arxiv: 2604.15096 · v1 · submitted 2026-04-16 · 💻 cs.CV · cs.LG

Recognition: unknown

Beyond Independent Frames: Latent Attention Masked Autoencoders for Multi-View Echocardiography

Simon B\"ohi , Irene Cannistraci , Sergio Mu\~noz Gonzalez , Moritz Vandenhirtz , Sonia Laguna , Samuel Ruiperez-Campillo , Max Kr\"ahenmann , Andrea Agostini

show 3 more authors

Ece Ozkan Thomas M. Sutter Julia E. Vogt

Authors on Pith no claims yet

Pith reviewed 2026-05-10 11:53 UTC · model grok-4.3

classification 💻 cs.CV cs.LG

keywords echocardiographymasked autoencodersmulti-view learningself-supervised learninglatent attentiontransfer learningcardiac imagingICD-10 prediction

0 comments

The pith

Adding a latent attention module allows masked autoencoders to integrate information across multiple echocardiographic views and frames in latent space.

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

The paper introduces Latent Attention Masked Autoencoder (LAMAE) to address the challenge of processing sparse and heterogeneous multi-view echocardiographic data. Standard masked autoencoders treat frames independently and miss the structural relationships between different views of the heart. By adding a latent attention module, the model can exchange information across frames and views, enabling reconstruction of a more complete representation of cardiac function. This approach is pretrained on a large uncurated dataset and shows effective transfer to pediatric data as well as the ability to predict ICD-10 codes from echo videos.

Core claim

LAMAE augments the standard MAE with a latent attention module that enables information exchange across frames and views directly in latent space. This allows the model to aggregate variable-length sequences and distinct views, reconstructing a holistic representation of cardiac function from partial observations. Pretraining on MIMIC-IV-ECHO yields the first reported results for ICD-10 code prediction from such videos and demonstrates that adult-learned representations transfer to pediatric cohorts despite anatomical differences.

What carries the argument

The latent attention module, which performs information exchange across multiple views and frames in the latent space of the masked autoencoder.

If this is right

Adult-trained representations transfer effectively to pediatric echocardiography despite substantial anatomical differences.
Multi-view attention leads to more robust representations compared to independent frame processing.
Self-supervised pretraining on uncurated echo videos supports downstream prediction of ICD-10 codes.
The approach aggregates variable-length multi-view sequences into coherent cardiac representations.

Where Pith is reading between the lines

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

Similar latent attention mechanisms could improve self-supervised models for other modalities with multiple views, such as CT or MRI scans of the heart.
Clinical systems might use these representations to assist diagnosis even when only partial views are available during an exam.
Further work could test if the model identifies specific cardiac abnormalities better than single-view baselines.

Load-bearing premise

The latent attention module aggregates variable-length sequences and distinct views in a way that captures genuine cardiac function rather than spurious correlations specific to the uncurated dataset.

What would settle it

An ablation study where removing the latent attention module results in no drop in performance on ICD-10 code prediction or adult-to-pediatric transfer tasks.

Figures

Figures reproduced from arXiv: 2604.15096 by Andrea Agostini, Ece Ozkan, Irene Cannistraci, Julia E. Vogt, Max Kr\"ahenmann, Moritz Vandenhirtz, Samuel Ruiperez-Campillo, Sergio Mu\~noz Gonzalez, Simon B\"ohi, Sonia Laguna, Thomas M. Sutter.

**Figure 1.** Figure 1: LAMAE architecture overview. During pretraining (left), masked frames from multiple views are encoded and fused through the Latent Attention (LA) module to learn shared representations, which are used to reconstruct full frames. During finetuning (right), frames are processed through that same encoder and LA module, followed by a lightweight classification head. LA𝜱 CLS … … CLS CLS CLS CLS CLS E𝜱 D𝝝 D𝝝 CLS… view at source ↗

**Figure 2.** Figure 2: Per-code AUROC results for the 10 topperforming ICD-10 codes under full finetuning. I48 I50 E78 Z79 N17 Z95 N18 I25 E87 E11 0.0 0.2 0.4 0.6 0.8 1.0 AUROC Full finetune Image-MAE VideoMAE LAMAE Video-LAMAE frame-based ones (i.e., 0.27–0.28 vs. 0.18–0.20), suggesting that temporal features are particularly robust for diagnosis when the feature extractor is fixed. Finally, the performance gap between fine… view at source ↗

