips or severe random-resized cropping during pretraining should degrade the composition/decomposition accuracy and downstream transfer.
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 / 6 minor

Summary. The paper proposes ACE, a self-supervised pretraining method for medical images (primarily chest X-rays) that combines a global consistency branch, which aligns averaged patch embeddings of overlapping grid-based crops in a DINO-style student--teacher setup, with a local consistency branch that enforces composition and decomposition relationships between patch embeddings through Gaussian-smoothed correspondence matrices. The authors claim that ACE learns anatomically consistent embeddings, exhibits emergent properties such as cross-patient anatomical correspondence and symmetry, and transfers better than existing SSL baselines to classification, segmentation, few-shot, and keypoint-detection tasks. The method is evaluated on six downstream datasets with ViT-B and Swin-B backbones, ablations show that each loss component contributes, and code and pretrained models are promised for release.

Significance. If the claims hold, ACE is a valuable contribution to medical image SSL: it explicitly models hierarchical part--whole structure, a prior that most SSL methods ignore, through an intuitive grid-wise cropping scheme and two complementary losses. The paper has several strengths: a clear and coherent two-branch objective; a broad downstream evaluation spanning classification, segmentation, few-shot transfer, and landmark detection on two backbones; an ablation study in Fig. 9 showing that each loss term improves performance; statistical testing (two-sample t-test) reported in Table 1; and a commitment to releasing code and pretrained models. The main uncertainties are whether the local correspondence mechanism is correctly implemented as described and whether the claimed superiority over the closest prior method, Adam-v2, actually holds, since that baseline is not evaluated anywhere in the paper.

major comments (4)
  1. [Section 3.3 / Eq. (4) / Algorithm 1] The local consistency branch hinges on precise 4-to-1 patch matching between C1 and C2 after resizing and augmentation, but the paper never specifies the augmentations applied to the local branch (T1 and T2 in Sec. 3.2; augment(C1) and augment(C2) in Algorithm 1). If these augmentations include flips or random-resized crops, the index-based target in Eq. (4) is wrong. Additionally, Sec. 3.1 describes 1024x1024 input with 32x32 patches (a 32x32 grid), while Sec. 4 and Appendix A.3 describe pretraining at 448x448 with a 14x14 grid, and Algorithm 1's ComputeLoss builds the target from idx1=nonzero(O1) and idx2=nonzero(O2) without any coordinate-alignment step. Please specify the augmentation pipeline for the local branch, describe how correspondences are recomputed when the two crops have different grid resolutions, and reconcile the input-size discrepancy. Without this, the central anatomical-correspondence premise is unverifiable and the local loss may be training the model to align mismatched tissues.
  2. [Related Work / Table 1] The closest prior work on compositional and decompositional anatomy SSL, Adam-v2 [46], is cited in the Related Work section as 'the latest research... exploiting learning from composition and decomposition in a hierarchical way,' but Adam-v2 is not included in any of the experiments: it is absent from Table 1, the few-shot experiments in Fig. 8, the ablation in Fig. 9, and the property analyses in Secs. 5.1 and 5.2. Since ACE's contribution is specifically the combination of global consistency with local composition/decomposition, the lack of a head-to-head comparison with Adam-v2 leaves the paper's central competitiveness claim unsubstantiated against the most relevant baseline. Please add Adam-v2 to the comparison tables and property analyses, or provide a specific technical justification for its exclusion.
  3. [Section 5.2 / Figs. 6 and 7] The 'emergent' properties are supported by t-SNE visualizations in Fig. 6 and a single mean landmark error of 61 pixels in Fig. 7, with no error bars, no statistical tests, and no comparison to any baseline feature extractor under the same zero-shot protocol. As written, these results do not demonstrate that ACE's embeddings are uniquely anatomically organized; for example, the 61-pixel mean error could plausibly be matched by other SSL features on this task, especially given the high degree of consistency in chest X-ray anatomy. Please report quantitative accuracy (e.g., success rate within a pixel tolerance), the error distribution across the 13 landmarks, and direct comparisons to baselines on the same unsupervised landmark-matching task, preferably with multiple seeds.
  4. [Appendix A.1 / Algorithm 1] The Pytorch pseudo-code contains implementation errors that make it irreproducible as written: loss_decomp is computed with s1 and t2 instead of the decomposition branch's s2 and t1; the matching matrix is computed with torch.mul (element-wise multiplication) rather than the matrix multiplication required by Eq. (3); and the comments contain typos (e.g., 'input composer head' for the decomposer, 'flattern'). Please correct the pseudo-code so that it matches the main-text equations, and proofread the supplementary material for such errors.
minor comments (6)
  1. [Sections 4 and 5.2] The notation '4482' and '10242' appears to be a formatting artifact for 448x448 and 1024x1024; please fix these superscripts for clarity (e.g., 448^2, 1024^2).
  2. [Table 1 caption] The caption describes t-tests and blue highlighting, but the typeset table does not show any highlighting and no p-values or effect sizes are reported; please include the full statistical results either in the table or in the supplementary material so the 'no statistically significant difference' claims can be verified.
  3. [Figure 3] The KDE distributions for compositionality are shown only qualitatively; please report the mean and standard deviation of the cosine similarities for each method so the 'narrower and taller' claim can be assessed numerically.
  4. [Figure 5] The retrieval experiment reports ACE's accuracy (94.37%) in the text but no baseline numbers; please provide a full comparison table or include the baseline numbers directly in the figure.
  5. [Section 3.1] Please clarify how the two random crops C1 and C2 are sampled from the 32x32 grid (e.g., random offsets, fixed relative positions) and how the overlap masks O1 and O2 are defined; this will help readers understand the correspondence construction that the method relies on.
  6. [Section 3.2] The operator '⊕' used for average pooling is not formally defined; please provide a precise definition to avoid confusion with the later uses of matrix notation.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: ACE's transfer and emergent-property claims rest on external benchmarks, not on its own training targets.

