Pith. sign in

REVIEW 4 major objections 5 minor 32 references

The paper claims that in graph-based breast ultrasound classification, the image encoder's representation—not the GCN design—sets the patient-similarity graph's homophily, and that homophily predicts classification accuracy almost linearly

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

Across five encoder backbones, better self-supervised vision-transformer features improve both GCN graph homophily and breast-ultrasound classification accuracy, with homophily correlating strongly with accuracy.

T0 review reviewed 2026-08-02 challenge →

load-bearing objection Useful empirical encoder comparison for GCN breast ultrasound, but the 'higher-capacity' claim is contradicted by its own Table 3; worth a serious referee if reframed. the 4 major comments →

arxiv 2607.12054 v3 pith:BCQA4YZ3 submitted 2026-07-13 eess.IV cs.CVcs.LG

Analyzing Image Encoder Choices and Graph Homophily in GCN Frameworks for Breast Ultrasound Classification

Sabahattin Mert Daloglu , Ceren Coskun , Harvey Castro , Soner Hacihaliloglu , Ilker Hacihaliloglu This is my paper
classification eess.IV cs.CVcs.LG
keywords breast ultrasound classificationgraph convolutional networksgraph homophilyimage encodersself-supervised vision transformersDINOmedical image analysis
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Breast ultrasound classification is hard because benign and malignant lesions look similar and speckle noise is pervasive. The paper's proposal is that instead of designing a better graph network, you should feed the graph better image representations: it compares five encoders—a small CNN, ResNet-18, and three self-supervised vision transformers—within one fixed GCN pipeline. Its central finding is that higher-capacity encoders produce patient-similarity graphs with more same-label edges (higher homophily) and, correspondingly, higher classification accuracy; the self-supervised DINO ViT reaches 0.8509 accuracy with 0.8152 homophily versus 0.7573/0.6860 for the CNN baseline. Across all folds, test-set homophily and accuracy sit on a nearly straight line (R²≈0.853). If this is right, encoder representation quality is a major—and measurable—driver of graph-based medical image classification, and homophily becomes a practical diagnostic for model selection.

Core claim

On a merged 6,619-scan breast-ultrasound dataset with patient-wise folds, all five backbones feed 512-dimensional embeddings into the same cosine top-7 nearest-neighbor graph and a one-layer GCN with a linear head. The paper's core observation is a monotone ordering—CustomCNN < MAE ViT < ResNet-18 < I-JEPA ViT < DINO ViT-S/8—appearing jointly in accuracy, AUC, sensitivity, specificity, F1, and graph homophily. DINO's embeddings yield the most class-consistent graph (homophily 0.8152) and the best accuracy (0.8509); the fold-wise homophily-accuracy scatter has R²≈0.853. The authors interpret this as evidence that encoder choice shapes the graph's neighborhood structure, and cleaner neighborho

What carries the argument

The central object is the cosine-similarity k-nearest-neighbor patient graph with homophily h = (1/|E|) Σ I[y_i = y_j]. Each scan is a node; edges connect the top-K (=7) most similar embeddings under cosine similarity after z-score normalization. The graph's topology is entirely determined by the encoder's embeddings, and the GCN then propagates features over that fixed topology using a normalized adjacency. Homophily is the paper's diagnostic: it quantifies how many graph edges join same-label nodes, and the paper's R²≈0.853 linear fit links that structural quantity to final test accuracy.

Load-bearing premise

The central claim treats 'encoder choice' as the explanatory variable, but the five encoders differ in architecture, parameter count, pretraining objective, and pretraining data all at once—and MAE is the only one initialized from an ultrasound-domain checkpoint—so the clean ranking could be driven by pretraining domain rather than encoder capacity or type.

