Pith. sign in

REVIEW 4 major objections 5 minor 28 references

CM-UNet: A Self-Supervised Learning-Based Model for Coronary Artery Segmentation in X-Ray Angiography

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Pith's one-line read The paper claims that CM-UNet, a UNet pre-trained with a contrastive masked autoencoder on unlabeled coronary X-ray angiography, keeps segmentation accurate when fine-tuned on as few as 18 annotated images, where a conventionally trained…

desk verdict The broad claim that SSL pre-training helps in low-data coronary segmentation holds up; the specific CMAE superiority claim is muddied by a parameter-count ambiguity and missing error bars. read the letter →

arxiv 2507.17779 v1 pith:GBEFWGVZ submitted 2025-07-22 q-bio.QM cs.LG

classification q-bio.QMcs.LG
keywords coronaryarterysegmentationX-rayangiographyself-supervisedlearningcontrastivemaskedautoencoderlow-datafine-tuningUNetDicescoretransfer
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 tries to show that self-supervised pre-training can replace most manual annotation in coronary artery segmentation from X-ray angiography. It introduces CM-UNet, a UNet pre-trained with a contrastive masked autoencoder on unlabeled angiograms, then fine-tuned on a small labeled set. The central result is that with only 18 labeled images for fine-tuning, CM-UNet reaches a Dice score of 0.626, while a supervised UNet without pre-training drops to 0.375. The authors argue this matters because annotation of coronary vessels is costly and scarce, and a model that learns from unlabeled data could make automated stenosis assessment practical in more clinics.

What carries the argument

The central object is the Contrastive Masked Autoencoder (CMAE) objective, adapted to a UNet backbone. One branch masks 65% of $16 \times 16$ patches and reconstructs them with an encoder-decoder; a second branch applies heavier augmentations and feeds a momentum encoder whose outputs are pulled together with the reconstruction branch's features through an InfoNCE loss. The two losses are combined as $\mathcal{L} = \mathcal{L}_r + \lambda_c \mathcal{L}_c$, and after pre-training the online encoder and pixel decoder become the initialization for the segmentation UNet. This joint objective is what carries the argument: reconstruction preserves local vessel continuity, while contrastive alignment keeps representations discriminative across images.

What would settle it

Train a plain UNet with the same 121.49-million-parameter architecture and the same 18-image fine-tuning protocol; if it reaches or exceeds Dice 0.626 without any pre-training, the reported gain is explained by capacity rather than by the contrastive masked autoencoder.

Watch

Extended reading notes

Core claim

CM-UNet combines two self-supervised objectives into one pre-training stage: masked image reconstruction, which forces the network to recover fine vessel details, and contrastive learning, which aligns global representations of different augmented views of the same angiogram. After pre-training on unlabeled data from the FAME2 dataset, the reconstruction encoder and decoder are reused as the segmentation UNet and fine-tuned with a Dice plus cross-entropy loss. The paper reports that this pipeline outperforms both a supervised UNet and other self-supervised baselines in low-data regimes; at a 79:1 pre-training-to-fine-tuning ratio, CM-UNet scores 0.626 Dice versus 0.563 for the best alternative SSL method and 0.375 without pre-training. The authors conclude that self-supervised pre-training, especially the combined contrastive-masked objective, is what lets segmentation survive extreme label scarcity.

Load-bearing premise

The comparison assumes CM-UNet's advantage comes from its self-supervised objective rather than from its much larger network, because no equally sized UNet or capacity-matched SSL baseline is trained.

Editorial extensions

If this is right

  • Fine-tuning CM-UNet on 18 labeled images instead of 500 costs 15.2% in Dice, whereas non-pre-trained models lose 46.5%, so the amount of annotation needed for usable coronary segmentation can shrink by more than an order of magnitude.
  • At the 79:1 data split, CM-UNet's Dice of 0.626 beats the best other self-supervised baseline (0.563) and the supervised UNet (0.375), suggesting the combined objective generalizes better than contrastive-only or reconstruction-only pre-training in this regime.
  • Across all tested pre-training-to-fine-tuning ratios, pre-trained models outperform the non-pre-trained model, and the gap widens as fine-tuning data shrinks, implying SSL becomes more valuable exactly when annotations are rarest.
  • CM-UNet's predicted vessel diameters correlate with manual ground-truth diameters (PCC 0.76 at 50:30 and 0.57 at 79:1), indicating the model preserves clinically meaningful measurements even under scarce labels.

