Pith. sign in

REVIEW 4 major objections 6 minor 88 references

Vector Contrastive Learning For Pixel-Wise Pretraining In Medical Vision

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Contrastive learning for pixel-wise medical pretraining can be reformulated as vector regression, and the resulting COVER framework outperforms 17 baselines across 8 downstream tasks.

desk verdict A broad, reproducible pixel-wise pretraining recipe whose empirical gains are plausible but whose motivating 'quantitative dispersion' story is not supported by the objective; worth reviewing, but the theory needs a major revision or removal. read the letter →

arxiv 2506.20850 v1 pith:2QT7HYWJ submitted 2025-06-25 cs.CV

Yuting He , Shuo Li This is my paper · ORCID
classification cs.CV
keywords vectorcontrastivelearningpixel-wiseself-supervisedpretrainingdisplacementfieldover-dispersiondistancemodelingmedicalimagesegmentationvisionfoundationmodels
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

This paper takes on a known failure mode of contrastive self-supervised learning when it is applied pixel by pixel to medical images: binary contrastive objectives push features apart without controlling how far, so nearby pixels that belong to the same anatomical structure end up over-dispersed and their correlations break. The proposed remedy is to reformulate contrastive learning as a vector regression problem: instead of maximizing and minimizing feature distances directly, the model predicts a dense displacement vector field between two randomly transformed views of the same image, with the known spatial transformation supplying free ground-truth vectors. The paper argues that regressing these vectors implicitly models feature distances, quantifying dispersion while preserving local correlations. It implements this idea in the COVER framework and reports that pretraining in this way beats 17 comparison methods across 8 downstream medical tasks in 2D and 3D. If the claim holds, pixel-wise self-supervised pretraining no longer needs binary contrast at all and can be driven entirely by geometric self-supervision.

What carries the argument

The load-bearing object is the displacement vector field (DVF): a dense field whose vector at each pixel points from its position in one augmented view to the corresponding position in the other view, generated for free by a random affine spatial transformation. Inside the mapping function $V$, the vector embedding unit (VEU) computes a normalized similarity (scaled dot-product attention) between a center feature and features in an $N\times N$ receptive field, then multiplies that distance map by a fixed vector template matrix $V^{N\times N}$ whose entries are the coordinate offsets from the center; the weighted sum is the predicted displacement vector. Multi-vector integration (MVI) splits features into groups and averages the resulting vectors to handle ambiguous correspondences, and vector pyramid aggregation (VPA) chains these units across feature scales so the final DVF fuses global and local correspondences. The theoretical work of the key machinery is the equivalence chain showing that if the true DVF vector is representable as a weighted sum of template vectors with weights equal to normalized feature similarities, then minimizing the vector regression loss is equivalent to constraining the aggregated feature distances.

What would settle it

Observe the correlation between learned feature distances and true spatial displacements on transformations outside the template's range, such as affine shifts larger than the template size or elastic deformations. If the Pearson correlation stays high, the distance-modeling equivalence generalizes; if it drops while the vector regression loss remains low, the model is fitting vectors without actually modeling distances, falsifying the central claim.

Watch

Extended reading notes

Core claim

The central claim is that standard binary contrastive learning, which treats every pixel pair as either positive or negative, is the wrong objective for pixel-wise pretraining because its max-min dispersion breaks the continuous feature correlations that medical images rely on. The paper defines vector contrastive learning (vector CL): a mapping function $V$ takes the dense features of two views and predicts the displacement vector field $\psi'_{ab}$ that would align them, and the pretraining loss is the regression error $|\psi_{ab} - \psi'_{ab}|$ against the ground-truth DVF generated by a random affine spatial transformation. Because the DVF encodes real coordinate displacements, the mapping from feature distances to vectors is trained to be consistent with geometric correspondence; the paper shows an optimization-flow chain from distance modeling $|\alpha_i - d'_i|$ to vector regression $|v - \sum_i V_i d'_i|$, and argues this yields quantitatively controlled dispersion. The COVER architecture instantiates this with three components: self-vector regression (SeVR) for annotation-free ground truth, a mixture of vectors (MoV) with a vector template and multi-vector integration to handle spatial continuity and correspondence ambiguity, and vector pyramid aggregation (VPA) to capture multiscale correspondences cheaply. The paper reports that this pretraining outperforms 17 baselines on 8 tasks spanning 2D and 3D and X-ray, fundus, MR, and CT, while also improving convergence, data efficiency, and feature clustering.

Load-bearing premise

The whole distance-modeling argument rests on the assumption that the true displacement vector at a pixel can be expressed as a weighted sum of the fixed template vectors, with weights equal to the normalized feature similarities; if a transformation produces a displacement that the template cannot represent, the vector-regression loss can shrink without ever constraining the embedding distances to carry the intended semantics.

Editorial extensions

If this is right

  • Pixel-wise self-supervised pretraining can be built entirely from random geometric transformations, with no paired images or annotations, so the same recipe applies to any medical modality that can be spatially transformed.
  • Because dispersion is quantified rather than binary, downstream segmentation and classification tasks should inherit feature correlations that survive scenes inconsistent with the pretraining data.
  • The same pretrained features support smaller fine-tuning budgets: the paper reports strong downstream performance with fewer fine-tuned parameters, less fine-tuning data, and fewer pretraining iterations than the compared methods.
  • No separate distance-labeling stage is needed; the ground-truth distances are effectively supplied by the DVF, which sidesteps the task-dependent-distance problem that blocks direct distance modeling.

Reading between the lines

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

  • Beyond the paper: if the vector-regression-to-distance equivalence is as tight as claimed, the same DVF-supervision recipe should transfer to non-medical dense tasks, for example pretraining on video or multi-view images where geometric transformations are known.
  • Beyond the paper: the fixed vector template could be replaced by a learnable or multi-resolution template; the paper's own ablation shows that a 7x7 template works best, which hints that the template size is a meaningful inductive bias rather than a free parameter.
  • Beyond the paper: a direct test of the central equivalence would be to probe the learned embedding with transformations outside the affine family, such as nonlinear deformations; if the model still quantifies distances correctly, the mechanism generalizes, and if not, the DVF supervision is learning affine-specific shortcuts.
  • Beyond the paper: because the method only needs a random spatial transformation, it could be pretrained jointly on heterogeneous medical datasets with shared weights, moving toward a single model that serves many organs without requiring aligned data.
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 a new self-supervised pretraining paradigm for pixel-wise medical representation learning, called vector contrastive learning (VCL), and instantiates it in the COVER framework. COVER generates two views of an image via appearance and spatial affine transformations, uses the known displacement vector field (DVF) as a free regression target, and predicts the DVF from dense feature distances through a vector embedding unit (VEU), a multi-vector integration (MVI), and a vector pyramid aggregation (VPA) architecture. The central theoretical claim is that regressing these displacement vectors is equivalent to modeling pixel-wise feature distances and therefore quantifies feature dispersion, avoiding the alleged over-dispersion of binary contrastive learning. The empirical section reports pretraining on 2D chest X-ray and 3D brain MRI, fine-tuning on 8 tasks spanning 4 modalities, and comparisons with 17 baselines, together with ablations, reliability analysis, and cross-architecture compatibility experiments.

