pith. sign in

arxiv: 2606.06725 · v1 · pith:4A5KPFBAnew · submitted 2026-06-04 · 📡 eess.IV · cs.CV

Compute-Optimal Network Design for Echocardiography Myocardial Segmentation and Perfusion Quantification using Neural Scaling Laws

Clara Rodrigo Gonz\'alez , Matthieu Toulemonde , Lasha Gvinianidze , Cameron A. B. Smith , Oscar Bates , Roxy Senior , Fu Siong Ng , Meng-Xing Tang This is my paper

Pith reviewed 2026-06-27 23:00 UTC · model grok-4.3

classification 📡 eess.IV cs.CV
keywords neural scaling lawsmyocardial segmentationechocardiographycontrast-enhanced ultrasoundperfusion quantificationcompute-optimal designCAMUS dataset
0
0 comments X

The pith

Neural scaling laws fitted on data subsets predict full-dataset performance and select compact networks that match expert cardiologist results on myocardial perfusion quantification.

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

The paper shows that neural scaling laws can be fitted to segmentation loss on subsets of echocardiography data and then extrapolated to identify the model size that minimizes loss on the full dataset. This extrapolation correctly forecasts test performance on the CAMUS dataset, leading to two networks that reach state-of-the-art accuracy while using 240 times fewer parameters than prior models. The same scaling relationship transfers to a 25-patient contrast-enhanced ultrasound dataset, offset only by a constant bias in predicted loss. Segmentations produced by these models yield myocardial perfusion parameters statistically indistinguishable from those computed by a senior cardiologist. The work therefore supplies a data-driven route to choose network size when annotated medical images remain scarce.

Core claim

Extrapolation based on the scaling law is predictive of test loss at the full dataset size, allowing selection of two networks that obtained state-of-the-art performance on CAMUS with a 240-fold reduction in parameter count. The gradient of the scaling law transfers from CAMUS to the CEUS dataset with a bias in the predicted losses. The automatically segmented masks perform equivalently to a senior cardiologist in myocardial perfusion quantification.

What carries the argument

Neural scaling laws that relate test loss to model parameter count, fitted on performance from data subsets to extrapolate optimal network sizes for myocardial segmentation.

Load-bearing premise

The scaling law fitted on performance from data subsets will accurately extrapolate to the full dataset size and its gradient will transfer to the CEUS dataset with only a bias that does not invalidate optimal model selection.

What would settle it

Train the two extrapolated optimal networks on the complete CAMUS training set and check whether their measured test loss matches the scaling-law prediction or whether substantially larger networks achieve lower loss.

Figures

Figures reproduced from arXiv: 2606.06725 by Cameron A. B. Smith, Clara Rodrigo Gonz\'alez, Fu Siong Ng, Lasha Gvinianidze, Matthieu Toulemonde, Meng-Xing Tang, Oscar Bates, Roxy Senior.

Figure 1
Figure 1. Figure 1: Our tunable U-Net is defined by parameters L, which determines the number of layers, and W, which [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The test BCE at the inflection point was predicted for each model using the power law fit on models trained on [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Example segmentations in CAMUS dataset using 3L2W-UNet and 4L5W-UNet, as well as the external SOTA [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Randomly selected example segmentations in CEUS dataset using 3L2W-UNet (61.1k parameters) and [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
read the original abstract

Myocardial perfusion quantification using contrast-enhanced ultrasound offers a bedside non-ionizing alternative to nuclear imaging modalities. However, its clinical adoption is hindered by time-consuming manual labelling. Automated segmentation has proved challenging due to a paucity of in-domain training data. Adapting strategies currently used to optimise large language models for large datasets, we apply neural scaling laws to predict network performance for myocardial segmentation. We extrapolate performance on subsets of the data to determine optimal network size on the CAMUS echocardiography dataset and a 25-patient contrast-enhanced ultrasound (CEUS) dataset. Finally, we validate the clinical utility of our models by comparing the final myocardial perfusion parameters with those obtained by a senior cardiologist. Extrapolation based on the scaling law is predictive of test loss at the full dataset size, allowing us to select two networks that obtained state-of-the-art performance on CAMUS with a 240-fold reduction in parameter count. We observe the gradient of the scaling law transfers from CAMUS to the CEUS dataset with a bias in the predicted losses. The automatically segmented masks perform equivalently to a senior cardiologist in myocardial perfusion quantification. These results establish neural scaling laws as a practical tool for data-driven compute-optimal model design for small imaging datasets.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper claims that neural scaling laws fitted to test losses on data subsets of the CAMUS echocardiography dataset can be extrapolated to predict performance at full dataset size. This enables selection of two networks achieving state-of-the-art myocardial segmentation on CAMUS with a 240-fold reduction in parameter count. The scaling-law gradient transfers to a 25-patient CEUS dataset after constant bias correction, and the resulting automatic segmentations produce myocardial perfusion quantification parameters equivalent to those from a senior cardiologist.