full rationale

ACE is an empirical self-supervised method: Eqs. (1), (5), (7), and (8) define training objectives rather than derived predictions, so there is no fitted constant later renamed as a prediction. The local-consistency target (Eq. 4) is constructed from the known grid geometry of the two crops; training the matching matrix toward it is the method's definition, and the paper does not present that training success as an external confirmation. The Sec. 5.1 compositionality/decompositionality checks are explicitly labeled "properties that ACE was explicitly trained to learn," so they are sanity checks of the learned objective, not emergent predictions. The central transferability claim (Sec. 5.3, Table 1) and the zero-shot cross-patient landmark matching (Sec. 5.2) are evaluated on held-out external labels and are therefore not forced by the training loss. Citations to the authors' PEAC, Adam, and Adam-v2 are contextual comparisons and building blocks; no load-bearing argument reduces to a self-citation. The missing augmentation specification and the 1024x1024 versus 448x448 resolution mismatch are reproducibility and correctness concerns, not circularity.

Assumptions & free parameters 6 free parameters · 5 assumptions · 2 invented entities

ACE is an empirical pretraining method, so the main load-bearing elements are hyperparameters chosen by hand and domain assumptions about anatomical consistency and spatial correspondence. The composer and decomposer heads are learned architectural components introduced by the paper without independent external validation. No physical constants or externally fitted parameters are involved.

free parameters (6)
  • Loss weights lambda1, lambda2, lambda3 = 0.1, 1, 1
    Chosen by hand in Eq. 8 to balance global, composition, and decomposition losses; no sensitivity study reported.
  • Composition loss alpha = 0.9
    Hyperparameter used in Eq. 5 to balance positive and negative samples; chosen without reported tuning.
  • Decomposition loss alpha = 0.99
    Hyperparameter used in Eq. 7; chosen separately from the composition alpha without reported tuning.
  • Gaussian kernel sigma and kernel size k = sigma=1, k=3
    Parameters of the smoothing target matrix in Eq. 4; the paper does not test other values.
  • Softmax temperatures tau_s, tau_t = not reported
    Used in the global consistency softmax distributions in Section 3.2; values are not given.
  • Crop grid sizes = C1: 14x14 patches, C2: 28x28 patches
    Design choice for the grid-wise cropping in Section 3.1; no ablation on other grid sizes.
assumptions (5)
  • domain assumption Medical images from standardized protocols show consistent global and local anatomical structures across patients.
    This premise motivates the whole method and is stated in Section 1; if anatomy were not consistent, global and local alignment would not transfer.
  • domain assumption Spatial proximity in the image grid corresponds to anatomical proximity, so the Gaussian target matrix encodes correct correspondences.
    Eq. 4 assigns high target values to nearby patches and zero to distant ones; broken by flips or resized-crop augmentations.
  • domain assumption A student-teacher setup with an EMA teacher and a cross-entropy matching loss avoids representational collapse.
    The method relies on DINO/BYOL-style self-distillation in Section 3.2 and Appendix A.3; collapse prevention is not analyzed for the local branches.
  • ad hoc to paper A learned 2-layer MLP composer and decomposer head can faithfully represent part-whole relations.
    Eqs. 2 and 6 define composition and decomposition through these heads; their ability to encode true anatomical compositionality is assumed and only indirectly evaluated.
  • domain assumption Pretraining on unlabeled ChestX-ray14 images transfers to the downstream classification, segmentation, and landmark tasks used in the paper.
    The evaluation in Section 5 assumes transfer from the pretraining domain to JSRT, SIIM, RSNA, Shenzhen, and fundus datasets.
invented entities (2)
  • Composer head (2-layer MLP)
    purpose: Merges 2x2 patch embeddings into a single embedding to enforce compositionality in Eq. 2.
    Its utility is only demonstrated through this paper's downstream tasks; there is no external falsifiable handle.
  • Decomposer head (2-layer MLP)
    purpose: Expands one patch embedding into 2x2 sub-embeddings to enforce decompositionality in Eq. 6.
    Validated only through the paper's internal evaluations and downstream fine-tuning.