Reading between the lines

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

  • A capacity-matched test the paper leaves open: if a 121M-parameter UNet without SSL also reaches roughly 0.6 Dice on 18 images, the contribution of the contrastive-masked objective would need to be re-examined.
  • If the effect is real, the same two-stage recipe should transfer to other thin-structure segmentation tasks, such as cerebral or peripheral angiography, where labeled examples are even rarer.
  • The paper lists ablation of the reconstruction and contrastive components as future work; one predictable outcome is that reconstruction matters most for vessel continuity and contrastive learning for discriminating arteries from background clutter.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes CM-UNet, a self-supervised pretraining framework for coronary artery segmentation from X-ray angiography. It combines a masked reconstruction branch with a contrastive branch during pretraining on unlabeled images, then retains the reconstruction encoder and decoder as a UNet that is fine-tuned on small labeled subsets. On the FAME2 dataset, the authors report that CM-UNet reaches a Dice of 0.626 with only 18 fine-tuning images versus 0.375 for a supervised UNet, and that its Dice drop when fine-tuning data shrinks from 500 to 18 images is 15.2% compared with 46.5% for the supervised baseline. The paper also benchmarks four other SSL methods, reports clinical metrics such as artery diameter difference and Bland-Altman agreement, and releases source code.

Significance. If the central comparisons were clean, the paper would make a useful practical contribution: it addresses a real annotation bottleneck in coronary angiography, evaluates on a clinical dataset with held-out test images, and provides clinical metrics beyond Dice. The release of source code and the use of a fixed seed for reproducibility are strengths, as is the breadth of SSL baselines considered. However, the head-to-head claim that CM-UNet outperforms other SSL methods currently rests on a capacity confound and on unreplicated single-run numbers, so the magnitude and attribution of the observed gains are not yet established.