Significance. If the subset-based extrapolation reliably identifies compute-optimal models that generalize and match clinical performance, the work would offer a practical, data-driven method for model-size selection in data-scarce medical imaging tasks, potentially reducing compute demands without sacrificing accuracy.

major comments (2)
  1. [Abstract] Abstract and scaling-law section: the central claim that extrapolation from subset performances is predictive of full-dataset test loss and enables optimal model selection rests on the power-law fit being robust; with high run-to-run variance typical in echocardiography segmentation and typically few subset sizes available for fitting, sensitivity of the fitted coefficients to exact subset choices or noise could invalidate both the CAMUS model selection and the CEUS transfer claim.
  2. [Results] CEUS transfer paragraph: the assumption that only a constant bias correction is needed for the scaling-law gradient to transfer from CAMUS to the 25-patient CEUS set is load-bearing for claiming the same networks remain optimal; without an explicit check that the bias-adjusted predictions correctly rank models on held-out CEUS performance, the transfer step risks circularity.
minor comments (2)
  1. Clarify the exact number of subset sizes used to fit the scaling law and report the uncertainty or variance in the fitted exponents.
  2. Provide the precise definition of the scaling law functional form (e.g., the equation relating loss to model size and data size) in the methods.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments, which highlight important aspects of robustness and validation in our scaling-law approach. We address each major comment below with point-by-point responses. Where the concerns identify areas for clarification or additional checks, we indicate that revisions will be made.

read point-by-point responses

  1. Referee: [Abstract] Abstract and scaling-law section: the central claim that extrapolation from subset performances is predictive of full-dataset test loss and enables optimal model selection rests on the power-law fit being robust; with high run-to-run variance typical in echocardiography segmentation and typically few subset sizes available for fitting, sensitivity of the fitted coefficients to exact subset choices or noise could invalidate both the CAMUS model selection and the CEUS transfer claim.

    Authors: We agree that sensitivity to subset choice and run variance is a valid concern for power-law fits with limited points. However, the manuscript already validates the extrapolation by showing that predictions from subset fits closely match the measured test loss on the full CAMUS dataset for the selected models (see scaling-law section and Figure 3). This empirical match provides evidence that the fit is sufficiently robust for model selection in this setting. To further address the referee's point, we will add a sensitivity analysis in the revision, including fits across different random subset selections and reporting the resulting variation in predicted optimal model sizes. revision: yes

  2. Referee: [Results] CEUS transfer paragraph: the assumption that only a constant bias correction is needed for the scaling-law gradient to transfer from CAMUS to the 25-patient CEUS set is load-bearing for claiming the same networks remain optimal; without an explicit check that the bias-adjusted predictions correctly rank models on held-out CEUS performance, the transfer step risks circularity.

    Authors: The manuscript reports that the scaling-law gradient observed on CAMUS transfers to CEUS after a constant bias correction, and the same networks selected via CAMUS scaling laws achieve strong performance on the CEUS task. We acknowledge that an explicit verification that the bias-adjusted predictions preserve model ranking on held-out CEUS data would strengthen the transfer claim and reduce any perception of circularity. We will add this check in the revision by training a small set of models on CEUS subsets, applying the bias correction, and confirming the ranking matches actual held-out performance. revision: yes