how reviews work

0 comments
Cite this review

Pith. "Pith review of ACE: Anatomically Consistent Embeddings in Composition and Decomposition." pith.science (2026). https://pith.science/paper/JGEZZGQQ

@misc{pith2026250110131,
  author       = {Pith},
  title        = {Pith review of: ACE: Anatomically Consistent Embeddings in Composition and Decomposition},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JGEZZGQQ}},
  note         = {Machine review of arXiv:2501.10131}
}
read the original abstract

Medical images acquired from standardized protocols show consistent macroscopic or microscopic anatomical structures, and these structures consist of composable/decomposable organs and tissues, but existing self-supervised learning (SSL) methods do not appreciate such composable/decomposable structure attributes inherent to medical images. To overcome this limitation, this paper introduces a novel SSL approach called ACE to learn anatomically consistent embedding via composition and decomposition with two key branches: (1) global consistency, capturing discriminative macro-structures via extracting global features; (2) local consistency, learning fine-grained anatomical details from composable/decomposable patch features via corresponding matrix matching. Experimental results across 6 datasets 2 backbones, evaluated in few-shot learning, fine-tuning, and property analysis, show ACE's superior robustness, transferability, and clinical potential. The innovations of our ACE lie in grid-wise image cropping, leveraging the intrinsic properties of compositionality and decompositionality of medical images, bridging the semantic gap from high-level pathologies to low-level tissue anomalies, and providing a new SSL method for medical imaging.

Figures

Figures reproduced from arXiv: 2501.10131 by the authors.

Figure 1
Figure 1. (a) Chest X-rays contain various large (global) and small [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. ACE learns anatomically consistent embedding with two key branches: (1) global consistency and (2) local consistency via composition decomposition. Using our proposed grid-wise image cropping strategy (detailed in Sec. 3.1), an input image is divided into a grid (see white grids), and two random crops, C1 (yellow grids) and C2 (green grids), are extracted. In the overlap region between C1 and C2, four patches in C1 … view at source ↗
Figure 3
Figure 3. ACE preserves the compositionality of anatomical struc￾tures in its learned embedding space. As seen, ACE’s distribution is narrower and taller compared with DINO [6], DropPos [52] and SelfPatch [60], with the mean similarity between embeddings of patches and their compositional parts closer to 1. (2) ACE enhances feature decompositionality. Experimental Setup: We examine ACE’s ability to main￾tain the decomposition… view at source ↗
Figures from the paper (8 more)
Figure 5
Figure 5. Figure 5: ACE captures semantics-rich features in its learned em￾bedding space. As seen, ACE achieves higher retrieval accuracy compared with other SSL baselines. flect the locality of anatomical structures in its learned em￾bedding space. To do so, we compile a dataset of 1,000…
Figure 6
Figure 6. Figure 6: ACE reflects locality of anatomical structures in its learned embedding space as an emergent property. As seen, ACE, unlike the SSL baselines, distinguishes different anatomical struc￾tures in its embedding space while keeping identical anatomical structures across pat…
Figure 8
Figure 8. Figure 8: ACE demonstrates superior performance in limited data regimes. As seen in both heart segmentation and pneumotho￾rax classification tasks, ACE surpasses SSL baselines (DINO and POPAR), particularly in few-shot transfer settings. (2) Fine-tuning evaluation Experimental S…
Figure 7
Figure 7. Figure 7: ACE demonstrates its ability to accurately identify [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 10
Figure 10. Figure 10: (a) Comparison of proposed ACE with other baselines. [PITH_FULL_IMAGE:figures/full_fig_p008_10.png]
Figure 12
Figure 12. Figure 12: ACE demonstrates its ability to boost downstream key [PITH_FULL_IMAGE:figures/full_fig_p011_12.png]
Figure 11
Figure 11. Figure 11: ACE reflects the symmetry of anatomical structures in its learned embedding space as an emergent property. As seen, ACE provides mirrored embeddings for mirrored anatomi￾cal structures (e.g., the right and left clavicles, and the right and left rib 5.). B.2. Fine-tuni…
Figure 13
Figure 13. Figure 13: Visualization of Grad-CAM heatmaps. For each column, we provide the heatmap examples for 8 thorax diseases that hold [PITH_FULL_IMAGE:figures/full_fig_p012_13.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

68 extracted references · 50 canonical work pages

  1. [46]

    Representing part-whole hierarchies in foundation models by learning localizability composability and decomposability from anatomy via self supervision