Significance. If the central equivalence claim were correct, COVER would be a notable contribution: it would offer a principled, annotation-free way to turn distance modeling into vector regression, with a concrete mechanism for preserving pixel-wise feature correlations and a tighter generalization bound. The empirical protocol is genuinely broad and is a strength of the paper: 8 downstream tasks, both 2D and 3D settings, 4 imaging modalities, 17 baselines, component and hyperparameter ablations, and a public code link. The reported gains are consistent across tasks, and the method is simple enough to reproduce. However, the theoretical justification in Sec. 3.1.2 and in Supplementary A.2 is not valid as stated: the loss constrains only a low-dimensional centroid of the distance map, not the per-pair distances or their dispersion. The paper itself concedes (Sec. 3.1.2) that 'vector regression is not strictly equivalent to conventional distance modeling,' which is in tension with the abstract and Sec. 3.1's stronger claims. The empirical value of COVER may still stand, but the paper currently over-interprets its mechanism, and the novel theoretical framing needs substantial rework.

major comments (4)
  1. [Sec. 3.1.2, Eqs. (2)-(5)] The claimed equivalence chain is not a valid derivation. Equation (2) asks every per-pair distance error |α_i - d'_i| to vanish; Eq. (3) asks the weighted norm of those errors to vanish. If the vectors V_i are not linearly independent (they are not in the VEU, because N×N > 2 in 2D), convergence of the weighted sum in Eq. (4) does not imply convergence of the individual terms in Eq. (3), so the double arrow between Eqs. (2) and (3) is false. Equation (4) itself only gives convergence of a weighted average, which is the same as Eq. (5). Thus the chain establishes only that the regressed vector approaches the ground truth vector v, not that feature distances approach the hypothesized ground truth distances α_i. The paper's own caveat, 'vector regression is not strictly equivalent to conventional distance modeling,' correctly identifies this gap, but that caveat undercuts the paper's central claim as stated in Sec. 3.1 and the abstract.
  2. [Sec. 3.3, Eq. (7)] The VEU output is a convex combination of the fixed template vectors: v'_i = Σ_j softmax_j(f)·V_j with softmax weights summing to 1. Consequently L_vec in Eq. (6) imposes at most two scalar constraints (for 2D) on the N×N distance map. For any ground-truth vector v in the convex hull of the template, there is a high-dimensional family of softmax/distance maps with exactly the same predicted vector, including one-hot maps (all mass on a single neighbor) and broad maps (mass spread over many neighbors). This means L_vec cannot quantify dispersion: a one-hot distance map and a broad map with the same centroid are indistinguishable to the loss. The claim in Sec. 3.1.1 that vector regression 'quantifies their dispersion degrees' is therefore not supported by the implemented objective. The authors need to either add an explicit concentration/spread penalty on the distance map or substantially revise the dispersion-quantification claims.
  3. [Supplementary A.2, Eqs. (21)-(24)] The theory assumes that the ground-truth displacement vector v can be written as v = Σ_j α_j V_j with α_j ≥ 0 and Σ_j α_j = 1, and then derives δ_VCL ≤ τ log(1/α_min) from α_j ∈ [α_min, 1]. Nothing in L_vec enforces α_j ≥ α_min > 0, nor does anything prevent the model from collapsing the softmax weights onto a single template vector. In the common situation where the ground-truth DVF equals a template offset (e.g., pure translation by one pixel, or a scaled version of a template vector), the loss is minimized by putting almost all softmax mass on that single offset, which reproduces exactly the binary over-dispersion the paper says it avoids. The bound in Eq. (24) and the tighter generalization bound in Eq. (25) therefore do not follow from the objective. If the authors want to keep this theory, they must show that the optimization dynamics or an explicit regularizer keep the weights bounded away from the simplex vertices.
  4. [Sec. 4.2.1 and Sec. 4.3.6] The empirical sections attribute the observed gains to 'distance modeling' and 'quantitative dispersion' (e.g., the component ablation and the t-SNE analysis). Because L_vec only constrains the centroid of the distance map, these attributions are not established by the current experiments. The reported improvements may be real and could result from the affine-prediction pretraining task, the multi-scale architecture, or the consistency loss, rather than from dispersion modeling. The authors should add a diagnostic that directly measures the dispersion of the predicted distance maps (e.g., entropy of the softmax distribution over the receptive field) under L_vec and under the compared baselines, or rephrase the claims to describe the method as centroid-based vector regression with multiscale fusion. Without such evidence, the mechanism-related conclusions in Secs. 4.2.1 and 4.3.6 are not supported by the objective.
minor comments (6)
  1. [Sec. 3.1.2, Eqs. (2)-(5)] The notation is garbled: 'PI i=0' appears to be an incomplete sum, and the limits and indexing are unclear (i = 1..I or i = 0..I). Please rewrite the chain with standard sum notation and define V_i and d'_i explicitly before the equations.
  2. [Sec. 4.1.1] The pretraining dataset is called 'ChestX-ray6' here but 'ChestX-ray8' in the datasets table (Table 2) and elsewhere. This inconsistency should be fixed.
  3. [Sec. 3.4, Eq. (8)] The fusion operation J is described qualitatively in the main text and formally only in Supplementary B.4. The main text should at least define J when it is first used, since Eq. (8) is the core architecture equation.
  4. [Table 1] The column 'A VG' is not defined in the caption or text; if it is the average rank, say so explicitly. Also, the numbers in the COVER row (e.g., 63.6 without a '+' sign for FeTA21) should be formatted consistently.
  5. [Supplementary D.5] The text references 'Fig.8 in the manuscript' when referring to the multi-scale vector visualization, but the main manuscript's Fig. 8 is the vector template diagram. The cross-references between main text and supplementary figures need to be corrected.
  6. [Throughout] There are several grammatical and typographical errors, e.g., 'putted', 'the the', and '2×10 5 iterations'. A careful proofreading pass is needed.