Circularity Check

0 steps flagged

No significant circularity in scaling-law extrapolation

full rationale

The paper fits scaling laws to measured test losses on data subsets of CAMUS, extrapolates to predict full-dataset performance, selects two architectures on that basis, and then reports that the selected models achieve SOTA when trained on the full set. This is a standard predictive use of scaling laws rather than a reduction by construction; the full-dataset losses serve as an independent check. No load-bearing self-citations, no fitted parameters renamed as predictions, and no ansatz or uniqueness claim imported from the authors' prior work appear in the provided text. The CEUS transfer with bias correction is an empirical observation, not a definitional loop.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the validity of neural scaling laws transferring from large language models to medical imaging tasks and the assumption that subset performance predicts full dataset behavior without major domain-specific deviations.

free parameters (1)
  • scaling law coefficients
    The scaling law is fitted to performance on data subsets to extrapolate to full dataset.
axioms (1)
  • domain assumption Performance on data subsets follows a predictable scaling law that can be extrapolated to larger data sizes
    Central to the extrapolation method described in the abstract.

pith-pipeline@v0.9.1-grok · 5787 in / 1447 out tokens · 30996 ms · 2026-06-27T23:00:50.716444+00:00 · methodology

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

44 extracted references · 12 canonical work pages · 4 internal anchors

  1. [1]

    Marc Dewey, Maria Siebes, Marc Kachelrieß, Klaus F. Kofoed, Pál Maurovich-Horvat, Konstantin Nikolaou, Wenjia Bai, Andreas Kofler, Robert Manka, Sebastian Kozerke, Amedeo Chiribiri, Tobias Schaeffter, Florian Michallek, Frank Bengel, Stephan Nekolla, Paul Knaapen, Mark Lubberink, Roxy Senior, Meng-Xing Tang, Jan J. Piek, Tim van de Hoef, Johannes Martens,...

    2020

  2. [2]

    Diagnostic Value of Extracellular V olume Quantification and Myocardial Perfusion Analysis at CT in Cardiac Amyloidosis.Radiology, 300(2):326–335, August 2021

    Jean-François Deux, Refaat Nouri, Vania Tacher, Amira Zaroui, Haytham Derbel, Islem Sifaoui, Virgile Chevance, Fourat Ridouani, Arnault Galat, Mounira Kharoubi, Silvia Oghina, Soulef Guendouz, Etienne Audureau, Em- manuel Teiger, Hicham Kobeiter, and Thibaud Damy. Diagnostic Value of Extracellular V olume Quantification and Myocardial Perfusion Analysis a...

    2021

  3. [3]

    Ultrasound study of right ventricular myocardial perfusion and functional changes in hypertrophic cardiomyopathy.BMC Cardiovascular Disorders, 24(1):63, January 2024

    Shan Cao, Lingjie Yang, Liyun Liu, Yuming Mu, and Lina Guan. Ultrasound study of right ventricular myocardial perfusion and functional changes in hypertrophic cardiomyopathy.BMC Cardiovascular Disorders, 24(1):63, January 2024

    2024

  4. [4]

    Einstein, Shepard D

    Andrew J. Einstein, Shepard D. Weiner, Adam Bernheim, Michal Kulon, Sabahat Bokhari, Lynne L. Johnson, Jeffrey W. Moses, and Stephen Balter. Multiple Testing, Cumulative Radiation Dose, and Clinical Indications in Patients Undergoing Myocardial Perfusion Imaging.JAMA : the journal of the American Medical Association, 304(19):2137–2144, November 2010

    2010

  5. [5]

    Myocardial perfusion scans: projected population cancer risks from current levels of use in the United States.Circulation, 122(23):2403–2410, December 2010

    Amy Berrington de Gonzalez, Kwang-Pyo Kim, Rebecca Smith-Bindman, and Dorothea McAreavey. Myocardial perfusion scans: projected population cancer risks from current levels of use in the United States.Circulation, 122(23):2403–2410, December 2010

    2010

  6. [6]

    Jerome, Peter L

    Scott D. Jerome, Peter L. Tilkemeier, Mary B. Farrell, and Leslee J. Shaw. Nationwide Laboratory Adherence to Myocardial Perfusion Imaging Radiation Dose Reduction Practices.JACC: Cardiovascular Imaging, 8(10):1170– 1176, October 2015

    2015

  7. [7]