    Mohammad Reza Hosseinzadeh Taher, Michael B Gotway, and Jianming Liang. Representing part-whole hierarchies in foundation models by learning localizability composability and decomposability from anatomy via self supervision. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition, pages 11269–11281, 2024. 2

    work page 2024

  2. [1]

    RSNA pneumonia detection ACE Nodule POPAR Effusion Mass Atelectasis DINO BYOL Cardiomegaly Infiltration Pneumonia Pneumothorax Adam Figure 13

    https://www.kaggle.com/c/rsna-pneumonia- detection- challenge. RSNA pneumonia detection ACE Nodule POPAR Effusion Mass Atelectasis DINO BYOL Cardiomegaly Infiltration Pneumonia Pneumothorax Adam Figure 13. Visualization of Grad-CAM heatmaps. For each column, we provide the heatmap examples for 8 thorax diseases that hold bounding boxes in official labelin...

    work page 2018

  3. [2]

    https : / / www . kaggle . com / competitions / siim-acr-pneumothorax-segmentation . SIIM- ACR Pneumothorax Segmentation (2019). 5, 7, 8, 10

    work page 2019

  4. [3]

    Variance- invariance-covariance regularization for self-supervised learning

    Adrien Bardes, Jean Ponce, and Yann LeCun. Variance- invariance-covariance regularization for self-supervised learning. ICLR, Vicreg, 2022. 2

    work page 2022

  5. [4]

    Vicregl: Self- supervised learning of local visual features

    Adrien Bardes, Jean Ponce, and Yann Lecun. Vicregl: Self- supervised learning of local visual features. In 36th Con- ference on Neural Information Processing Systems (NeurIPS 2022), 2022. 1, 2

    work page 2022

  6. [5]

    Unsupervised learning of visual features by contrasting cluster assignments

    Mathilde Caron, Ishan Misra, Julien Mairal, Priya Goyal, Pi- otr Bojanowski, and Armand Joulin. Unsupervised learning of visual features by contrasting cluster assignments. Ad- vances in neural information processing systems , 33:9912– 9924, 2020. 2

    2020

  7. [6]

    Emerg- ing properties in self-supervised vision transformers

    Mathilde Caron, Hugo Touvron, Ishan Misra, Herv ´e J´egou, Julien Mairal, Piotr Bojanowski, and Armand Joulin. Emerg- ing properties in self-supervised vision transformers. In Pro- ceedings of the IEEE/CVF international conference on com- puter vision, pages 9650–9660, 2021. 1, 5, 6, 8, 11

    work page 2021

  8. [7]

    A simple framework for contrastive learning of visual representations

    Ting Chen, Simon Kornblith, Mohammad Norouzi, and Ge- offrey Hinton. A simple framework for contrastive learning of visual representations. In International conference on ma- chine learning, pages 1597–1607. PMLR, 2020. 1, 2

    work page 2020

Show all 68 references
  1. [8]

    Exploring simple siamese rep- resentation learning

    Xinlei Chen and Kaiming He. Exploring simple siamese rep- resentation learning. In Proceedings of the IEEE/CVF con- ference on computer vision and pattern recognition , pages 15750–15758, 2021. 2

    2021

  2. [9]

    Eyepacs: an adapt- able telemedicine system for diabetic retinopathy screening

    Jorge Cuadros and George Bresnick. Eyepacs: an adapt- able telemedicine system for diabetic retinopathy screening. Journal of diabetes science and technology , 3(3):509–516,

  3. [10]

    An image is worth 16x16 words: Trans- formers for image recognition at scale

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Syl- vain Gelly, et al. An image is worth 16x16 words: Trans- formers for image recognition at scale. arXiv preprint a...

    2010 arXiv

  4. [11]

    Anatomy-aware con- trastive representation learning for fetal ultrasound

    Zeyu Fu, Jianbo Jiao, Robail Yasrab, Lior Drukker, Aris T Papageorghiou, and J Alison Noble. Anatomy-aware con- trastive representation learning for fetal ultrasound. In Eu- ropean Conference on Computer Vision , pages 422–436. Springer, 2022. 2

    2022

  5. [12]

    Bootstrap your own latent-a new approach to self-supervised learning

    Jean-Bastien Grill, Florian Strub, Florent Altch ´e, Corentin Tallec, Pierre Richemond, Elena Buchatskaya, Carl Doersch, Bernardo Avila Pires, Zhaohan Guo, Mohammad Ghesh- laghi Azar, et al. Bootstrap your own latent-a new approach to self-supervised learning. Advances in neur...