What would settle it

Run the same pipeline with a ViT-Base MAE initialized from standard ImageNet weights (not the ultrasound-domain checkpoint) and compare its homophily and accuracy against the current MAE ViT row. If the gap disappears, the result is about pretraining domain, not encoder architecture; if the gap remains, the paper's encoder framing survives. A second check: feed DINO embeddings into the GCN and also into a feature-only linear classifier; if the linear classifier already matches the GCN, graph message passing—and hence homophily—is not the operative mechanism.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • With the GCN architecture fixed, swapping the custom CNN for DINO ViT-S/8 raises test accuracy from 0.7573 to 0.8509 and graph homophily from 0.6860 to 0.8152, making encoder selection a first-order design decision rather than a tuning detail.
  • Self-supervised vision transformer features (DINO, I-JEPA) structure breast-ultrasound similarity graphs cleanly enough that a single graph convolution plus linear head reaches the best results, evidence that large-scale self-supervised pretraining transfers to this small-data medical task.
  • Test-graph homophily explains about 85% of the fold-wise variance in accuracy across encoder choices, so homophily can serve as a practical, interpretable predictor of which encoder will serve the GCN best.
  • Accuracy, AUC, sensitivity, specificity, and F1 all move in the same direction across the encoder ranking, so the gain from better encoders is not a sensitivity-specificity trade-off in this dataset.
  • Higher-capacity encoders consistently occupy the high-homophily, high-accuracy region across all three patient-wise folds, not just on average.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The paper leaves implicit that the same embeddings should help any message-passing architecture; testing a 2- or 3-layer GCN or an attention-based graph model under the same five encoders would show whether the ordering and the R²≈0.85 fit are specific to this one-layer GCN.
  • The MAE ViT arm is the only one initialized from an ultrasound-domain checkpoint (>230,000 ultrasound images), so the causal framing conflates architecture, capacity, and pretraining domain; a matched run with generic-weights MAE would isolate which factor drives the trend.
  • The homophily–accuracy correlation is also consistent with a common cause: stronger features independently improve both the graph and the classifier. A natural check is to keep DINO features but rewire or randomize graph edges; if accuracy does not fall with homophily, the graph-structure mechanism is not the active ingredient.
  • The homophily diagnostic could transfer to other medical modalities where frozen embeddings are used to build patient-similarity graphs, such as pathology or retinal imaging, where representation quality varies widely across encoders.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper evaluates five image encoders — CustomCNN, ResNet-18, MAE ViT, I-JEPA ViT, and DINO ViT-S/8 — in a unified GCN pipeline for binary benign/malignant breast ultrasound classification. It assembles 6619 scans from eight public datasets, constructs patient-wise train/validation/test splits, builds cosine top-K (K=7) nearest-neighbor graphs per split, and applies a single-layer GCN followed by a linear head. The central claim is that 'higher-capacity encoders consistently improve graph homophily and downstream classification performance' (Abstract; Section 3.1; Conclusion), with DINO ViT-S/8 performing best (accuracy 0.8509±0.0081, homophily 0.8152±0.0062) and CustomCNN worst. The paper also reports a strong linear fit between test-set homophily and test accuracy (R2≈0.853, Figure 3) and interprets this as evidence that encoder-driven graph-structure improvements are a key performance mechanism.

Significance. If the central causal claim held, the paper would be a useful empirical guide for encoder selection in graph-based medical image classification, and graph homophily would be a practically valuable diagnostic. The authors deserve credit for assembling a large multi-source dataset with patient-wise splits, fixing the GCN architecture across all backbones, and reporting a broad set of metrics. However, the headline claim is not supported by the reported data as stated: the backbone ordering in Table 3 is not monotone in capacity, and the MAE ViT backbone is confounded by ultrasound-domain pretraining. The homophily-accuracy regression is also post-hoc and uses ground-truth labels on the same test folds whose accuracy is predicted. The paper therefore currently establishes a set of empirical correlations and a plausible hypothesis, not a causal conclusion. Its value would increase substantially if the capacity claim were rephrased to match the actual controlled interventions and if the homophily diagnostic were validated as a forward predictor.