    Scannell, and Anedeo Chiribiri

    Ebraham Alskaf, Utkarsh Dutta, Cian M. Scannell, and Anedeo Chiribiri. Deep learning applications in myocardial perfusion imaging, a systematic review and meta-analysis.Informatics in Medicine Unlocked, 32:101055, January 2022. 12 APREPRINT- JUNE8, 2026

    2022

  8. [8]

    Kasprzak

    Roxy Senior, Antonella Moreo, Nicola Gaibazzi, Luciano Agati, Klaus Tiemann, Bharati Shivalkar, Stephan von Bardeleben, Leonarda Galiuto, Hervé Lardoux, Giuseppe Trocino, Ignasi Carrió, Dominique Le Guludec, Gianmario Sambuceti, Harald Becher, Paolo Colonna, Folkert Ten Cate, Ezio Bramucci, Ariel Cohen, Gianpaolo Bezante, Costantina Aggeli, and Jaroslaw D...

    2013

  9. [9]

    Mina Amiri, Rupert Brooks, and Hassan Rivaz. Fine-Tuning U-Net for Ultrasound Image Segmentation: Dif- ferent Layers, Different Outcomes.IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, 67(12):2510–2518, December 2020

    2020

  10. [10]

    Deep Neural Architectures for Contrast Enhanced Ultrasound (CEUS) Focal Liver Lesions Automated Diagnosis.Sensors, 21(12):4126, January 2021

    C˘at˘alin Daniel C˘aleanu, Cristina Laura Sîrbu, and Georgiana Simion. Deep Neural Architectures for Contrast Enhanced Ultrasound (CEUS) Focal Liver Lesions Automated Diagnosis.Sensors, 21(12):4126, January 2021. Number: 12

    2021

  11. [11]

    Joint Segmentation and Differential Diagnosis of Thyroid Nodule in Contrast-Enhanced Ultrasound Images.IEEE Transactions on Biomedical Engineering, 70(9):2722–2732, September 2023

    Fang Chen, Haojie Han, Peng Wan, Hongen Liao, Chunrui Liu, and Daoqiang Zhang. Joint Segmentation and Differential Diagnosis of Thyroid Nodule in Contrast-Enhanced Ultrasound Images.IEEE Transactions on Biomedical Engineering, 70(9):2722–2732, September 2023

    2023

  12. [12]

    Semantic segmentation method for myocardial contrast echocardiogram based on DeepLabV3+ deep learning architecture.Mathematical Biosciences and Engineering, 20(2):2081–2093, 2023

    Huan Cheng, Jucheng Zhang, Yinglan Gong, Zhaoxia Pu, Jun Jiang, Yonghua Chu, Ling Xia, Huan Cheng, Jucheng Zhang, Yinglan Gong, Zhaoxia Pu, Jun Jiang, Yonghua Chu, and Ling Xia. Semantic segmentation method for myocardial contrast echocardiogram based on DeepLabV3+ deep learning architecture.Mathematical Biosciences and Engineering, 20(2):2081–2093, 2023....

    2081

  13. [13]

    Lee R. Dice. Measures of the Amount of Ecologic Association Between Species.Ecology, 26(3):297–302, 1945. _eprint: https://esajournals.onlinelibrary.wiley.com/doi/pdf/10.2307/1932409

    work page doi:10.2307/1932409 1945

  14. [14]

    Left-Ventricular V olume Estimation in Contrast-Enhanced Echocardiography Using Deep Learning

    Jieyu Hu, Erik Smistad, Bjørnar Grenne, Espen Holte, Håvard Dalen, and Lasse Lovstakken. Left-Ventricular V olume Estimation in Contrast-Enhanced Echocardiography Using Deep Learning. In2023 IEEE International Ultrasonics Symposium (IUS), pages 1–4, September 2023

    2023

  15. [15]

    Scaling Laws for Neural Language Models

    Jared Kaplan, Sam McCandlish, Tom Henighan, Tom B. Brown, Benjamin Chess, Rewon Child, Scott Gray, Alec Radford, Jeffrey Wu, and Dario Amodei. Scaling Laws for Neural Language Models, January 2020. arXiv:2001.08361 [cs]

    work page internal anchor Pith review Pith/arXiv arXiv 2020

  16. [16]

    Training Compute-Optimal Large Language Models

    Jordan Hoffmann, Sebastian Borgeaud, Arthur Mensch, Elena Buchatskaya, Trevor Cai, Eliza Rutherford, Diego de Las Casas, Lisa Anne Hendricks, Johannes Welbl, Aidan Clark, Tom Hennigan, Eric Noland, Katie Millican, George van den Driessche, Bogdan Damoc, Aurelia Guy, Simon Osindero, Karen Simonyan, Erich Elsen, Jack W. Rae, Oriol Vinyals, and Laurent Sifre...