    2020

  6. [13]

    Discriminative, restorative, and adversarial learning: Stepwise incremental pretraining

    Zuwei Guo, Nahid UI Islam, Michael B Gotway, and Jian- ming Liang. Discriminative, restorative, and adversarial learning: Stepwise incremental pretraining. In MICCAI Workshop on Domain Adaptation and Representation Trans- fer, pages 66–76. Springer, 2022. 2

    2022

  7. [14]

    Stepwise incremental pretraining for integrating discriminative, restorative, and adversarial learning.Medical Image Analysis, 95:103159, 2024

    Zuwei Guo, Nahid Ul Islam, Michael B Gotway, and Jian- ming Liang. Stepwise incremental pretraining for integrating discriminative, restorative, and adversarial learning.Medical Image Analysis, 95:103159, 2024. 2

    2024

  8. [15]

    Learning semantics-enriched representation via self- discovery, self-classification, and self-restoration

    Fatemeh Haghighi, Mohammad Reza Hosseinzadeh Taher, Zongwei Zhou, Michael B Gotway, and Jianming Liang. Learning semantics-enriched representation via self- discovery, self-classification, and self-restoration. In Medical Image Computing and Computer Assisted Intervention–MICC...

    2020

  9. [16]

    Dira: Discrimina- tive, restorative, and adversarial learning for self-supervised medical image analysis

    Fatemeh Haghighi, Mohammad Reza Hosseinzadeh Taher, Michael B Gotway, and Jianming Liang. Dira: Discrimina- tive, restorative, and adversarial learning for self-supervised medical image analysis. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognit...

    2022

  10. [17]

    Self-supervised learning for medical image analysis: Discriminative, restora- tive, or adversarial? Medical Image Analysis , 94:103086,

    Fatemeh Haghighi, Mohammad Reza Hosseinzadeh Taher, Michael B Gotway, and Jianming Liang. Self-supervised learning for medical image analysis: Discriminative, restora- tive, or adversarial? Medical Image Analysis , 94:103086,

  11. [18]

    Transferable visual words: Exploiting the semantics of anatomical patterns for self-supervised learning.IEEE trans- actions on medical imaging, 40(10):2857–2868, 2021

    Fatemeh Haghighi, Mohammad Reza Hosseinzadeh Taher, Zongwei Zhou, Michael B Gotway, and Jianming Liang. Transferable visual words: Exploiting the semantics of anatomical patterns for self-supervised learning.IEEE trans- actions on medical imaging, 40(10):2857–2868, 2021. 2

    2021

  12. [19]

    Momentum contrast for unsupervised visual rep- resentation learning

    Kaiming He, Haoqi Fan, Yuxin Wu, Saining Xie, and Ross Girshick. Momentum contrast for unsupervised visual rep- resentation learning. In Proceedings of the IEEE/CVF con- ference on computer vision and pattern recognition , pages 9729–9738, 2020. 1, 2

    2020

  13. [20]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceed- ings of the IEEE conference on computer vision and pattern recognition, pages 770–778, 2016. 5

    2016

  14. [21]

    Fire: fundus image registration dataset

    Carlos Hernandez-Matas, Xenophon Zabulis, Areti Tri- antafyllou, Panagiota Anyfanti, Stella Douma, and Anto- nis A Argyros. Fire: fundus image registration dataset. Mod- eling and Artificial Intelligence in Ophthalmology, 1(4):16– 28, 2017. 8, 10

    2017

  15. [22]

    How to represent part-whole hierarchies in a neural network

    Geoffrey Hinton. How to represent part-whole hierarchies in a neural network. Neural Computation, 35(3):413–452,

  16. [23]

    Distill- ing the knowledge in a neural network

    Geoffrey Hinton, Oriol Vinyals, and Jeff Dean. Distill- ing the knowledge in a neural network. arXiv preprint arXiv:1503.02531, 2015. 10

    2015 arXiv

  17. [24]

    Towards foundation models learned from anatomy in medical imaging via self-supervision

    Mohammad Reza Hosseinzadeh Taher, Michael B Gotway, and Jianming Liang. Towards foundation models learned from anatomy in medical imaging via self-supervision. In MICCAI Workshop on Domain Adaptation and Representa- tion Transfer, pages 94–104. Springer, 2023. 2, 5, 8, 11

    2023

  18. [25]

    Anatomy-aware self-supervised learn- ing for aligned multi-modal medical data

    Hongyu Hu, Tiancheng Lin, Yuanfan Guo, Chunxiao Li, Rong Wu, and Yi Xu. Anatomy-aware self-supervised learn- ing for aligned multi-modal medical data. In BMVC, page 877, 2022. 2

    2022

  19. [26]

    Self- supervised learning with local contrastive loss for detec- tion and semantic segmentation

    Ashraful Islam, Benjamin Lundell, Harpreet Sawhney, Sudipta N Sinha, Peter Morales, and Richard J Radke. Self- supervised learning with local contrastive loss for detec- tion and semantic segmentation. In Proceedings of the IEEE/CVF Winter Conference on Applications of Compute...