major comments (4)
  1. [Abstract; §3.1; Table 3] The claim that 'higher-capacity encoders consistently improve graph homophily and downstream classification performance' is contradicted by the paper's own data. Table 3 orders backbones as CustomCNN < MAE ViT < ResNet-18 < I-JEPA ViT < DINO ViT-S/8. Parameter counts are approximately: ResNet-18 ~11M, DINO ViT-S/8 ~21M, MAE ViT-B/16 ~86M, and I-JEPA at least ViT-B scale. Thus MAE ViT, the largest model, ranks below ResNet-18 and DINO, while DINO, one of the smallest models, is best. The reported trend is therefore not monotone in capacity. Please replace 'capacity' with a variable that is actually controlled or measured (e.g., architecture family, pretraining objective, or a quantitative representation-quality metric) and revise the abstract, Section 3.1, and conclusion accordingly.
  2. [§2.4] The MAE ViT backbone is initialized from 'an ultrasound-domain MAE checkpoint pretrained on more than 230,000 deidentified ultrasound images,' whereas ResNet-18, DINO, and I-JEPA are initialized from generic public weights. This means MAE's performance and homophily values reflect a combined intervention (architecture + pretraining data), not encoder choice alone. The comparison is confounded for the central claim. At minimum, this must be stated as an explicit limitation; stronger remedies are to include an ImageNet-pretrained MAE baseline or to use ultrasound-domain checkpoints for DINO/I-JEPA as well.
  3. [§3.3; Eq. (6); Figure 3] The homophily-accuracy fit uses test-graph homophily computed from ground-truth labels (Eq. 6) on exactly the same test folds whose accuracy is then regressed against it. This is a post-hoc correlation, not a forward prediction or a causal indicator: both quantities derive from the same labels and the same test allocation. Additionally, the 15 fold-wise points are clustered by backbone, so the reported R2≈0.853 likely reflects between-backbone differences rather than a general diagnostic relation. To support the 'key indicator' claim, the authors should report a predictive experiment (e.g., train the fit on some backbones/folds and use it to predict held-out accuracy), exclude labels from the homophily computation if a feature-only diagnostic is intended, and provide confidence intervals for the fit.
  4. [Table 3; §3.1] With only three folds, the reported standard deviations are large enough to undermine the claimed 'consistent' ordering. For example, MAE ViT accuracy is 0.7890±0.0383 and ResNet-18 is 0.7958±0.0039; these intervals overlap considerably. The paper should provide per-fold results and either paired significance tests or effect sizes with confidence intervals. Without this, the intermediate ordering (MAE vs ResNet-18 vs I-JEPA) is not statistically substantiated.
minor comments (5)
  1. [Throughout] The terms 'patientwise' and 'patient-wise' are used inconsistently; please standardize.
  2. [§2.2] I-JEPA ViT is not fully specified: the text should state whether it is ViT-S, ViT-B, or ViT-L and give the parameter count, as is done for DINO ViT-S/8 and MAE ViT-Base/16.
  3. [§2.3] The choice K=7 is described as based on 'initial hyperparameter sweeps,' but no sweep results or sensitivity analysis are given. Reporting how performance/homophily vary with K would strengthen the paper.
  4. [§2.3] The z-score normalization step is applied 'to all node features' before graph construction. Please clarify whether this is per-dimension across nodes or per-node across feature dimensions; this affects the cosine similarity.
  5. [§3.2; Figure 2] The distinction between pooled AUC in Figure 2 and fold-averaged AUC in Table 3 is explained, but the different absolute values may confuse readers. Consider stating this explicitly in the figure caption as well as the text.

Circularity Check

1 steps flagged

Mostly self-contained empirical comparison; only the homophily-as-'predictor' framing is post-hoc/self-referential, and the capacity claim is confounded (not circular).

specific steps
  1. fitted input called prediction [Section 3.3 / Figure 3; Eq. 6; Discussion]
    "Test-graph homophily is obtained using: h= 1/|E| ∑_{(i,j)∈E} I[y_i = y_j], (6) ... Figure 3 further shows a strong association between homophily and test accuracy (R2≈ 0.853), supporting homophily as a useful cross-encoder predictor of GCN performance."

    The 'predictor' h is computed from ground-truth labels y_i/y_j on the same test set whose classification accuracy is being regressed, so it is not a label-free forward predictor. The reported R2 is a post-hoc association between two quantities that share the same label source and the same encoder-derived graph, rather than an independent prediction of accuracy. This does not make the encoder-ranking results circular, but the homophily-as-predictor claim is self-referential.

full rationale

Most of the paper is an empirical bake-off with no analytic derivation chain: five encoder embeddings → cosine kNN graph → single-layer GCN → test metrics. Table 3 and Figure 2 are direct measurements, not fitted predictions. The only borderline item is the homophily-accuracy 'predictor' in Eq. 6 / Fig. 3, which uses ground-truth labels on the same test set whose accuracy is then correlated with it; that is a post-hoc diagnostic, not a forward prediction. The central capacity claim is additionally confounded (MAE ViT-B/16 is the largest model but ranks below ResNet-18 and DINO ViT-S/8, and MAE additionally uses an ultrasound-domain checkpoint), but confounding is a validity/correctness concern, not circularity. The self-citation [7] supplies only the CustomCNN baseline and is not load-bearing. Thus no 6+ circularity is warranted; the minor self-referential framing gives score 2.