Circularity Check

2 steps flagged · score 6.0 of 10

Central 'distance modeling' step is wired in by assuming v decomposes over the same vector template used by the predictor; the dispersion bound assumes the spread it claims to prove.

  1. self definitional [Section 3.1.2, Eq. (5), and Supplementary A.2, Eq. (21)]
    "Therefore, our COVER hypothesizes that existing a value v (i.e., vector in DVF ψab) can make Σ_i V_i α_i = v, the Equ.5 will be derived from Equ.4, formulating distance modeling in vector regression. ... We assume that the v can be modeled as v=Σ_j α_j V_j (α_j ≥0, Σ_j α_j =1)."

    The formal bridge from distance modeling to vector regression is obtained by assuming the ground-truth vector v decomposes in the same vector template V as the predictor's weighted sum (Eq. 7: v'_i = softmax(f_i f_b^T/τ) V). Under this assumption L_vec = |Σ_j V_j(α_j − d'_j)|, which is a weighted aggregate of distance errors, not a constraint on individual distances. The paper then interprets vector regression as 'formulating distance modeling', but the α_j plays no independent role: it is defined as a coefficient of v in the template basis. The claimed distance/dispersion content is thus stipulated by the decomposition, not derived from the loss; the paper concedes 'vector regression is not strictly equivalent to conventional distance modeling'.

  2. other [Supplementary A.2, Eqs. (23)-(24)]
    "Owing to the α_j, α_k ∈ [α_min,1] (α_min > 0), when α_j = 1, α(k) = α_min the δ_VCL will be: δ_VCL ≤ τ log 1/α_min. ... this situation where one weight is exactly 1 and all others are zero is degenerate (and would essentially revert to binary CL), so typically the effective δ_VCL is much smaller than the worst-case bound."

    The claimed tighter generalization bound rests on assuming α_j ≥ α_min > 0, i.e., that the softmax/attention weights are spread across several template vectors. That is exactly the non-over-dispersed property the paper sets out to prove: nothing in L_vec (Eq. 6/19) prevents the one-hot minimizer α_min=0, and for any ground-truth vector inside the template hull the loss can be minimized by collapsing attention to one offset. The premise of the bound is therefore the conclusion it is used to justify.

full rationale

The empirical comparison against 17 baselines on 8 tasks uses independent downstream labels and is not circular. Self-citations to GVSL [26] and GEMINI [29] are used as baselines and context, not as load-bearing justification for the central derivation. The circularity is confined to the theoretical conversion from vector regression to distance/dispersion modeling: Eq. (5) is made to follow from Eq. (4) only by positing v=ΣV_iα_i, and the supplement's bound assumes the weight spread that the loss never enforces. The vector-regression pretext itself has an external target (the affine DVF), so the method is not circular in the self-generated-label sense; however, the paper's central claim of dispersion quantification is partly definitional and partly assumes the conclusion it is supposed to prove. The paper's own limitation statement that 'vector regression is not strictly equivalent' to distance modeling supports scoring this as partial, rather than total, circularity.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claim rests on the geometric augmentation producing meaningful DVF targets, on the convex-decomposition assumption in Eq. 21 that links displacement vectors to embedding distances, and on the Rademacher generalization framework. Two new hyperparameters (N and J) are tuned on a downstream benchmark, and the affine transform ranges that create the ground truth vectors are hand-chosen. No new physical entities are introduced.

free parameters (3)
  • Receptive field size N in VEU = 7x7 (best among 3x3, 5x5, 7x7, 9x9)
    Ablated on SCR 25% linear evaluation in Sec 4.3.2; the central vector regression mechanism depends on this window size.
  • Number of VEU groups J per scale level = [4,4,4,1,1] (best among listed settings)
    Ablated on SCR 25% linear evaluation in Sec 4.3.2; controls multi-vector integration and the ambiguity modeling.
  • Affine transformation ranges for T_sp = translation in [-0.2, 0.2], rotation in [-pi/9, pi/9], shear in [-pi/32, pi/32], scale in [0.5, 1.5]
    Hand-chosen in Supplementary B.2 with no ablation; these ranges generate the ground truth displacement vector fields that define the pretraining target.
assumptions (4)
  • ad hoc to paper Every ground truth displacement vector v is representable as a convex combination of fixed template vectors with nonnegative coefficients alpha_j summing to 1 (Eq. 21 in Supplementary A.2).
    This decomposition is needed to equate vector regression with distance modeling. It is asserted, not derived, and is not guaranteed for arbitrary affine displacements.
  • domain assumption Pixel-wise features obey a local smoothness constraint with bounded local variation delta (Eq. 10), and the effective local Rademacher complexity scales as (delta/Delta) R_n(F) (Eq. 13).
    The improved generalization bound for vector CL depends on this reduction of effective complexity, which is stated without a proof specific to the proposed framework.
  • domain assumption Random affine transformations of medical images produce displacement fields that are semantically meaningful supervision for pixel-wise representation learning.
    SeVR (Sec 3.2) relies on this to make the self-generated DVF a useful pretraining signal for downstream medical tasks.
  • standard math Rademacher complexity generalization bounds of Mohri et al. apply to the considered pixel-wise feature hypothesis class.
    Used in Supplementary A.1 to formalize the over-dispersion and vector CL bounds.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Vector Contrastive Learning For Pixel-Wise Pretraining In Medical Vision." pith.science (2026). https://pith.science/paper/2QT7HYWJ

@misc{pith2026250620850,
  author       = {Pith},
  title        = {Pith review of: Vector Contrastive Learning For Pixel-Wise Pretraining In Medical Vision},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2QT7HYWJ}},
  note         = {Machine review of arXiv:2506.20850}
}
read the original abstract

Contrastive learning (CL) has become a cornerstone of self-supervised pretraining (SSP) in foundation models, however, extending CL to pixel-wise representation, crucial for medical vision, remains an open problem. Standard CL formulates SSP as a binary optimization problem (binary CL) where the excessive pursuit of feature dispersion leads to an over-dispersion problem, breaking pixel-wise feature correlation thus disrupting the intra-class distribution. Our vector CL reformulates CL as a vector regression problem, enabling dispersion quantification in pixel-wise pretraining via modeling feature distances in regressing displacement vectors. To implement this novel paradigm, we propose the COntrast in VEctor Regression (COVER) framework. COVER establishes an extendable vector-based self-learning, enforces a consistent optimization flow from vector regression to distance modeling, and leverages a vector pyramid architecture for granularity adaptation, thus preserving pixel-wise feature correlations in SSP. Extensive experiments across 8 tasks, spanning 2 dimensions and 4 modalities, show that COVER significantly improves pixel-wise SSP, advancing generalizable medical visual foundation models.