major comments (4)
  1. [§III.B, Table I, §IV.C] The capacity comparison that underlies the claim that CM-UNet outperforms other SSL methods is confounded. §III.B states that after pre-training only the reconstruction encoder E_r and pixel decoder D_r are retained and combined into a UNet; this is the same four-level UNet architecture described in §IV.C for all baselines. Table I nevertheless reports 121.49M parameters for CM-UNet versus 31.04-37.70M for the baselines, and the text interprets this as "increased complexity enhances segmentation accuracy." Since 121.49M appears to include the momentum encoder, feature decoder, and projection/prediction heads used only during pre-training, the deployed segmentation model is not 121.49M parameters. If the 121.49M model is actually the fine-tuned model, then no capacity-matched baseline exists. Under either reading, the +0.063 Dice margin over the best baseline in Table I cannot be attributed specifically to the contrastive masked autoencoder objective. Please report the parameter count of the fine-tuned CM-UNet and add a same-capacity baseline or an SSL baseline with a matched inference architecture.
  2. [Fig. 3, Table I, §V] All segmentation results are reported as single numbers with no variance. Table I reports one Dice value per method at the 79:1 ratio, Fig. 3 plots curves without error bars or confidence bands, and §V calls the 4.3% improvement over the non-pre-trained model "statistically significant (p<0.05)" without naming a test or describing the distribution. With 18 fine-tuning images, random initialization, augmentation, and pretraining stochasticity can materially change results. Please run multiple seeds (or repeated fine-tuning) and report means with standard deviations or confidence intervals, and state the hypothesis test used for any significance claim.
  3. [§III.A, §V] The paper itself lists the absence of an ablation of the reconstruction and contrastive components as a limitation. This omission is load-bearing because §III.A motivates the combined objective as balancing "global" and "fine-grained" features, and the main numerical advantage of CM-UNet is presented as the product of this combination. A controlled ablation (reconstruction-only, contrastive-only, and combined within the same framework and budget) is needed to establish that the fusion, rather than one of the branches, drives the gains; the baselines MAE and MoCo do not substitute for this because they use different objectives, implementations, and pre-training budgets.
  4. [Table I, §IV.D1] The claim that CM-UNet "outperforms competing SSL approaches" is not supported on all metrics in Table I. For ADD, CM-UNet scores 0.965, which is worse than Model Genesis's 0.893; the sentence reporting a -13.33% reduction in ADD therefore needs a clearly stated reference value. If the comparison is to the mean of the SSL baselines, the table implies about a 16.8% reduction, not 13.33%; if the comparison is to the best baseline, the claim is false on ADD. Please recompute and report all percentage improvements against an explicitly named reference, and qualify the outperformance claim per metric.
minor comments (5)
  1. [§IV.D1] The sentence "only 1% of the training dataset (18 images)" is inconsistent with a 1,390-image training set; 18 images is approximately 1% of all 1,738 images. Clarify the denominator for all PT-FT ratios.
  2. [§III.A3, Fig. 2] The relationship among the contrastive encoder E_c, the momentum encoder, and the projection heads is unclear. The caption says the momentum encoder outputs are compared with the feature decoder outputs, while the text says E_c processes X_k^B. Specify which encoder is the momentum encoder, whether gradients flow through it, and how negative examples are collected.
  3. [§IV.D3, Fig. 6] The text says Fig. 6 visualizes reconstruction results, but the caption describes segmentation outputs; align the text, the caption, and the figure panels.
  4. [§III.A2, Fig. 5, §IV.C] There are several typos that should be corrected: "employes" in §III.A2, "pPT-FT ratios" in the Fig. 5 caption, and "Pytorch" in §IV.C.
  5. [§IV.D2] The Bland-Altman and PCC analyses are reported only for CM-UNet at two PT-FT ratios; reporting the same clinical metrics for the best SSL baseline would make the clinical-validity comparison more informative.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: CM-UNet's performance claims are empirical measurements on a held-out test set, and no derivation step reduces to its own inputs.

full rationale

The paper is an empirical evaluation of a self-supervised pre-training method followed by supervised fine-tuning for coronary artery segmentation. The central quantitative claims, such as Dice 0.626 for CM-UNet versus 0.375 for a supervised UNet at a 79:1 pre-training/fine-tuning split, are measured on a held-out test set of 348 images that are not used during pre-training, fine-tuning, or hyperparameter selection. Hyperparameters are selected by three-fold cross-validation on the fine-tuning set only, so no test-label information enters model selection. The loss functions in Eqs. (1) and (3) are standard reconstruction and InfoNCE objectives, and the evaluation metrics in Eqs. (5)-(7) are independent overlap and distance measures; none of these equations defines the reported outcome in terms of a fitted parameter that is then renamed a prediction. The paper cites works by co-authors, notably the FAME2 dataset [6] and AngioPy [13], but these are data and related-work references, not load-bearing justifications for the method's validity. The paper also self-reports a missing ablation of the reconstruction and contrastive components; this is an incompleteness that affects attribution of the gain to specific components, but it is not a circular reduction. The parameter-count mismatch (121.49M for CM-UNet versus 31.04-37.70M for baselines) is a potential confound in comparing methods, but it concerns experimental fairness and causal interpretation, not circularity, because the comparison itself is still an empirical measurement on held-out data. Overall, no derivation chain collapses into its inputs, and no self-citation chain forces the conclusion; the circularity burden is therefore minimal.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The paper introduces CM-UNet, an architecture combining existing components (UNet, CMAE, momentum encoder, feature decoder). No new physical entities are postulated. The free parameters are standard ML hyperparameters, several of which are unreported. The axioms are domain assumptions about data quality and the transfer learning premise, plus an ad hoc design choice for the loss combination.