    2023

  20. [27]

    Two public chest x-ray datasets for computer-aided screening of pulmonary diseases

    Stefan Jaeger, Sema Candemir, Sameer Antani, Y `ı-Xi´ang J W´ang, Pu-Xuan Lu, and George Thoma. Two public chest x-ray datasets for computer-aided screening of pulmonary diseases. Quantitative imaging in medicine and surgery , 4(6):475, 2014. 5, 9

    2014

  21. [28]

    Anatomical invariance modeling and semantic alignment for self-supervised learning in 3d medical image analysis

    Yankai Jiang, Mingze Sun, Heng Guo, Xiaoyu Bai, Ke Yan, Le Lu, and Minfeng Xu. Anatomical invariance modeling and semantic alignment for self-supervised learning in 3d medical image analysis. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 158...

    2023

  22. [29]

    Self-supervised visual fea- ture learning with deep neural networks: A survey

    Longlong Jing and Yingli Tian. Self-supervised visual fea- ture learning with deep neural networks: A survey. IEEE transactions on pattern analysis and machine intelligence , 43(11):4037–4058, 2020. 1

    2020

  23. [30]

    Mimic-cxr, a de- identified publicly available database of chest radiographs with free-text reports

    Alistair EW Johnson, Tom J Pollard, Seth J Berkowitz, Nathaniel R Greenbaum, Matthew P Lungren, Chih-ying Deng, Roger G Mark, and Steven Horng. Mimic-cxr, a de- identified publicly available database of chest radiographs with free-text reports. Scientific data, 6(1):317, 2019. 5

    2019

  24. [31]

    Contrastive learning of global and local features for medical image segmentation with limited annotations

    KCEEN Karani and Ender Konukoglu. Contrastive learning of global and local features for medical image segmentation with limited annotations. Advances in Neural Information Processing Systems, 33, 2020. 2

    2020

  25. [32]

    Compressive visual representa- tions

    Kuang-Huei Lee, Anurag Arnab, Sergio Guadarrama, John Canny, and Ian Fischer. Compressive visual representa- tions. Advances in Neural Information Processing Systems , 34:19538–19552, 2021. 2

    2021

  26. [33]

    A structure-aware relation network for thoracic diseases detection and segmentation

    Jie Lian, Jingyu Liu, Shu Zhang, Kai Gao, Xiaoqing Liu, Dingwen Zhang, and Yizhou Yu. A structure-aware relation network for thoracic diseases detection and segmentation. IEEE Transactions on Medical Imaging , 40(8):2042–2052,

  27. [34]

    Chestx-det10: chest x-ray dataset on detection of thoracic abnormalities

    Jingyu Liu, Jie Lian, and Yizhou Yu. Chestx-det10: chest x-ray dataset on detection of thoracic abnormalities. arXiv preprint arXiv:2006.10550, 2020. 9

    2006 arXiv

  28. [35]

    Swin transformer: Hierarchical vision transformer using shifted windows

    Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo. Swin transformer: Hierarchical vision transformer using shifted windows. In Proceedings of the IEEE/CVF international conference on computer vision, pages 10012–10022, 2021. 5, 10

    2021

  29. [36]

    Devide: Faceted medical knowledge for improved medical vision-language pre-training

    Haozhe Luo, Ziyu Zhou, Corentin Royer, Anjany Sekuboy- ina, and Bjoern Menze. Devide: Faceted medical knowledge for improved medical vision-language pre-training. arXiv preprint arXiv:2404.03618, 2024. 5, 8

    2024 arXiv

  30. [37]

    Benchmarking and boosting transformers for medical image classification

    DongAo Ma, Mohammad Reza Hosseinzadeh Taher, Jiax- uan Pang, Nahid UI Islam, Fatemeh Haghighi, Michael B Gotway, and Jianming Liang. Benchmarking and boosting transformers for medical image classification. In MICCAI Workshop on Domain Adaptation and Representation Trans- fer, ...

    2022

  31. [38]

    Foundation ark: Accruing and reusing knowl- edge for superior and robust performance

    DongAo Ma, Jiaxuan Pang, Michael B Gotway, and Jian- ming Liang. Foundation ark: Accruing and reusing knowl- edge for superior and robust performance. In International Conference on Medical Image Computing and Computer- Assisted Intervention, pages 651–662. Springer, 2023. 7

    2023

  32. [39]

    Lvm-med: Learning large- scale self-supervised vision models for medical imaging via second-order graph matching

    Duy MH Nguyen, Hoang Nguyen, Nghiem Diep, Tan Ngoc Pham, Tri Cao, Binh Nguyen, Paul Swoboda, Nhat Ho, Shadi Albarqouni, Pengtao Xie, et al. Lvm-med: Learning large- scale self-supervised vision models for medical imaging via second-order graph matching. Advances in Neural Info...

    2024

  33. [40]

    Popar: Patch order prediction and appearance recovery for self-supervised medical image anal- ysis

    Jiaxuan Pang, Fatemeh Haghighi, DongAo Ma, Nahid Ul Is- lam, Mohammad Reza Hosseinzadeh Taher, Michael B Got- way, and Jianming Liang. Popar: Patch order prediction and appearance recovery for self-supervised medical image anal- ysis. In MICCAI Workshop on Domain Adaptation an...