**Figure 3.** Figure 3: AUROC and F1 scores for the top 10 ICD-10 codes. [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗

read the original abstract

Echocardiography is a widely used modality for cardiac assessment due to its non-invasive and cost-effective nature, but the sparse and heterogeneous spatiotemporal views of the heart pose distinct challenges. Existing masked autoencoder (MAE) approaches typically process images or short clips independently, failing to capture the inherent multi-view structure required for coherent cardiac representation. We introduce Latent Attention Masked Autoencoder (LAMAE), a foundation model architecture tailored to the multi-view nature of medical imaging. LAMAE augments the standard MAE with a latent attention module that enables information exchange across frames and views directly in latent space. This allows the model to aggregate variable-length sequences and distinct views, reconstructing a holistic representation of cardiac function from partial observations. We pretrain LAMAE on MIMIC-IV-ECHO, a large-scale, uncurated dataset reflecting real-world clinical variability. To the best of our knowledge, we present the first results for predicting ICD-10 codes from MIMIC-IV-ECHO videos. Furthermore, we empirically demonstrate that representations learned from adult data transfer effectively to pediatric cohorts despite substantial anatomical differences. These results provide evidence that incorporating structural priors, such as multi-view attention, yields significantly more robust and transferable representations.

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.

Desk Editor's Note private letter to a colleague

LAMAE adds a latent attention module to MAE so it can mix info across echo views in latent space, with pretraining on MIMIC-IV-ECHO and an adult-to-pediatric transfer test, but the abstract supplies no numbers or controls to show the module actually helps.

read the letter

The core idea is straightforward: standard masked autoencoders handle frames one at a time, so the authors insert a latent attention block that lets representations exchange information across different views and variable-length sequences before reconstruction. They pretrain on the uncurated MIMIC-IV-ECHO collection and report results on ICD-10 code prediction plus transfer to pediatric data despite anatomical differences. That combination of architecture and dataset is not in the prior work they cite, and the focus on multi-view structure matches a real constraint in echocardiography practice. Pretraining on messy hospital data rather than curated clips is also a reasonable choice for this modality. The transfer experiment at least tests whether the learned features survive domain shift, which is more than many self-supervised medical papers attempt. The main weakness is the complete absence of quantitative evidence. No baseline comparisons, no ablation of the attention module, no single-view controls, and no error bars appear in the text. Without those, it is impossible to know whether any gains come from the structural prior or from other training details. The risk that the model is picking up acquisition biases or demographic correlations in MIMIC-IV-ECHO rather than cardiac function remains open. The paper is aimed at researchers building self-supervised models for cardiac ultrasound. A reader who needs concrete starting points for handling multi-view video would get an architecture sketch and a dataset pointer, even if the performance claims need verification. It is worth sending to peer review because the problem is clinically relevant and the proposed fix is simple enough to test. Referees can ask for the missing ablations and baselines; the work does not look like a desk-reject on its face.

Referee Report

1 major / 0 minor

Summary. The paper introduces Latent Attention Masked Autoencoder (LAMAE), which augments a standard masked autoencoder with a latent attention module to enable information exchange across variable-length frames and distinct views directly in latent space. Pretrained on the uncurated MIMIC-IV-ECHO dataset, the model is claimed to produce the first results on ICD-10 code prediction from echocardiography videos and to support effective transfer from adult to pediatric cohorts despite anatomical differences, with the central assertion that incorporating structural priors such as multi-view attention yields significantly more robust and transferable representations than independent-frame processing.

Significance. If the empirical claims are substantiated with proper controls, this work could advance self-supervised learning for multi-view medical imaging by demonstrating the value of domain-specific structural priors over frame-independent MAE baselines. It would provide a foundation model for echocardiography that better handles real-world clinical variability and cross-population generalization, with potential downstream impact on automated cardiac assessment tasks.

major comments (1)

[Abstract] Abstract: The central claim that multi-view attention produces 'significantly more robust and transferable representations' is load-bearing for the contribution, yet the abstract supplies no quantitative results, ablation studies, single-view baselines, or error bars to isolate the effect of the latent attention module from potential dataset-specific correlations in the uncurated MIMIC-IV-ECHO collection (e.g., view selection biases or demographic artifacts). This directly engages the stress-test concern and prevents verification that gains reflect genuine cardiac function modeling rather than spurious cues.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their constructive feedback and for highlighting the need to strengthen the abstract. We address the major comment below and will revise the manuscript accordingly.