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  17. [17]

    Scaling Laws For Deep Learning Based Image Reconstruction, February 2023

    Tobit Klug and Reinhard Heckel. Scaling Laws For Deep Learning Based Image Reconstruction, February 2023. arXiv:2209.13435 [eess]

    work page arXiv 2023

  18. [18]

    Manmatha, Ashwin Swaminathan, Zhuowen Tu, Stefano Ermon, and Stefano Soatto

    Hao Li, Yang Zou, Ying Wang, Orchid Majumder, Yusheng Xie, R. Manmatha, Ashwin Swaminathan, Zhuowen Tu, Stefano Ermon, and Stefano Soatto. On the Scalability of Diffusion-based Text-to-Image Generation. In2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 9400–9409, Seattle, WA, USA, June 2024. IEEE

    2024

  19. [19]

    Revisiting model scaling with a U-net benchmark for 3D medical image segmentation.Scientific Reports, 15(1):29795, August 2025

    Ziyan Huang, Jin Ye, Haoyu Wang, Zhongying Deng, Zhikai Yang, Yanzhou Su, Jie Liu, Tianbin Li, Yun Gu, Shaoting Zhang, Yu Qiao, Lixu Gu, and Junjun He. Revisiting model scaling with a U-net benchmark for 3D medical image segmentation.Scientific Reports, 15(1):29795, August 2025

    2025

  20. [20]

    Fubao Zhu, Longxi Li, Jinyu Zhao, Chen Zhao, Shaojie Tang, Jiaofen Nan, Yanting Li, Zhongqiang Zhao, Jianzhou Shi, Zenghong Chen, Chuang Han, Zhixin Jiang, and Weihua Zhou. A new method incorporating deep learning with shape priors for left ventricular segmentation in myocardial perfusion SPECT images.Computers in Biology and Medicine, 160:106954, June 2023

    2023

  21. [21]

    Seyed Mohammad Entezarmahdi, Reza Faghihi, Mehran Yazdi, Negar Shahamiri, Parham Geramifar, and Mahdi Haghighatafshar. QCard-NM: Developing a semiautomatic segmentation method for quantitative analysis of the right ventricle in non-gated myocardial perfusion SPECT imaging.EJNMMI Physics, 10(1):21, March 2023

    2023

  22. [22]

    How to perform an ultrasound contrast myocardial perfusion examination?European Heart Journal - Cardiovascular Imaging, 23(6):727–729, June 2022

    Bernard Cosyns, Andreas Helfen, Howard Leong-Poi, and Roxy Senior. How to perform an ultrasound contrast myocardial perfusion examination?European Heart Journal - Cardiovascular Imaging, 23(6):727–729, June 2022. 13 APREPRINT- JUNE8, 2026

    2022

  23. [23]

    Image Segmentation Using Deep Learning: A Survey.IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(7):3523–3542, July 2022

    Shervin Minaee, Yuri Boykov, Fatih Porikli, Antonio Plaza, Nasser Kehtarnavaz, and Demetri Terzopoulos. Image Segmentation Using Deep Learning: A Survey.IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(7):3523–3542, July 2022

    2022

  24. [24]

    Modality specific U-Net variants for biomedical image segmentation: A survey.Artificial Intelligence Review, 55(7):5845–5889, October 2022

    Narinder Singh Punn and Sonali Agarwal. Modality specific U-Net variants for biomedical image segmentation: A survey.Artificial Intelligence Review, 55(7):5845–5889, October 2022. arXiv:2107.04537 [eess]

    work page arXiv 2022

  25. [25]