    2022

  34. [41]

    Asa: Learning anatomical con- sistency, sub-volume spatial relationships and fine-grained appearance for ct images

    Jiaxuan Pang, DongAo Ma, Ziyu Zhou, Michael B Got- way, and Jianming Liang. Asa: Learning anatomical con- sistency, sub-volume spatial relationships and fine-grained appearance for ct images. In International Conference on Medical Image Computing and Computer-Assisted Interven...

    2024

  35. [42]

    Learning transferable visual models from natural language supervi- sion

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervi- sion. In International conference on machine learning, ...

    2021

  36. [43]

    Grad-cam: Visual explanations from deep networks via gradient-based localization

    Ramprasaath R Selvaraju, Michael Cogswell, Abhishek Das, Ramakrishna Vedantam, Devi Parikh, and Dhruv Batra. Grad-cam: Visual explanations from deep networks via gradient-based localization. In Proceedings of the IEEE in- ternational conference on computer vision , pages 618–626,

  37. [44]

    Junji Shiraishi, Shigehiko Katsuragawa, Junpei Ikezoe, Tsu- neo Matsumoto, Takeshi Kobayashi, Ken-ichi Komatsu, Mitate Matsui, Hiroshi Fujita, Yoshie Kodera, and Kunio Doi. Development of a digital image database for chest radiographs with and without a lung nodule: receiver o...

    2000

  38. [45]

    Deep high-resolution representation learning for human pose es- timation

    Ke Sun, Bin Xiao, Dong Liu, and Jingdong Wang. Deep high-resolution representation learning for human pose es- timation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 5693–5703,

  39. [47]

    Guest editorial annotation-efficient deep learning: the holy grail of medical imaging

    Nima Tajbakhsh, Holger Roth, Demetri Terzopoulos, and Jianming Liang. Guest editorial annotation-efficient deep learning: the holy grail of medical imaging. IEEE trans- actions on medical imaging, 40(10):2526–2533, 2021. 1

    2021

  40. [48]

    Multimodal self-supervised learning for medical im- age analysis

    Aiham Taleb, Christoph Lippert, Tassilo Klein, and Moin Nabi. Multimodal self-supervised learning for medical im- age analysis. In International conference on information pro- cessing in medical imaging, pages 661–673. Springer, 2021. 1

    2021

  41. [49]

    Con- trastive multiview coding

    Yonglong Tian, Dilip Krishnan, and Phillip Isola. Con- trastive multiview coding. In Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part XI 16 , pages 776–794. Springer,

    2020

  42. [50]

    Expert-level detec- tion of pathologies from unannotated chest x-ray images via self-supervised learning

    Ekin Tiu, Ellie Talius, Pujan Patel, Curtis P Langlotz, An- drew Y Ng, and Pranav Rajpurkar. Expert-level detec- tion of pathologies from unannotated chest x-ray images via self-supervised learning. Nature Biomedical Engineering , 6(12):1399–1406, 2022. 5, 8

    2022

  43. [51]

    Visualiz- ing data using t-sne

    Laurens Van der Maaten and Geoffrey Hinton. Visualiz- ing data using t-sne. Journal of machine learning research, 9(11), 2008. 6

    2008

  44. [52]

    Droppos: Pre-training vision transformers by reconstructing dropped positions

    Haochen Wang, Junsong Fan, Yuxi Wang, Kaiyou Song, Tong Wang, and Zhaoxiang Zhang. Droppos: Pre-training vision transformers by reconstructing dropped positions. arXiv preprint arXiv:2309.03576, 2023. 5, 6, 8

    2023 arXiv

  45. [53]

    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, Zhiyong Lu, Mo- hammadhadi Bagheri, and Ronald M Summers. Chestx- ray8: Hospital-scale chest x-ray database and benchmarks on weakly-supervised classification and localization of common thorax diseases. In Proceedings of the IEEE conference on...

  46. [54]

    Dense contrastive learning for self-supervised visual pre-training

    Xinlong Wang, Rufeng Zhang, Chunhua Shen, Tao Kong, and Lei Li. Dense contrastive learning for self-supervised visual pre-training. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 3024–3033, 2021. 2

    2021

  47. [55]

    Unified perceptual parsing for scene understand- ing

    Tete Xiao, Yingcheng Liu, Bolei Zhou, Yuning Jiang, and Jian Sun. Unified perceptual parsing for scene understand- ing. In Proceedings of the European conference on computer vision (ECCV), pages 418–434, 2018. 5, 10

    2018

  48. [56]

    Region similarity representation learn- ing

    Tete Xiao, Colorado J Reed, Xiaolong Wang, Kurt Keutzer, and Trevor Darrell. Region similarity representation learn- ing. In Proceedings of the IEEE/CVF International Confer- ence on Computer Vision, pages 10539–10548, 2021. 2

    2021

  49. [57]

    Propagate yourself: Exploring pixel-level consistency for unsupervised visual representation learning

    Zhenda Xie, Yutong Lin, Zheng Zhang, Yue Cao, Stephen Lin, and Han Hu. Propagate yourself: Exploring pixel-level consistency for unsupervised visual representation learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 16684–16693...