Axiom & Free-Parameter Ledger

3 free parameters · 4 axioms · 0 invented entities

The paper is an empirical benchmark rather than a derivation, so the ledger contains mostly domain assumptions and model hyperparameters. The most consequential entries are the ultrasound-domain MAE checkpoint confound and the unreported data-exclusion/hyperparameter choices, both of which affect the strength of the causal claim.

free parameters (3)
  • K (top-K neighbors) = 7
    Selected via 'initial hyperparameter sweeps' (Section 2.3), not derived; directly controls graph sparsity and thus homophily and GCN behavior.
  • GCN hidden width / dropout / edge dropout = 256, p=0.5, p=0.2
    Fixed by design (Sections 2.3-2.4); not ablated, but standard values chosen by the authors.
  • Fold-wise test allocation = ~10% (640/663/555 test scans per fold)
    Chosen to be roughly balanced per sub-dataset (Section 2.1); introduced by hand, not by a formal procedure.
axioms (4)
  • domain assumption The eight public datasets' benign/malignant labels are correct and consistent, and patient-wise folds prevent patient overlap.
    Invoked in Section 2.1; no patient-level IDs or cross-dataset deduplication is shown.
  • standard math Standard GCN propagation (Eqs. 2-3) and cosine kNN graph (Eq. 1) are appropriate for this task.
    Used throughout; standard methods from Kipf & Welling (2017).
  • domain assumption Ultrasound-domain MAE pretraining on >230k images transfers to these breast scans and does not confound the comparison.
    Section 2.4; MAE alone uses an ultrasound-specific checkpoint, so any MAE advantage could come from pretraining data rather than encoder choice.
  • ad hoc to paper 'Low-quality scans' removal and K=7 selection do not introduce selection bias.
    Section 2.1 and Section 2.3; exclusion criteria and sweep details are not specified.

reviewed 2026-08-02 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Analyzing Image Encoder Choices and Graph Homophily in GCN Frameworks for Breast Ultrasound Classification." pith.science (2026). https://pith.science/paper/BCQA4YZ3

@misc{pith2026260712054,
  author       = {Pith},
  title        = {Pith review of: Analyzing Image Encoder Choices and Graph Homophily in GCN Frameworks for Breast Ultrasound Classification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BCQA4YZ3}},
  note         = {Machine review of arXiv:2607.12054}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Breast ultrasound is widely used for screening, yet automated analysis remains challenging due to speckle noise, acquisition variability, and weak separation of benign and malignant cases in standard ultrasound imaging. Graph convolutional networks (GCNs) have recently emerged as a promising approach by leveraging relationships among similar patient samples. However, it remains unclear how the choice of image encoder influences graph construction and downstream classification performance. In this work, we systematically evaluate five image encoders spanning convolutional and transformer-based architectures for GCN-based breast ultrasound classification. Image embeddings are used to construct cosine similarity k-nearest-neighbor graphs, which are classified using a single-layer GCN with a linear classification head. Across three patientwise cross-validation folds, higher-capacity encoders consistently improve graph homophily and downstream classification performance, yielding gains in accuracy, AUC, sensitivity, specificity, and F1-score. Moreover, test-set graph homophily exhibits a strong linear correlation with classification accuracy, with higher-capacity encoders consistently occupying the high-homophily, high-accuracy region suggesting that encoder-driven improvements in graph structure are a key mechanism underlying the observed performance gains. These findings establish encoder selection as a critical factor in graph-based breast ultrasound classification and identify graph homophily as a key indicator linking representation quality to downstream classification performance.

Figures

Figures reproduced from arXiv: 2607.12054 by Ceren Coskun, Harvey Castro, Ilker Hacihaliloglu, Sabahattin Mert Daloglu, Soner Hacihaliloglu.