    UNet 3+: A Full-Scale Connected UNet for Medical Image Segmentation, April 2020

    Huimin Huang, Lanfen Lin, Ruofeng Tong, Hongjie Hu, Qiaowei Zhang, Yutaro Iwamoto, Xianhua Han, Yen-Wei Chen, and Jian Wu. UNet 3+: A Full-Scale Connected UNet for Medical Image Segmentation, April 2020. arXiv:2004.08790 [eess]

    work page arXiv 2020

  26. [26]

    Medical Image Segmentation Review: The success of U-Net, November 2022

    Reza Azad, Ehsan Khodapanah Aghdam, Amelie Rauland, Yiwei Jia, Atlas Haddadi Avval, Afshin Bozorgpour, Sanaz Karimijafarbigloo, Joseph Paul Cohen, Ehsan Adeli, and Dorit Merhof. Medical Image Segmentation Review: The success of U-Net, November 2022. arXiv:2211.14830 [eess]

    work page arXiv 2022

  27. [27]

    Deep Learning for Segmentation Using an Open Large-Scale Dataset in 2D Echocardiography.IEEE transactions on medical imaging, 38(9):2198–2210, September 2019

    Sarah Leclerc, Erik Smistad, Joao Pedrosa, Andreas Ostvik, Frederic Cervenansky, Florian Espinosa, Torvald Espeland, Erik Andreas Rye Berg, Pierre-Marc Jodoin, Thomas Grenier, Carole Lartizien, Jan Dhooge, Lasse Lovstakken, and Olivier Bernard. Deep Learning for Segmentation Using an Open Large-Scale Dataset in 2D Echocardiography.IEEE transactions on med...

    2019

  28. [28]

    Myocardial Segmentation of Contrast Echocardiograms Using Random Forests Guided by Shape Model

    Yuanwei Li, Chin Pang Ho, Navtej Chahal, Roxy Senior, and Meng-Xing Tang. Myocardial Segmentation of Contrast Echocardiograms Using Random Forests Guided by Shape Model, June 2018. arXiv:1806.07490 [cs]

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  29. [29]

    Vaitkus, Marek Belohlavek, Vinayak Krishnamurthy, and Iman Borazjani

    Taeouk Kim, Mohammadali Hedayat, Veronica V . Vaitkus, Marek Belohlavek, Vinayak Krishnamurthy, and Iman Borazjani. Automatic segmentation of the left ventricle in echocardiographic images using convolutional neural networks.Quantitative Imaging in Medicine and Surgery, 11(5):1763–1781, May 2021

    2021

  30. [30]

    Segmentation of Anatomical Structures of the Left Heart from Echocardiographic Images Using Deep Learning

    Mhd Jafar Mortada, Selene Tomassini, Haidar Anbar, Micaela Morettini, Laura Burattini, and Agnese Sbrollini. Segmentation of Anatomical Structures of the Left Heart from Echocardiographic Images Using Deep Learning. Diagnostics (Basel, Switzerland), 13(10):1683, May 2023

    2023

  31. [31]

    Compositional Segmentation of Cardiac Images Leveraging Metadata, October 2024

    Abbas Khan, Muhammad Asad, Martin Benning, Caroline Roney, and Gregory Slabaugh. Compositional Segmentation of Cardiac Images Leveraging Metadata, October 2024. arXiv:2410.23130 [eess]

    work page arXiv 2024

  32. [32]

    CoST-UNet: Convolution and swin transformer based deep learning architecture for cardiac segmentation.Biomedical Signal Processing and Control, 96:106633, October 2024

    Md Rabiul Islam, Marwa Qaraqe, and Erchin Serpedin. CoST-UNet: Convolution and swin transformer based deep learning architecture for cardiac segmentation.Biomedical Signal Processing and Control, 96:106633, October 2024

    2024

  33. [33]

    Anatomically accurate cardiac segmentation using Dense Associative Networks

    Zahid Ullah and Jihie Kim. Anatomically accurate cardiac segmentation using Dense Associative Networks. Engineering Applications of Artificial Intelligence, 162:112742, December 2025

    2025

  34. [34]

    DAF- Mamba: Dynamic Selective and Adaptive Fused Mamba for Cardiac Image Segmentation.Pattern Recognition, page 112664, November 2025