Figures

Figures reproduced from arXiv: 2506.20850 by the authors.

Figure 1
Figure 1. Over-dispersion problem: Binary CL excessively pur￾sues to disperse features, disrupting intra-class distribution and struggling models to disentangle the underlying semantics. [48], leading to a lack of global diversity that will cause CL models to converge to trivial solutions [60, 74]. Recent evidence highlights the critical importance of local features [42, 76, 80, 83] and leading results [21, 56, 81] in medical… view at source ↗
Figure 2
Figure 2. Motivation: Our vector CL unifies the distance model￾ing |d − d ′ | and displacement vector regression |v − V(d ′ )|, en￾abling CL with quantitative dispersion via the optimization flow. wise correspondences between two transformed views of a medical image. The pixel-wise distances are modeled through vector regression supervised by the DVF. This synthesis-based process enables a scalable and annotation￾free SSP alg… view at source ↗
Figure 3
Figure 3. Framework: Our COVER formulates the distance modeling as the regression of displacement vectors (coordinate correspon￾dences), quantifying dispersion degree in pixel-wise CL. a) SeVR (Sec.3.2) implements vector CL in a self-space transformation process. b) MoV (Sec.3.3) formulates feature distances as vectors with the properties of spatial continuity (VEU (i)) and correspondence ambiguity (MVI (ii)). c) VPA (Sec.3.4… view at source ↗
Figures from the paper (7 more)
Figure 5
Figure 5. Figure 5: Convergence analysis: a) Our COVER is able to pro￾mote convergence ability in fine-tuning. b) In the pretraining, the consistency learning causes representation collapse, and when adding our vector regression, the discrimination is enhanced. When adding our VEU, it eff…
Figure 7
Figure 7. Figure 7: Multi-scale vectors analysis: Our VPA predicts multi￾level correspondence for both global and detailed representation. 4.3.6 Distributed clusters from COVER The clusters show that our COVER has strong clustering capabilities by quantifying the dispersion of features. U…
Figure 8
Figure 8. Figure 8: The details of our vector template matrix. It is multiplied [PITH_FULL_IMAGE:figures/full_fig_p011_8.png]
Figure 9
Figure 9. Figure 9: The detailed implementations in our downstream tasks, [PITH_FULL_IMAGE:figures/full_fig_p013_9.png]
Figure 10
Figure 10. Figure 10: The analysis of the pre-training data amount. When [PITH_FULL_IMAGE:figures/full_fig_p014_10.png]
Figure 11
Figure 11. Figure 11: The visualization of the segmentation results for the methods with the top 2 average scores in each type. Our COVER achieves [PITH_FULL_IMAGE:figures/full_fig_p015_11.png]
Figure 12
Figure 12. Figure 12: Extension of multi-scale vectors analysis (Fig.8 in manuscript): The visualization of the regressed vectors via the alignment of two views. Our VPA is able to discover the correspondence of the semantics at multiple levels for both global and detailed representation. …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

88 extracted references · 70 canonical work pages

  1. [1]

    V oxelmorph: a learning framework for deformable medical image registration.IEEE transac- tions on medical imaging, 38(8):1788–1800, 2019

    Guha Balakrishnan, Amy Zhao, Mert R Sabuncu, John Gut- tag, and Adrian V Dalca. V oxelmorph: a learning framework for deformable medical image registration.IEEE transac- tions on medical imaging, 38(8):1788–1800, 2019. 3

    2019

  2. [2]

    Rademacher and gaussian complexities: Risk bounds and structural results

    Peter L Bartlett and Shahar Mendelson. Rademacher and gaussian complexities: Risk bounds and structural results. Journal of Machine Learning Research, 3(Nov):463–482,

  3. [3]

    Rep- resentation learning: A review and new perspectives.IEEE transactions on pattern analysis and machine intelligence, 35(8):1798–1828, 2013

    Yoshua Bengio, Aaron Courville, and Pascal Vincent. Rep- resentation learning: A review and new perspectives.IEEE transactions on pattern analysis and machine intelligence, 35(8):1798–1828, 2013. 2, 3, 7

    2013

  4. [4]

    On the opportunities and risks of foundation models.arXiv preprint arXiv:2108.07258, 2021

    Rishi Bommasani, Drew A Hudson, Ehsan Adeli, Russ Alt- man, Simran Arora, Sydney von Arx, Michael S Bernstein, Jeannette Bohg, Antoine Bosselut, Emma Brunskill, et al. On the opportunities and risks of foundation models.arXiv preprint arXiv:2108.07258, 2021. 2

    arXiv 2021

  5. [5]

    A survey of image registration tech- niques.ACM computing surveys (CSUR), 24(4):325–376,

    Lisa Gottesfeld Brown. A survey of image registration tech- niques.ACM computing surveys (CSUR), 24(4):325–376,

  6. [6]

    Swin-unet: Unet-like pure transformer for medical image segmentation

    Hu Cao, Yueyue Wang, Joy Chen, Dongsheng Jiang, Xi- aopeng Zhang, Qi Tian, and Manning Wang. Swin-unet: Unet-like pure transformer for medical image segmentation. InEuropean conference on computer vision, pages 205–218. Springer, 2022. 13

    2022

  7. [7]

    Deep clustering for unsupervised learning of visual features

    Mathilde Caron, Piotr Bojanowski, Armand Joulin, and Matthijs Douze. Deep clustering for unsupervised learning of visual features. InEuropean Conference on Computer Vi- sion, 2018. 1, 3, 6, 9

    2018

  8. [8]

    Krishna Chaitanya, Ertunc Erdil, Neerav Karani, and Ender Konukoglu. Contrastive learning of global and local fea- tures for medical image segmentation with limited annota- tions.Advances in neural information processing systems, 33:12546–12558, 2020. 1, 6

    2020

Show all 88 references
  1. [9]

    Transunet: Rethinking the u-net architec- ture design for medical image segmentation through the lens of transformers.Medical Image Analysis, 97:103280, 2024

    Jieneng Chen, Jieru Mei, Xianhang Li, Yongyi Lu, Qihang Yu, Qingyue Wei, Xiangde Luo, Yutong Xie, Ehsan Adeli, Yan Wang, et al. Transunet: Rethinking the u-net architec- ture design for medical image segmentation through the lens of transformers.Medical Image Analysis, 97:1032...