Figure 1
Figure 1. Figure 1: Overview of the GCN pipeline. Ultrasound scans are encoded by one backbone (CustomCNN, ResNet-18, MAE ViT, DINO ViT-S/8, or I-JEPA ViT), z-score normal￾ized, converted to a cosine top-K graph (K = 7), and processed by a GCN with one graph convolutional layer followed by a linear output layer to produce benign/malignant predictions and evaluation metrics (accuracy, AUC, homophily, sensitivity, specificity, … view at source ↗
Figure 2
Figure 2. Figure 2: Combined ROC plot (all backbones on one axis). For each backbone, predictions from all test folds were pooled to compute a single ROC curve; legend AUC values correspond to the pooled ROC AUC. that encoder quality is associated with both discriminative performance and graph structure. DINO ViT-S/8 is the strongest overall model in most metrics. These results suggest that stronger encoder representations ma… view at source ↗
Figure 3
Figure 3. Figure 3: Test accuracy vs. test-graph homophily for GCN models using fold-wise points (15 total: 3 folds per backbone), with one additional per-backbone average marker for visualization. The linear fit is computed only from the fold-wise points. 3.3 Accuracy vs Homophily Fit [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

Review history (2 revisions) →

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

32 extracted references · 5 canonical work pages

  1. [1]

    The Cancer Imaging Archive (2024)

    A curated benchmark dataset for ultrasound based breast lesion analy- sis (breast-lesions-usg) (version 1). The Cancer Imaging Archive (2024). https://doi.org/10.7937/9WKK-Q141

    doi:10.7937/9wkk-q141 2024

  2. [2]

    Data in Brief28, 104863 (2020)

    Al-Dhabyani, W., Gomaa, M., Khaled, H., Fahmy, A.: Dataset of breast ultrasound images. Data in Brief28, 104863 (2020). https://doi.org/10.1016/j.dib.2019.104863

    arXiv 2020

  3. [3]

    Computers in Biology and Medicine152, 106438 (2023)

    Ardakani, A.A., Mohammadi, A., Mirza-Aghazadeh-Attari, M., Acharya, U.R.: An open-access breast lesion ultrasound image database: Applicable in artifi- cial intelligence studies. Computers in Biology and Medicine152, 106438 (2023). https://doi.org/10.1016/j.compbiomed.2022.106438

    arXiv 2023

  4. [4]

    arXiv preprint arXiv:2301.08243 (2023)

    Assran, M., Duval, Q., Misra, I., Bojanowski, P., Vincent, P., Rabbat, M., LeCun, Y., Ballas, N.: Self-supervised learning from images with a joint-embedding predictive architecture. arXiv preprint arXiv:2301.08243 (2023)

    Pith/arXiv arXiv 2023

  5. [5]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)

    Caron, M., Touvron, H., Misra, I., Jégou, H., Mairal, J., Bojanowski, P., Joulin, A.: Emerging properties in self-supervised vision transformers. In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV). pp. 9650–9660 (2021)

    2021

  6. [6]

    In: Medical Image Computing and Computer Assisted Intervention (MICCAI)

    Chen, Y., Zhang, C., Liu, L., et al.: Uscl: Pretraining deep ultrasound image diagnosis model through video contrastive representation learning. In: Medical Image Computing and Computer Assisted Intervention (MICCAI). pp. 627–637. Springer (2021)

    2021

  7. [7]

    Mayo Clinic Proceedings: Innovations, Quality & Outcomes10(1), 100691 (2026)

    Daloglu, S.M., Coskun, C., Bekar, G., Sahin, S., Letter, H.P., Castro, H., Haci- haliloglu, S., Hacihaliloglu, I.: Multifeature ultrasound-based classification for breast lesions: A comparative study of pons image enhancement technology. Mayo Clinic Proceedings: Innovations, Quality & Outcomes10(1), 100691 (2026). https://doi.org/10.1016/j.mayocpiqo.2025.100691

    arXiv 2026

  8. [8]

    In: International Conference on Learning Representations (ICLR) (2021)

    Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., Uszkoreit, J., Houlsby, N.: An image is worth 16x16 words: Transformers for image recognition at scale. In: International Conference on Learning Representations (ICLR) (2021)

    2021

  9. [9]

    Zenodo (2023)

    Gomez-Flores, W., Gregorio-Calas, M.J., Albuquerque Pereira, W.C.d.: Bus-bra: A breast ultrasound dataset for assessing computer-aided diagnosis systems (1.0). Zenodo (2023). https://doi.org/10.5281/zenodo.8231412

    work page doi:10.5281/zenodo.8231412 2023

  10. [10]