    Yonglin Chen, Yixiang Wang, Zhongyuan Liu, Yuyan Weng, Chihui Long, Yalong Yang, and Jinhui Tang. DAF- Mamba: Dynamic Selective and Adaptive Fused Mamba for Cardiac Image Segmentation.Pattern Recognition, page 112664, November 2025

    2025

  35. [35]

    Diego D. B. Carvalho, Zeynettin Akkus, Stijn C. H. van den Oord, Arend F. L. Schinkel, Antonius F. W. van der Steen, Wiro J. Niessen, Johan G. Bosch, and Stefan Klein. Lumen Segmentation and Motion Estimation in B-Mode and Contrast-Enhanced Ultrasound Images of the Carotid Artery in Patients With Atherosclerotic Plaque. IEEE Transactions on Medical Imagin...

    2015

  36. [36]

    Deep-learning-based washout classification for decision support in contrast-enhanced ultrasound examinations of the liver.Journal of Medical Imaging, 12(4):044502, July 2025

    Hannah Strohm, Sven Rothlübbers, Jürgen Jenne, Dirk-André Clevert, Thomas Fischer, Niklas Hitschrich, Bernhard Mumm, Paul Spiesecke, and Matthias Günther. Deep-learning-based washout classification for decision support in contrast-enhanced ultrasound examinations of the liver.Journal of Medical Imaging, 12(4):044502, July 2025

    2025

  37. [37]

    CEUSegNet: A Cross-Modality Lesion Segmentation Network for Contrast-Enhanced Ultrasound

    Zheling Meng, Yangyang Zhu, Xiao Fan, Jie Tian, Fang Nie, and Kun Wang. CEUSegNet: A Cross-Modality Lesion Segmentation Network for Contrast-Enhanced Ultrasound. In2022 IEEE 19th International Symposium on Biomedical Imaging (ISBI), pages 1–5, March 2022

    2022

  38. [38]

    CEUS-SAM: Cross-Modal Prompt-Based SAM Network for Breast CEUS Image Segmentation

    Xutao Li, Min Xu, Ximiao Zhang, Sihua Niu, Jiaan Zhu, and Xiuzhuang Zhou. CEUS-SAM: Cross-Modal Prompt-Based SAM Network for Breast CEUS Image Segmentation. In2024 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), pages 994–999, December 2024

    2024

  39. [39]

    Yuxiang Duan, Jili Long, Shunyi Zhao, Hao Wang, and Yuriy S. Shmaliy. AI-Enhanced High-Precision Segmenta- tion and Perfusion Analysis in Myocardial Contrast Echocardiography.IEEE Transactions on Cybernetics, pages 1–13, 2026

    2026

  40. [40]

    Deep Learning Scaling is Predictable, Empirically

    Joel Hestness, Sharan Narang, Newsha Ardalani, Gregory Diamos, Heewoo Jun, Hassan Kianinejad, Md Mostofa Ali Patwary, Yang Yang, and Yanqi Zhou. Deep Learning Scaling is Predictable, Empirically, December 2017. arXiv:1712.00409 [cs]. 14 APREPRINT- JUNE8, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  41. [41]

    Rosenfeld, Amir Rosenfeld, Yonatan Belinkov, and Nir Shavit

    Jonathan S. Rosenfeld, Amir Rosenfeld, Yonatan Belinkov, and Nir Shavit. A Constructive Prediction of the Generalization Error Across Scales, December 2019. arXiv:1909.12673 [cs]

    work page arXiv 2019

  42. [42]

    Scaling Vision Transformers

    Xiaohua Zhai, Alexander Kolesnikov, Neil Houlsby, and Lucas Beyer. Scaling Vision Transformers. In2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 1204–1213, New Orleans, LA, USA, June 2022. IEEE

    2022

  43. [43]

    Emanuel, Rick I

    Anna L. Emanuel, Rick I. Meijer, Erik van Poelgeest, Pien Spoor, Erik H. Serné, and Etto C. Eringa. Contrast- enhanced ultrasound for quantification of tissue perfusion in humans.Microcirculation, 27(1):e12588, 2020. _eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1111/micc.12588

    work page doi:10.1111/micc.12588 2020

  44. [44]

    G. Hommel. A Stagewise Rejective Multiple Test Procedure Based on a Modified Bonferroni Test.Biometrika, 75(2):383–386, 1988. 15

    1988