read point-by-point responses

Referee: [Abstract] Abstract: The central claim that multi-view attention produces 'significantly more robust and transferable representations' is load-bearing for the contribution, yet the abstract supplies no quantitative results, ablation studies, single-view baselines, or error bars to isolate the effect of the latent attention module from potential dataset-specific correlations in the uncurated MIMIC-IV-ECHO collection (e.g., view selection biases or demographic artifacts). This directly engages the stress-test concern and prevents verification that gains reflect genuine cardiac function modeling rather than spurious cues.

Authors: We agree that the abstract should include quantitative support for the central claims to allow readers to assess the contribution of the latent attention module. In the revised version, we will update the abstract to report key performance metrics, including ICD-10 code prediction accuracy on MIMIC-IV-ECHO, transfer learning results from adult to pediatric cohorts, and direct comparisons against independent-frame MAE baselines and single-view variants, with associated error bars where applicable. These additions will draw from the empirical results already presented in the main text and will help isolate the effect of multi-view latent attention from potential dataset artifacts. revision: yes

Circularity Check

0 steps flagged

No significant circularity: empirical architecture with no derivations

full rationale

The paper introduces LAMAE as an architectural extension of standard MAE with a latent attention module to handle multi-view echocardiography sequences. All central claims (robust representations, ICD-10 prediction on MIMIC-IV-ECHO, adult-to-pediatric transfer) are presented as empirical results from pretraining and evaluation rather than any closed-form derivation, first-principles prediction, or mathematical chain. No equations, fitted parameters renamed as predictions, self-definitional constructs, or load-bearing self-citations appear in the abstract or described content. The work is self-contained as an empirical ML contribution against external benchmarks and datasets.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The paper introduces one new architectural component (latent attention module) whose benefit is asserted empirically; no free parameters are explicitly fitted in the abstract, and no new physical entities are postulated.

axioms (1)

domain assumption Standard masked autoencoder reconstruction objective is a suitable self-supervised signal for cardiac video.
Implicit in the choice of MAE backbone.

invented entities (1)

Latent attention module no independent evidence
purpose: Enable information exchange across frames and views directly in latent space.
New component added to standard MAE; no independent evidence of its necessity is provided in the abstract.

pith-pipeline@v0.9.0 · 5566 in / 1302 out tokens · 23403 ms · 2026-05-10T11:53:04.677172+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

30 extracted references · 22 canonical work pages

[1]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION format.date year duplicate empty "emp...
[2]

@esa (Ref

\@ifxundefined[1] #1\@undefined \@firstoftwo \@secondoftwo \@ifnum[1] #1 \@firstoftwo \@secondoftwo \@ifx[1] #1 \@firstoftwo \@secondoftwo [2] @ #1 \@temptokena #2 #1 @ \@temptokena \@ifclassloaded agu2001 natbib The agu2001 class already includes natbib coding, so you should not add it explicitly Type <Return> for now, but then later remove the command n...
[3]

\@lbibitem[] @bibitem@first@sw\@secondoftwo \@lbibitem[#1]#2 \@extra@b@citeb \@ifundefined br@#2\@extra@b@citeb \@namedef br@#2 \@nameuse br@#2\@extra@b@citeb \@ifundefined b@#2\@extra@b@citeb @num @parse #2 @tmp #1 NAT@b@open@#2 NAT@b@shut@#2 \@ifnum @merge>\@ne @bibitem@first@sw \@firstoftwo \@ifundefined NAT@b*@#2 \@firstoftwo @num @NAT@ctr \@secondoft...
[4]

@open @close @open @close and [1] URL: #1 \@ifundefined chapter * \@mkboth \@ifxundefined @sectionbib * \@mkboth * \@mkboth\@gobbletwo \@ifclassloaded amsart * \@ifclassloaded amsbook * \@ifxundefined @heading @heading NAT@ctr thebibliography [1] @ \@biblabel @NAT@ctr \@bibsetup #1 @NAT@ctr @ @openbib .11em \@plus.33em \@minus.07em 4000 4000 `\.\@m @bibit...
[5]

Echocardiographic assessment of cardiac structure and function in chronic renal disease