    2024

  2. [10]

    Towards a general-purpose foundation model for computational pathology.Nature Medicine, 30(3):850–862,

    Richard J Chen, Tong Ding, Ming Y Lu, Drew FK Williamson, Guillaume Jaume, Andrew H Song, Bowen Chen, Andrew Zhang, Daniel Shao, Muhammad Shaban, et al. Towards a general-purpose foundation model for computational pathology.Nature Medicine, 30(3):850–862,

  3. [11]

    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. InInternational conference on ma- chine learning, pages 1597–1607. PMLR, 2020. 1, 3, 6, 9

    2020

  4. [12]

    Exploring simple siamese representation learning

    Xinlei Chen and Kaiming He. Exploring simple siamese representation learning. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 15750–15758, 2021. 3, 6

    2021

  5. [13]

    Improved baselines with momentum contrastive learning

    Xinlei Chen, Haoqi Fan, Ross Girshick, and Kaiming He. Improved baselines with momentum contrastive learning. 15 Appearance transformed xa Space transformed xb Level 0 ψ' 0 ab(xa) Level 1 ψ' 1 ab(xa) Level 2 ψ' 2 ab(xa) Level 3 ψ' 3 ab(xa) Level 4 ψ' 4 ab(xa) Appearance transf...

    2003 arXiv

  6. [14]

    Pearson correlation coefficient.Noise reduction in speech processing, pages 1–4, 2009

    Israel Cohen, Yiteng Huang, Jingdong Chen, Jacob Benesty, Jacob Benesty, Jingdong Chen, Yiteng Huang, and Israel Cohen. Pearson correlation coefficient.Noise reduction in speech processing, pages 1–4, 2009. 13

    2009

  7. [15]

    Dis- tributed representations of behaviour-derived object dimen- sions in the human visual system.Nature Human Behaviour, 8(11):2179–2193, 2024

    Oliver Contier, Chris I Baker, and Martin N Hebart. Dis- tributed representations of behaviour-derived object dimen- sions in the human visual system.Nature Human Behaviour, 8(11):2179–2193, 2024. 3

    2024

  8. [16]

    Rethinking op- tical flow from geometric matching consistent perspective

    Qiaole Dong, Chenjie Cao, and Yanwei Fu. Rethinking op- tical flow from geometric matching consistent perspective. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 1337–1347, 2023. 3

    2023

  9. [17]

    Partial-to- partial shape matching with geometric consistency

    Viktoria Ehm, Maolin Gao, Paul Roetzer, Marvin Eisen- berger, Daniel Cremers, and Florian Bernard. Partial-to- partial shape matching with geometric consistency. InPro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 27488–27497, 2024. 3 16

    2024

  10. [18]

    Addison-Wesley Longman Publishing Co., Inc., 1987

    Rafael C Gonzales and Paul Wintz.Digital image process- ing. Addison-Wesley Longman Publishing Co., Inc., 1987. 1

    1987

  11. [19]

    Boot- strap 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 Pires, Zhaohan Guo, Mohammad Azar, et al. Boot- strap your own latent: A new approach to self-supervised learning. InNeural Information Processing ...

    2020

  12. [20]

    A basis for analyzing test-retest reliability

    Louis Guttman. A basis for analyzing test-retest reliability. Psychometrika, 10(4):255–282, 1945. 13

    1945

  13. [21]

    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,

  14. [22]

    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. InProceedings of the IEEE/CVF con- ference on computer vision and pattern recognition, pages 9729–9738, 2020. 1, 3

    2020

  15. [23]

    Yuting He, Tiantian Li, Rongjun Ge, Jian Yang, Youyong Kong, Jian Zhu, Huazhong Shu, Guanyu Yang, and Shuo Li. Few-shot learning for deformable medical image regis- tration with perception-correspondence decoupling and re- verse teaching.IEEE Journal of Biomedical and Health I...

    2021

  16. [24]

    Meta grayscale adaptive network for 3d integrated renal structures segmen- tation.Medical Image Analysis, 71:102055, 2021

    Yuting He, Guanyu Yang, Jian Yang, Rongjun Ge, Youy- ong Kong, Xiaomei Zhu, Shaobo Zhang, Pengfei Shao, Huazhong Shu, Jean-Louis Dillenseger, et al. Meta grayscale adaptive network for 3d integrated renal structures segmen- tation.Medical Image Analysis, 71:102055, 2021. 6, 12

    2021

  17. [25]

    Learning better registration to learn better few-shot medical image segmentation: Authenticity, diversity, and robustness

    Yuting He, Rongjun Ge, Xiaoming Qi, Yang Chen, Jia- song Wu, Jean-Louis Coatrieux, Guanyu Yang, and Shuo Li. Learning better registration to learn better few-shot medical image segmentation: Authenticity, diversity, and robustness. IEEE Transactions on Neural Networks and Lear...

    2022

  18. [26]

    Geometric visual similarity learning in 3d medical image self-supervised pre- training

    Yuting He, Guanyu Yang, Rongjun Ge, Yang Chen, Jean- Louis Coatrieux, Boyu Wang, and Shuo Li. Geometric visual similarity learning in 3d medical image self-supervised pre- training. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 202...

    2023

  19. [27]

    Learning better registration to learn better few-shot medical image segmentation: Authenticity, diversity, and robustness

    Yuting He, Rongjun Ge, Xiaoming Qi, Yang Chen, Jia- song Wu, Jean-Louis Coatrieux, Guanyu Yang, and Shuo Li. Learning better registration to learn better few-shot medical image segmentation: Authenticity, diversity, and robustness. IEEE Transactions on Neural Networks and Lear...

    2024

  20. [28]

    Founda- tion model for advancing healthcare: Challenges, opportuni- ties, and future directions.arXiv preprint arXiv:2404.03264,

    Yuting He, Fuxiang Huang, Xinrui Jiang, Yuxiang Nie, Minghao Wang, Jiguang Wang, and Hao Chen. Founda- tion model for advancing healthcare: Challenges, opportuni- ties, and future directions.arXiv preprint arXiv:2404.03264,

    arXiv

  21. [29]

    Homeomorphism prior for false posi- tive and negative problem in medical image dense contrastive representation learning, 2025

    Yuting He, Boyu Wang, Rongjun Ge, Yang Chen, Guanyu Yang, and Shuo Li. Homeomorphism prior for false posi- tive and negative problem in medical image dense contrastive representation learning, 2025. 2, 3, 5, 6

    2025

  22. [30]

    Overdispersion: mod- els and estimation.Computational statistics & data analysis, 27(2):151–170, 1998

    John Hinde and Clarice GB Dem ´etrio. Overdispersion: mod- els and estimation.Computational statistics & data analysis, 27(2):151–170, 1998. 1