    European Journal of Radiology157, 110591 (2022)

    Hamyoon, H., Yee Chan, W., Mohammadi, A., Kuzan, T.Y., Mirza-Aghazadeh- Attari, M., Leong, W.L., Acharya, N.A.M., Rajendra Acharya, U.: Artificial intelli- gence, bi-rads evaluation and morphometry: A novel combination to diagnose breast cancer using ultrasonography, results from multi-center cohorts. European Journal of Radiology157, 110591 (2022). https...

    arXiv 2022

  11. [11]

    arXiv preprint arXiv:2111.06377 (2021)

    He, K., Chen, X., Xie, S., Li, Y., Dollár, P., Girshick, R.: Masked autoencoders are scalable vision learners. arXiv preprint arXiv:2111.06377 (2021)

    Pith/arXiv arXiv 2021

  12. [12]

    In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)

    He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recognition. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). pp. 770–778 (2016). https://doi.org/10.1109/CVPR.2016.90

    doi:10.1109/cvpr.2016.90 2016

  13. [13]

    PLoS One 13(8), e0200412 (2018)

    van den Heuvel, T.L.A., de Bruijn, D., de Korte, C.L., van Ginneken, B.: Automated measurement of fetal head circumference using 2d ultrasound images. PLoS One 13(8), e0200412 (2018). https://doi.org/10.1371/journal.pone.0200412

    doi:10.1371/journal.pone.0200412 2018

  14. [14]

    Biocybernetics and Biomedical Engineering42, 921–933 (2022)

    Homayoun, H., Chan, W.Y., Kuzan, T.Y., Leong, W.L., Altintoprak, K.M., Mohammadi, A., Vijayananthan, A., Ardakani, A.A., Acharya, U.R., et al.: Applications of machine-learning algorithms for prediction of benign and malignant breast lesions using ultrasound radiomics signatures: A multi- center study. Biocybernetics and Biomedical Engineering42, 921–933 ...

    work page doi:10.1016/j.bbe.2022.07.004 2022

  15. [15]

    arXiv preprint arXiv:1609.02907 (2016) Encoder Choice and Graph Homophily in GCN Classification 11

    Kipf, T.N., Welling, M.: Semi-supervised classification with graph convolutional networks. arXiv preprint arXiv:1609.02907 (2016) Encoder Choice and Graph Homophily in GCN Classification 11

    Pith/arXiv arXiv 2016

  16. [16]

    IEEE Transactions on Medical Imaging38(9), 2198–2210 (2019)

    Leclerc, S., Smistad, E., Pedrosa, J., et al.: Deep learning for segmentation using an open large-scale dataset in 2d echocardiography. IEEE Transactions on Medical Imaging38(9), 2198–2210 (2019). https://doi.org/10.1109/TMI.2019.2900516

    arXiv 2019

  17. [17]

    In: Medical Image Computing and Computer Assisted Intervention (MICCAI)

    Li, J., Zheng, Q., Li, M., et al.: Rethinking breast lesion segmentation in ultrasound: A new video dataset and a baseline network. In: Medical Image Computing and Computer Assisted Intervention (MICCAI). pp. 391–400. Springer (2022)

    2022

  18. [18]

    Luan, S., Hua, C., Xu, M., Lu, Q., Zhu, J., Chang, X.W., Fu, J., Leskovec, J., Precup, D.: When do graph neural networks help with node classification? investigating the impact of homophily principle on node distinguishability (2024)

    2024

  19. [19]

    IEEE Reviews in Biomedical Engineering (2024)

    Luo, L., Wang, X., Lin, Y., Ma, X., Tan, A., Chan, R., Vardhanabhuti, V., Chu, C.W., Cheng, K.T., Chen, H.: Deep learning in breast cancer imaging: A decade of progress and future directions. IEEE Reviews in Biomedical Engineering (2024). https://doi.org/10.1109/rbme.2024.3357877

    arXiv 2024

  20. [20]

    Ma, Y., Liu, X., Shah, N., Tang, J.: Is homophily a necessity for graph neural networks? CoRRabs/2106.06134(2021)

    Pith/arXiv arXiv 2021

  21. [21]

    IEEE Journal of Biomedical and Health Informatics22(4), 1218– 1226 (Aug 2017)