free parameters (5)
  • InfoNCE temperature tau = not reported
    Temperature in Eq. (3) scales cosine similarities for the contrastive loss; chosen by hand without reported sensitivity analysis.
  • Contrastive loss weight lambda_c = not reported
    Weight in L = Lr + lambda_c * Lc (Section III.A.3); no ablation or optimization details provided.
  • Mask coverage ratio = 65%
    Masking ratio in pre-training (Section IV.C); chosen without reported optimization or sensitivity analysis.
  • Mask patch size = 16x16
    Patch size for masking (Section IV.C); chosen without reported sensitivity analysis.
  • Fine-tuning learning rate and batch size = selected via CV from ranges 1e-1 to 1e-5, batch 16 or 32
    Hyperparameter tuning on the fine-tuning set via three-fold cross-validation (Section IV.C); final values not explicitly stated.
assumptions (4)
  • domain assumption Manual annotations of coronary arteries in the FAME2 dataset are accurate and consistent enough to serve as ground truth.
    All training and evaluation relies on these labels (Section IV.A). If the labels are noisy or inconsistent, the reported Dice and clinical metrics lose meaning.
  • domain assumption The preprocessing pipeline (unsharp masking, corner inpainting, z-score normalization) preserves vessel boundaries and does not bias comparisons across models.
    Applied to all images before training and testing (Section IV.A); no analysis of its effect on segmentation is provided.
  • domain assumption Self-supervised pre-training on unlabeled ICA images from the same dataset provides a useful initialization for downstream segmentation.
    This is the transfer-learning premise underlying the entire method (Section III). It is tested empirically but not justified theoretically.
  • ad hoc to paper The weighted combination of reconstruction and contrastive losses is an appropriate objective for learning vessel representations.
    L = Lr + lambda_c * Lc (Section III.A.3) is a design choice without ablation or theoretical derivation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CM-UNet: A Self-Supervised Learning-Based Model for Coronary Artery Segmentation in X-Ray Angiography." pith.science (2026). https://pith.science/paper/GBEFWGVZ

@misc{pith2026250717779,
  author       = {Pith},
  title        = {Pith review of: CM-UNet: A Self-Supervised Learning-Based Model for Coronary Artery Segmentation in X-Ray Angiography},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GBEFWGVZ}},
  note         = {Machine review of arXiv:2507.17779}
}
read the original abstract

Accurate segmentation of coronary arteries remains a significant challenge in clinical practice, hindering the ability to effectively diagnose and manage coronary artery disease. The lack of large, annotated datasets for model training exacerbates this issue, limiting the development of automated tools that could assist radiologists. To address this, we introduce CM-UNet, which leverages self-supervised pre-training on unannotated datasets and transfer learning on limited annotated data, enabling accurate disease detection while minimizing the need for extensive manual annotations. Fine-tuning CM-UNet with only 18 annotated images instead of 500 resulted in a 15.2% decrease in Dice score, compared to a 46.5% drop in baseline models without pre-training. This demonstrates that self-supervised learning can enhance segmentation performance and reduce dependence on large datasets. This is one of the first studies to highlight the importance of self-supervised learning in improving coronary artery segmentation from X-ray angiography, with potential implications for advancing diagnostic accuracy in clinical practice. By enhancing segmentation accuracy in X-ray angiography images, the proposed approach aims to improve clinical workflows, reduce radiologists' workload, and accelerate disease detection, ultimately contributing to better patient outcomes. The source code is publicly available at https://github.com/CamilleChallier/Contrastive-Masked-UNet.

Figures

Figures reproduced from arXiv: 2507.17779 by the authors.