    1998

  23. [31]

    Deep metric learning using triplet network

    Elad Hoffer and Nir Ailon. Deep metric learning using triplet network. InSimilarity-based pattern recognition: third inter- national workshop, SIMBAD 2015, Copenhagen, Denmark, October 12-14, 2015. Proceedings 3, pages 84–92. Springer,

    2015

  24. [32]

    Automatic lung segmentation in routine imaging is primarily a data di- versity problem, not a methodology problem.European Ra- diology Experimental, 4:1–13, 2020

    Johannes Hofmanninger, Forian Prayer, Jeanny Pan, Sebas- tian R ¨ohrich, Helmut Prosch, and Georg Langs. Automatic lung segmentation in routine imaging is primarily a data di- versity problem, not a methodology problem.European Ra- diology Experimental, 4:1–13, 2020. 12

    2020

  25. [33]

    Automated brain extraction of multisequence mri using artificial neural networks.Human brain mapping, 40(17): 4952–4964, 2019

    Fabian Isensee, Marianne Schell, Irada Pflueger, Gianluca Brugnara, David Bonekamp, Ulf Neuberger, Antje Wick, Heinz-Peter Schlemmer, Sabine Heiland, Wolfgang Wick, et al. Automated brain extraction of multisequence mri using artificial neural networks.Human brain mapping, 40(...

    2019

  26. [34]

    A programming language

    Kenneth E Iverson. A programming language. InProceed- ings of the May 1-3, 1962, spring joint computer conference, pages 345–351, 1962. 11

    1962

  27. [35]

    Fives: A fundus image dataset for artificial intelligence based vessel segmentation.Scientific data, 9(1):475, 2022

    Kai Jin, Xingru Huang, Jingxing Zhou, Yunxiang Li, Yan Yan, Yibao Sun, Qianni Zhang, Yaqi Wang, and Juan Ye. Fives: A fundus image dataset for artificial intelligence based vessel segmentation.Scientific data, 9(1):475, 2022. 6, 12

    2022

  28. [36]

    Can- dishare: A resource for pediatric neuroimaging data.Neu- roinformatics, 10(3):319, 2012

    David N Kennedy, Christian Haselgrove, Steven M Hodge, Pallavi S Rane, Nikos Makris, and Jean A Frazier. Can- dishare: A resource for pediatric neuroimaging data.Neu- roinformatics, 10(3):319, 2012. 6, 11, 12

    2012

  29. [37]

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

    Daniel S Kermany, Michael Goldbaum, Wenjia Cai, Car- olina CS Valentim, Huiying Liang, Sally L Baxter, Alex McKeown, Ge Yang, Xiaokang Wu, Fangbing Yan, et al. Identifying medical diagnoses and treatable diseases by image-based deep learning.cell, 172(5):1122–1131, 2018. 6, 12

    2018

  30. [38]

    Adam: A method for stochastic optimization.arXiv preprint arXiv:1412.6980,

    Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization.arXiv preprint arXiv:1412.6980,

    arXiv

  31. [39]

    Local rademacher complexities and oracle inequalities in risk minimization

    Vladimir Koltchinskii. Local rademacher complexities and oracle inequalities in risk minimization. 2006. 9

    2006

  32. [40]

    Unsupervised repre- sentation learning by predicting image rotations

    Nikos Komodakis and Spyros Gidaris. Unsupervised repre- sentation learning by predicting image rotations. InInter- national Conference on Learning Representations (ICLR),

  33. [41]

    U-kan makes strong backbone for medical image segmentation and generation.arXiv preprint,

    Chenxin Li, Xinyu Liu, Wuyang Li, Cheng Wang, Hengyu Liu, and Yixuan Yuan. U-kan makes strong backbone for medical image segmentation and generation.arXiv preprint,

  34. [42]

    Dense seman- tic contrast for self-supervised visual representation learning

    Xiaoni Li, Yu Zhou, Yifei Zhang, Aoting Zhang, Wei Wang, Ning Jiang, Haiying Wu, and Weiping Wang. Dense seman- tic contrast for self-supervised visual representation learning. InProceedings of the 29th ACM International Conference on Multimedia, pages 1368–1376, 2021. 1, 2, 3, 5, 6

    2021

  35. [43]

    Self-supervised visual representation learning for medical 17 image analysis: A comprehensive survey.Transactions on Machine Learning Research

    Siladittya Manna, Saumik Bhattacharya, and Umapada Pal. Self-supervised visual representation learning for medical 17 image analysis: A comprehensive survey.Transactions on Machine Learning Research. 1, 3

  36. [44]

    Cross-entropy loss functions: Theoretical analysis and applications

    Anqi Mao, Mehryar Mohri, and Yutao Zhong. Cross-entropy loss functions: Theoretical analysis and applications. InIn- ternational conference on Machine learning, pages 23803– 23828. PMLR, 2023. 13

    2023

  37. [45]

    The parkin- son progression marker initiative (ppmi).Progress in neuro- biology, 95(4):629–635, 2011

    Kenneth Marek, Danna Jennings, Shirley Lasch, Andrew Siderowf, Caroline Tanner, Tanya Simuni, Chris Coffey, Karl Kieburtz, Emily Flagg, Sohini Chowdhury, et al. The parkin- son progression marker initiative (ppmi).Progress in neuro- biology, 95(4):629–635, 2011. 6, 12, 13

    2011

  38. [46]

    Foundations of machine learning, 2018

    Mehryar Mohri. Foundations of machine learning, 2018. 4, 9

    2018

  39. [47]

    Foundation models for generalist medi- cal artificial intelligence.Nature, 616(7956):259–265, 2023

    Michael Moor, Oishi Banerjee, Zahra Shakeri Hossein Abad, Harlan M Krumholz, Jure Leskovec, Eric J Topol, and Pranav Rajpurkar. Foundation models for generalist medi- cal artificial intelligence.Nature, 616(7956):259–265, 2023. 1, 3

    2023

  40. [48]

    com Access with full downloadable image Bank

    Frank H Netter.Atlas of human anatomy, Professional Edi- tion E-Book: including NetterReference. com Access with full downloadable image Bank. Elsevier health sciences,

  41. [49]

    Unsupervised learning of dense visual representations.Advances in Neural Infor- mation Processing Systems, 33:4489–4500, 2020

    Pedro O O Pinheiro, Amjad Almahairi, Ryan Benmalek, Flo- rian Golemo, and Aaron C Courville. Unsupervised learning of dense visual representations.Advances in Neural Infor- mation Processing Systems, 33:4489–4500, 2020. 6

    2020

  42. [50]

    Repre- sentation learning with contrastive predictive coding.arXiv preprint arXiv:1807.03748, 2018

    Aaron van den Oord, Yazhe Li, and Oriol Vinyals. Repre- sentation learning with contrastive predictive coding.arXiv preprint arXiv:1807.03748, 2018. 1, 9

    2018 arXiv

  43. [51]

    Dinov2: Learning robust visual features without supervision

    Maxime Oquab, Timoth ´ee Darcet, Th ´eo Moutakanni, Huy V o, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel Haziza, Francisco Massa, Alaaeldin El-Nouby, et al. Dinov2: Learning robust visual features without supervision. Transactions on Machine Learning Research Jou...