    2021

  50. [58]

    Sam: Self-supervised learning of pixel-wise anatom- ical embeddings in radiological images

    Ke Yan, Jinzheng Cai, Dakai Jin, Shun Miao, Dazhou Guo, Adam P Harrison, Youbao Tang, Jing Xiao, Jingjing Lu, and Le Lu. Sam: Self-supervised learning of pixel-wise anatom- ical embeddings in radiological images. IEEE Transactions on Medical Imaging, 41(10):2658–2669, 2022. 2

    2022

  51. [59]

    Densedino: Boosting dense self-supervised learning with token-based point-level consistency

    Yike Yuan, Xinghe Fu, Yunlong Yu, and Xi Li. Densedino: Boosting dense self-supervised learning with token-based point-level consistency. arXiv preprint arXiv:2306.04654 ,

    arXiv

  52. [60]

    Patch-level representation learning for self-supervised vision transformers

    Sukmin Yun, Hankook Lee, Jaehyung Kim, and Jinwoo Shin. Patch-level representation learning for self-supervised vision transformers. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 8354–8363, 2022. 2, 5, 8

    2022

  53. [61]

    Barlow twins: Self-supervised learning via redundancy reduction

    Jure Zbontar, Li Jing, Ishan Misra, Yann LeCun, and St´ephane Deny. Barlow twins: Self-supervised learning via redundancy reduction. In International Conference on Ma- chine Learning, pages 12310–12320. PMLR, 2021. 2

    2021

  54. [62]

    How does sim- siam avoid collapse without negative samples? a unified un- derstanding with self-supervised contrastive learning

    Chaoning Zhang, Kang Zhang, Chenshuang Zhang, Trung X Pham, Chang D Yoo, and In So Kweon. How does sim- siam avoid collapse without negative samples? a unified un- derstanding with self-supervised contrastive learning. arXiv preprint arXiv:2203.16262, 2022. 2

    2022 arXiv

  55. [63]

    Pre- cise location matching improves dense contrastive learning in digital pathology

    Jingwei Zhang, Saarthak Kapse, Ke Ma, Prateek Prasanna, Maria Vakalopoulou, Joel Saltz, and Dimitris Samaras. Pre- cise location matching improves dense contrastive learning in digital pathology. In International Conference on In- formation Processing in Medical Imaging , page...

    2023

  56. [64]

    Leverage your local and global repre- sentations: A new self-supervised learning strategy

    Tong Zhang, Congpei Qiu, Wei Ke, Sabine S ¨usstrunk, and Mathieu Salzmann. Leverage your local and global repre- sentations: A new self-supervised learning strategy. In Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 16580–16589, 2022. 1

    2022

  57. [65]

    Knowledge-enhanced visual-language pre- training on chest radiology images.Nature Communications, 14(1):4542, 2023

    Xiaoman Zhang, Chaoyi Wu, Ya Zhang, Weidi Xie, and Yanfeng Wang. Knowledge-enhanced visual-language pre- training on chest radiology images.Nature Communications, 14(1):4542, 2023. 5, 8

    2023

  58. [66]

    Learning anatom- ically consistent embedding for chest radiography

    Ziyu Zhou, Haozhe Luo, Jiaxuan Pang, Xiaowei Ding, Michael Gotway, and Jianming Liang. Learning anatom- ically consistent embedding for chest radiography. In BMVC: proceedings of the British Machine Vision Confer- ence. British Machine Vision Conference, volume 2023. NIH Publi...

    2023

  59. [67]

    Fine-tuning convo- lutional neural networks for biomedical image analysis: ac- tively and incrementally

    Zongwei Zhou, Jae Shin, Lei Zhang, Suryakanth Gurudu, Michael Gotway, and Jianming Liang. Fine-tuning convo- lutional neural networks for biomedical image analysis: ac- tively and incrementally. In Proceedings of the IEEE con- ference on computer vision and pattern recognition...

    2017

  60. [68]

    Models genesis: Generic autodidactic models for 3d medical image analysis

    Zongwei Zhou, Vatsal Sodha, Md Mahfuzur Rahman Sid- diquee, Ruibin Feng, Nima Tajbakhsh, Michael B Gotway, and Jianming Liang. Models genesis: Generic autodidactic models for 3d medical image analysis. In Medical Image Computing and Computer Assisted Intervention–MICCAI 2019: ...

    2019

Pith tools

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