Figure 1
Figure 1. Overview of the CM-UNet training pipeline, [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Illustration of Contrastive Masked UNet (CM-UNet) method for self-supervised pre-training. The framework consists [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Evalution of the Dice score and Hausdorff distance [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (3 more)
Figure 5
Figure 5. Figure 5: Segmentation examples of major vessels using UNet [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 4
Figure 4. Figure 4: Comparison of CM-UNet diameter predictions with [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 6
Figure 6. Figure 6: Segmentation examples of major vessels comparing [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

28 extracted references · 27 canonical work pages

  1. [1]

    Dataset for Automatic Region-based Coronary Artery Disease Diagnostics Using X-Ray Angiography Images,

    M. Popov et al. , “Dataset for Automatic Region-based Coronary Artery Disease Diagnostics Using X-Ray Angiography Images,” Sci- entific Data, vol. 11, p. 20, 2024

  2. [2]

    Image quality in coronary CT angiography: challenges and technical solutions,

    O. Ghekiere et al. , “Image quality in coronary CT angiography: challenges and technical solutions,” The British Journal of Radiology , vol. 90, 2017

  3. [3]

    Extraction of coronary arterial tree using cine x-ray angiograms,

    C.-Y . Lin and Y .-T. Ching, “Extraction of coronary arterial tree using cine x-ray angiograms,” Biomedical Engineering: Applications, Basis and Communications, vol. 17, pp. 111–120, 2005

  4. [4]

    Self-supervised learning for medical image classification: a systematic review and implementation guidelines,

    S.-C. Huang, A. Pareek, M. Jensen, M. P. Lungren, S. Yeung, and A. S. Chaudhari, “Self-supervised learning for medical image classification: a systematic review and implementation guidelines,” npj Digital Medicine, vol. 6, pp. 1–16, 2023

  5. [5]

    Contrastive Masked Autoencoders are Stronger Vision Learners,

    Z. Huang et al. , “Contrastive Masked Autoencoders are Stronger Vision Learners,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 46, pp. 2506–2517, 2024

  6. [6]

    Fractional flow reserve-guided PCI versus medical therapy in stable coronary disease,

    B. De Bruyne et al. , “Fractional flow reserve-guided PCI versus medical therapy in stable coronary disease,” The New England Journal of Medicine, vol. 367, pp. 991–1001, 2012

  7. [7]

    ImageNet Classifi- cation with Deep Convolutional Neural Networks,

    A. Krizhevsky, I. Sutskever, and G. E. Hinton, “ImageNet Classifi- cation with Deep Convolutional Neural Networks,” in Advances in Neural Information Processing Systems , vol. 25, 2012

  8. [8]

    U-Net: Convolutional Networks for Biomedical Image Segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-Net: Convolutional Networks for Biomedical Image Segmentation,” in Medical Image Computing and Computer-Assisted Intervention – MICCAI 2015 , pp. 234–241, Cham, 2015

Show all 28 references
  1. [9]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , pp. 770–778, 2016

  2. [10]

    Densely connected convolutional networks,

    G. Huang, Z. Liu, L. Van Der Maaten, and K. Q. Weinberger, “Densely connected convolutional networks,” in 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , pp. 2261–2269, 2017

  3. [11]

    Deep learning segmentation of major vessels in X-ray coronary angiography,

    S. Yang et al., “Deep learning segmentation of major vessels in X-ray coronary angiography,” Scientific Reports, vol. 9, p. 16897, 2019

  4. [12]

    AngioNet: a convolutional neural network for vessel segmentation in X-ray angiography,

    K. Iyer et al. , “AngioNet: a convolutional neural network for vessel segmentation in X-ray angiography,” Scientific Reports , p. 18066, 2021

  5. [13]

    AngioPy Segmentation: An open-source, user- guided deep learning tool for coronary artery segmentation,

    T. Mahendiran et al., “AngioPy Segmentation: An open-source, user- guided deep learning tool for coronary artery segmentation,” Interna- tional Journal of Cardiology , vol. 418, p. 132598, 2025

  6. [14]

    Echo-Rhythm Net: Semi-Supervised Learning For Automatic Detection of Atrial Fibrillation in Echocardiography,

    F. T. Dezaki et al. , “Echo-Rhythm Net: Semi-Supervised Learning For Automatic Detection of Atrial Fibrillation in Echocardiography,” in 2021 IEEE 18th International Symposium on Biomedical Imaging (ISBI), pp. 110–113, 2021

  7. [15]