    2024

  44. [52]

    Pytorch: An im- perative style, high-performance deep learning library.Ad- vances in neural information processing systems, 32, 2019

    Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, et al. Pytorch: An im- perative style, high-performance deep learning library.Ad- vances in neural information processing systems, ...

    2019

  45. [53]

    Context encoders: Feature learning by inpainting

    Deepak Pathak, Philipp Krahenbuhl, Jeff Donahue, Trevor Darrell, and Alexei A Efros. Context encoders: Feature learning by inpainting. InProceedings of the IEEE con- ference on computer vision and pattern recognition, pages 2536–2544, 2016. 3, 6

    2016

  46. [54]

    Fetal brain tissue annotation and segmentation challenge re- sults.Medical image analysis, 88:102833, 2023

    Kelly Payette, Hongwei Bran Li, Priscille de Dumast, Rox- ane Licandro, Hui Ji, Md Mahfuzur Rahman Siddiquee, Daguang Xu, Andriy Myronenko, Hao Liu, Yuchen Pei, et al. Fetal brain tissue annotation and segmentation challenge re- sults.Medical image analysis, 88:102833, 2023. 6, 11, 12

    2023

  47. [55]

    A new de- pendency and correlation analysis for features.IEEE Trans- actions on Knowledge and Data Engineering, 17(9):1199– 1207, 2005

    Guangzhi Qu, Salim Hariri, and Mazin Yousif. A new de- pendency and correlation analysis for features.IEEE Trans- actions on Knowledge and Data Engineering, 17(9):1199– 1207, 2005. 2, 3, 4

    2005

  48. [56]

    Igu-aug: Information-guided unsupervised augmentation and pixel-wise contrastive learning for medical image analy- sis.IEEE Transactions on Medical Imaging, 2024

    Quan Quan, Qingsong Yao, Heqin Zhu, and S Kevin Zhou. Igu-aug: Information-guided unsupervised augmentation and pixel-wise contrastive learning for medical image analy- sis.IEEE Transactions on Medical Imaging, 2024. 1

    2024

  49. [57]

    Study of thoracic ct in covid-19: The stoic project.Radiology, 301(1):E361–E370,

    Marie-Pierre Revel, Samia Boussouar, Constance de Margerie-Mellon, In `es Saab, Thibaut Lapotre, Dominique Mompoint, Guillaume Chassagnon, Audrey Milon, Math- ieu Lederlin, Souhail Bennani, et al. Study of thoracic ct in covid-19: The stoic project.Radiology, 301(1):E361–E370,

  50. [58]

    Convo- lutional neural network architecture for geometric matching

    Ignacio Rocco, Relja Arandjelovic, and Josef Sivic. Convo- lutional neural network architecture for geometric matching. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 6148–6157, 2017. 3

    2017

  51. [59]

    U- net: Convolutional networks for biomedical image segmen- tation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U- net: Convolutional networks for biomedical image segmen- tation. InInternational Conference on Medical image com- puting and computer-assisted intervention, pages 234–241. Springer, 2015. 6, 13

    2015

  52. [60]

    A theoretical analy- sis of contrastive unsupervised representation learning

    Nikunj Saunshi, Orestis Plevrakis, Sanjeev Arora, Mikhail Khodak, and Hrishikesh Khandeparkar. A theoretical analy- sis of contrastive unsupervised representation learning. InIn- ternational Conference on Machine Learning, pages 5628–

  53. [61]

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

    2000

  54. [62]

    Improved deep metric learning with multi- class n-pair loss objective.Advances in neural information processing systems, 29, 2016

    Kihyuk Sohn. Improved deep metric learning with multi- class n-pair loss objective.Advances in neural information processing systems, 29, 2016. 3

    2016

  55. [63]

    A tutorial on distance metric learning: Mathematical founda- tions, algorithms, experimental analysis, prospects and chal- lenges.Neurocomputing, 425:300–322, 2021

    Juan Luis Su ´arez, Salvador Garc´ıa, and Francisco Herrera. A tutorial on distance metric learning: Mathematical founda- tions, algorithms, experimental analysis, prospects and chal- lenges.Neurocomputing, 425:300–322, 2021. 2, 3

    2021

  56. [64]

    Metrics for evaluat- ing 3d medical image segmentation: analysis, selection, and tool.BMC medical imaging, 15(1):1–28, 2015

    Abdel Aziz Taha and Allan Hanbury. Metrics for evaluat- ing 3d medical image segmentation: analysis, selection, and tool.BMC medical imaging, 15(1):1–28, 2015. 6

    2015

  57. [65]

    Soft contrastive learning for visual localization.Advances in Neu- ral Information Processing Systems, 33:11119–11130, 2020

    Janine Thoma, Danda Pani Paudel, and Luc V Gool. Soft contrastive learning for visual localization.Advances in Neu- ral Information Processing Systems, 33:11119–11130, 2020. 1

    2020

  58. [66]

    What makes for good views for contrastive learning?Advances in Neural Infor- mation Processing Systems, 33:6827–6839, 2020

    Yonglong Tian, Chen Sun, Ben Poole, Dilip Krishnan, Cordelia Schmid, and Phillip Isola. What makes for good views for contrastive learning?Advances in Neural Infor- mation Processing Systems, 33:6827–6839, 2020. 1, 3, 9

    2020

  59. [67]

    Visualizing data using t-sne.Journal of machine learning research, 9 (11), 2008

    Laurens Van der Maaten and Geoffrey Hinton. Visualizing data using t-sne.Journal of machine learning research, 9 (11), 2008. 8

    2008

  60. [68]

    Segmentation of anatomical structures in chest radiographs using supervised methods: a comparative study on a public database.Medical image analysis, 10(1):19–40, 2006