    Marti, M.H., Pons, G., Marti, J., Ganau, S., Sentis, M., Zwiggelaar, R., Davison, A.K., Marti, R.: Automated breast ultrasound lesions detection using convolutional neural networks. IEEE Journal of Biomedical and Health Informatics22(4), 1218– 1226 (Aug 2017)

    2017

  22. [22]

    Meyer, A., Murali, A., Zarin, F., Mutter, D., Padoy, N.: Ultrasam: A foundation model for ultrasound using large open-access segmentation datasets. Int. J. Comput. Assist. Radiol. Surg. (2025). https://doi.org/10.1007/s11548-025-03517-8

    work page doi:10.1007/s11548-025-03517-8 2025

  23. [23]

    IEEE Trans

    Mo, Y., et al.: Hover-trans: Anatomy-aware hover-transformer for roi-free breast cancer diagnosis in ultrasound images. IEEE Trans. Med. Imaging pp. 1–1 (2023). https://doi.org/10.1109/TMI.2023.3236011

    arXiv 2023

  24. [24]

    Rutgers University (2020)

    Mohabir, J.: Knee cartilage segmentation of ultrasound images using convolutional neural networks and local phase enhancement. Rutgers University (2020)

    2020

  25. [25]

    Digital Health10, 20552076241251660 (2024)

    Montaha, S., Azam, S., Bhuiyan, M.R.I., Chowa, S.S., Mukta, M.S.H., Jonkman, M.: Malignancy pattern analysis of breast ultrasound images using clinical features and a graph convolutional network. Digital Health10, 20552076241251660 (2024). https://doi.org/10.1177/20552076241251660

    work page doi:10.1177/20552076241251660 2024

  26. [26]

    Nature580(7802), 252–256 (2020)

    Ouyang, D., He, B., Ghorbani, A., et al.: Video-based ai for beat-to- beat assessment of cardiac function. Nature580(7802), 252–256 (2020). https://doi.org/10.1038/s41586-020-2145-8

    doi:10.1038/s41586-020-2145-8 2020

  27. [27]

    Journal of the American Society of Echocardiography36(5), 482–489 (2023)

    Reddy, C.D., Lopez, L., Ouyang, D., Zou, J.Y., He, B.: Video-based deep learn- ing for automated assessment of left ventricular ejection fraction in pediatric patients. Journal of the American Society of Echocardiography36(5), 482–489 (2023). https://doi.org/10.1016/j.echo.2023.01.015

    doi:10.1016/j.echo.2023.01.015 2023

  28. [28]

    Zenodo (2024)

    Sappia, M.S.: Acouslic-ai: Abdominal circumference operator-agnostic ultrasound measurement in low-income countries using artificial intelligence. Zenodo (2024). https://doi.org/10.5281/zenodo.12697994

    doi:10.5281/zenodo.12697994 2024

  29. [29]

    Computerized Medical Imaging and Graphics87, 101829 (2021)

    Shia, W.C., Chen, D.R.: Classification of malignant tumors in breast ultra- sound using a pretrained deep residual network model and support vector machine. Computerized Medical Imaging and Graphics87, 101829 (2021). https://doi.org/10.1016/j.compmedimag.2020.101829

    arXiv 2021

  30. [30]

    Mendeley Data, V3 (2025)

    Vallez, N., Bueno, G., Deniz, O., Rienda, M.A., Pastor, C.: Bus-uclm: Breast ultrasound lesion segmentation dataset. Mendeley Data, V3 (2025). https://doi.org/10.17632/7fvgj4jsp7.3

    work page doi:10.17632/7fvgj4jsp7.3 2025

  31. [31]

    In: Medical Image Computing and Computer-Assisted Intervention (MICCAI)

    Wang, P., Patel, V.M., Hacihaliloglu, I.: Simultaneous segmentation and classifica- tion of bone surfaces from ultrasound using a multi-feature guided cnn. In: Medical Image Computing and Computer-Assisted Intervention (MICCAI). pp. 134–142. Springer (2018)

    2018

  32. [32]

    Information Fusion p

    Yan, P., Gong, W., Li, M., Zhang, J., Li, X., Jiang, Y., Luo, H., Zhou, H.: Tdf-net: Trusted dynamic feature fusion network for breast cancer diagnosis using incomplete multimodal ultrasound. Information Fusion p. 102592 (2024)

    2024

This paper was first reviewed by deepseek-v4-flash on August 2, 2026.