    Self-supervised contrastive video-speech representation learning for ultrasound,

    J. Jiao, Y . Cai, M. Alsharid, L. Drukker, A. T. Papageorghiou, and J. A. Noble, “Self-supervised contrastive video-speech representation learning for ultrasound,” in Medical Image Computing and Computer Assisted Intervention – MICCAI 2020 , p. 534–543, 2020

  8. [16]

    Models Gen- esis: Generic Autodidactic Models for 3D Medical Image Analysis,

    Z. Zhou, V . Sodha, J. Pang, M. B. Gotway, and J. Liang, “Models Gen- esis: Generic Autodidactic Models for 3D Medical Image Analysis,” Medical Image Analysis , vol. 67, p. 101840, 2021

  9. [17]

    A simple frame- work for contrastive learning of visual representations,

    T. Chen, S. Kornblith, M. Norouzi, and G. Hinton, “A simple frame- work for contrastive learning of visual representations,” inProceedings of the 37th International Conference on Machine Learning , 2020

  10. [18]

    Momentum con- trast for unsupervised visual representation learning,

    K. He, H. Fan, Y . Wu, S. Xie, and R. Girshick, “Momentum con- trast for unsupervised visual representation learning,” in IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pp. 9726–9735, 2020

  11. [19]

    Self-supervised pre-training with contrastive and masked autoencoder methods for dealing with small datasets in deep learning for medical imaging,

    D. Wolf et al. , “Self-supervised pre-training with contrastive and masked autoencoder methods for dealing with small datasets in deep learning for medical imaging,” Scientific Reports, p. 20260, 2023

  12. [20]

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

    K. Chaitanya, E. Erdil, N. Karani, and E. Konukoglu, “Contrastive learning of global and local features for medical image segmentation with limited annotations,” in Proceedings of the 34th International Conference on Neural Information Processing Systems , 2020

  13. [21]

    Masked autoencoders are scalable vision learners,

    K. He, X. Chen, S. Xie, Y . Li, P. Doll ´ar, and R. Girshick, “Masked autoencoders are scalable vision learners,” in 2022 IEEE/CVF Con- ference on Computer Vision and Pattern Recognition (CVPR) , pp. 15 979–15 988, 2022

  14. [22]

    Designing BERT for Convolutional Networks: Sparse and Hierarchical Masked Modeling,

    K. Tian, Y . Jiang, Q. Diao, C. Lin, L. Wang, and Z. Yuan, “Designing BERT for Convolutional Networks: Sparse and Hierarchical Masked Modeling,” 2023

  15. [23]

    Self-supervised vessel segmentation via adversarial learning,

    Y . Ma et al. , “Self-supervised vessel segmentation via adversarial learning,” in 2021 IEEE/CVF International Conference on Computer Vision (ICCV), pp. 7516–7525, 2021

  16. [24]

    Pretrained subtraction and segmentation model for coronary angiograms,

    Y . Zeng, H. Liu, J. Hu, Z. Zhao, and Q. She, “Pretrained subtraction and segmentation model for coronary angiograms,” Scientific Reports, vol. 14, p. 19888, 2024

  17. [25]

    Representation Learning with Contrastive Predictive Coding,

    A. v. d. Oord, Y . Li, and O. Vinyals, “Representation Learning with Contrastive Predictive Coding,” 2019

  18. [26]

    cldice-a novel topology-preserving loss function for tubular structure segmentation,

    S. Shit et al. , “cldice-a novel topology-preserving loss function for tubular structure segmentation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pp. 16 560– 16 569, 2021

  19. [27]

    A modified Hausdorff distance for object matching,

    M.-P. Dubuisson and A. Jain, “A modified Hausdorff distance for object matching,” in Proceedings of 12th International Conference on Pattern Recognition, vol. 1, pp. 566–568, 1994

  20. [28]

    Statistical methods for assessing agreement between two methods of clinical measurement,

    J. M. Bland and D. G. Altman, “Statistical methods for assessing agreement between two methods of clinical measurement,” Lancet, pp. 307–310, 1986

Pith tools

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