    Bram Van Ginneken, Mikkel B Stegmann, and Marco Loog. Segmentation of anatomical structures in chest radiographs using supervised methods: a comparative study on a public database.Medical image analysis, 10(1):19–40, 2006. 6, 12

    2006

  61. [69]

    Attention is all you need.Advances in Neural Information Processing Systems, 2017

    A Vaswani. Attention is all you need.Advances in Neural Information Processing Systems, 2017. 5 18

    2017

  62. [70]

    Stacked denoising autoencoders: Learning useful represen- tations in a deep network with a local denoising criterion

    Pascal Vincent, Hugo Larochelle, Isabelle Lajoie, Yoshua Bengio, Pierre-Antoine Manzagol, and L ´eon Bottou. Stacked denoising autoencoders: Learning useful represen- tations in a deep network with a local denoising criterion. Journal of machine learning research, 11(12), 2010. 3, 6

    2010

  63. [71]

    A foundation model for clinical-grade computational pathology and rare cancers detection.Nature medicine, pages 1–12, 2024

    Eugene V orontsov, Alican Bozkurt, Adam Casson, George Shaikovski, Michal Zelechowski, Kristen Severson, Eric Zimmermann, James Hall, Neil Tenenholtz, Nicolo Fusi, et al. A foundation model for clinical-grade computational pathology and rare cancers detection.Nature medicine, ...

    2024

  64. [72]

    Understanding the behaviour of contrastive loss

    Feng Wang and Huaping Liu. Understanding the behaviour of contrastive loss. InProceedings of the IEEE/CVF con- ference on computer vision and pattern recognition, pages 2495–2504, 2021. 1, 3

    2021

  65. [73]

    Modet: Learning de- formable image registration via motion decomposition trans- former

    Haiqiao Wang, Dong Ni, and Yi Wang. Modet: Learning de- formable image registration via motion decomposition trans- former. InInternational Conference on Medical Image Com- puting and Computer-Assisted Intervention, pages 740–749. Springer, 2023. 3

    2023

  66. [74]

    Understanding contrastive representation learning through alignment and uniformity on the hypersphere

    Tongzhou Wang and Phillip Isola. Understanding contrastive representation learning through alignment and uniformity on the hypersphere. InInternational conference on machine learning, pages 9929–9939. PMLR, 2020. 1, 3

    2020

  67. [75]

    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. InProceedings of the IEEE conference on ...

  68. [76]

    Densecl: A simple framework for self-supervised dense visual pre-training.Visual Informatics, 2022

    Xinlong Wang, Rufeng Zhang, Chunhua Shen, and Tao Kong. Densecl: A simple framework for self-supervised dense visual pre-training.Visual Informatics, 2022. 1, 2, 3, 5, 6

    2022

  69. [77]

    Exploring set similarity for dense self-supervised rep- resentation learning

    Zhaoqing Wang, Qiang Li, Guoxin Zhang, Pengfei Wan, Wen Zheng, Nannan Wang, Mingming Gong, and Tongliang Liu. Exploring set similarity for dense self-supervised rep- resentation learning. InProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pa...

    2022

  70. [78]

    Large-scale 3d medical image pre-training with geometric context priors

    Linshan Wu, Jiaxin Zhuang, and Hao Chen. Large-scale 3d medical image pre-training with geometric context priors. arXiv preprint arXiv:2410.09890, 2024. 1, 3

    2024 arXiv

  71. [79]

    V oco: A simple- yet-effective volume contrastive learning framework for 3d medical image analysis

    Linshan Wu, Jiaxin Zhuang, and Hao Chen. V oco: A simple- yet-effective volume contrastive learning framework for 3d medical image analysis. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 22873–22882, 2024. 1

    2024

  72. [80]

    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. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 16684–16693,...

    2021

  73. [81]

    Sam: Self-supervised learning of pixel-wise anatom- ical embeddings in radiological images.IEEE Transactions on Medical Imaging, 41(10):2658–2669, 2022

    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. 1

    2022

  74. [82]

    Distance metric learning: A com- prehensive survey.Michigan State Universiy, 2(2):4, 2006

    Liu Yang and Rong Jin. Distance metric learning: A com- prehensive survey.Michigan State Universiy, 2(2):4, 2006. 2, 3

    2006

  75. [83]

    From single to universal: tiny lesion detection in medical imaging.Artificial Intelligence Review, 57(8):192, 2024

    Yi Zhang, Yiji Mao, Xuanyu Lu, Xingyu Zou, Hao Huang, Xinyang Li, Jiayue Li, and Haixian Zhang. From single to universal: tiny lesion detection in medical imaging.Artificial Intelligence Review, 57(8):192, 2024. 1

    2024

  76. [84]

    Rethinking dice loss for medical image segmentation

    Rongjian Zhao, Buyue Qian, Xianli Zhang, Yang Li, Rong Wei, Yang Liu, and Yinggang Pan. Rethinking dice loss for medical image segmentation. In2020 IEEE International Conference on Data Mining (ICDM), pages 851–860. IEEE,

  77. [85]

    A foundation model for generalizable disease detection from retinal images.Nature, 622(7981):156–163,

    Yukun Zhou, Mark A Chia, Siegfried K Wagner, Murat S Ayhan, Dominic J Williamson, Robbert R Struyven, Tim- ing Liu, Moucheng Xu, Mateo G Lozano, Peter Woodward- Court, et al. A foundation model for generalizable disease detection from retinal images.Nature, 622(7981):156–163,

  78. [86]

    Models genesis.Medical image analysis, 67:101840, 2021

    Zongwei Zhou, Vatsal Sodha, Jiaxuan Pang, Michael B Got- way, and Jianming Liang. Models genesis.Medical image analysis, 67:101840, 2021. 3, 6

    2021

  79. [87]

    Rubik’s cube+: A self-supervised feature learning framework for 3d medical image analysis.Medical image analysis, 64:101746, 2020

    Jiuwen Zhu, Yuexiang Li, Yifan Hu, Kai Ma, S Kevin Zhou, and Yefeng Zheng. Rubik’s cube+: A self-supervised feature learning framework for 3d medical image analysis.Medical image analysis, 64:101746, 2020. 3 19

    2020

  80. [2024]

    The visualization of the segmentation results for the methods with the top 2 average scores in each type

    1, 3 14 Ground Truth SimSiamCOVER GVSL PixPro DSC-PM BYOL Model Genesis Rotation Scratch FIVESS KiPA222D S SCR25% S GEMINI FeTA21S FeTA21S KiPA223D S KiPA223D S CANDIS CANDIS Figure 11. The visualization of the segmentation results for the methods with the top 2 average scores...

Pith tools

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