Kaoru Dohi. Echocardiographic assessment of cardiac structure and function in chronic renal disease. Journal of Echocardiography, 17 0 (3): 0 115--122, September 2019. ISSN 1880-344X. doi:10.1007/s12574-019-00436-x. URL https://doi.org/10.1007/s12574-019-00436-x

work page doi:10.1007/s12574-019-00436-x 2019
[6]

MIMIC - IV - ECHO : Echocardiogram Matched Subset , 2023

Brian Gow, Tom Pollard, Nathaniel Greenbaum, Benjamin Moody, Alistair Johnson, Elizabeth Herbst, Jonathan W Waks, Parastou Eslami, Ashish Chaudhari, Tanner Carbonati, Seth Berkowitz, Roger Mark, and Steven Horng. MIMIC - IV - ECHO : Echocardiogram Matched Subset , 2023. URL https://physionet.org/content/mimic-iv-echo/0.1/

2023
[7]

org/abs/2111.06377

Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, and Ross Girshick. Masked Autoencoders Are Scalable Vision Learners , December 2021. URL http://arxiv.org/abs/2111.06377. arXiv:2111.06377 [cs]

work page arXiv 2021
[8]

Kukučka et al

Maximilian Ilse, Jakub M. Tomczak, and Max Welling. Attention-based Deep Multiple Instance Learning , June 2018. URL http://arxiv.org/abs/1802.04712. arXiv:1802.04712 [cs]

work page arXiv 2018
[9]

USFM : A universal ultrasound foundation model generalized to tasks and organs towards label efficient image analysis

Jing Jiao, Jin Zhou, Xiaokang Li, Menghua Xia, Yi Huang, Lihong Huang, Na Wang, Xiaofan Zhang, Shichong Zhou, Yuanyuan Wang, and Yi Guo. USFM : A universal ultrasound foundation model generalized to tasks and organs towards label efficient image analysis. Medical Image Analysis, 96: 0 103202, August 2024. ISSN 1361-8415. doi:10.1016/j.media.2024.103202. U...

work page doi:10.1016/j.media.2024.103202 2024
[10]

MIMIC - IV , 2024

Alistair Johnson, Lucas Bulgarelli, Tom Pollard, Steven Horng, Leo Anthony Celi, and Roger Mark. MIMIC - IV , 2024. URL https://physionet.org/content/mimiciv/1.0/

2024
[11]

Deblurring Masked Autoencoder is Better Recipe for Ultrasound Image Recognition , July 2023

Qingbo Kang, Jun Gao, Kang Li, and Qicheng Lao. Deblurring Masked Autoencoder is Better Recipe for Ultrasound Image Recognition , July 2023. URL http://arxiv.org/abs/2306.08249. arXiv:2306.08249 [cs]

work page arXiv 2023
[12]

D\ \ 2\ \ MAE : Diffusional Deblurring MAE for Ultrasound Image Pre -training

Qingbo Kang, Jun Gao, Hongkai Zhao, Zhu He, Kang Li, and Qicheng Lao. D\ \ 2\ \ MAE : Diffusional Deblurring MAE for Ultrasound Image Pre -training. In James C. Gee, Daniel C. Alexander, Jaesung Hong, Juan Eugenio Iglesias, Carole H. Sudre, Archana Venkataraman, Polina Golland, Jong Hyo Kim, and Jinah Park (eds.), Medical Image Computing and Computer Assi...

work page doi:10.1007/978-3-032-05169-1_11 2025
[13]

EchoFM : Foundation Model for Generalizable Echocardiogram Analysis , January 2025

Sekeun Kim, Pengfei Jin, Sifan Song, Cheng Chen, Yiwei Li, Hui Ren, Xiang Li, Tianming Liu, and Quanzheng Li. EchoFM : Foundation Model for Generalizable Echocardiogram Analysis , January 2025. URL http://arxiv.org/abs/2410.23413. arXiv:2410.23413 [cs]

work page arXiv 2025
[14]

Kosiorek, Seungjin Choi, and Yee Whye Teh

Juho Lee, Yoonho Lee, Jungtaek Kim, Adam R. Kosiorek, Seungjin Choi, and Yee Whye Teh. Set Transformer : A Framework for Attention -based Permutation - Invariant Neural Networks , May 2019. URL http://arxiv.org/abs/1810.00825. arXiv:1810.00825 [cs]

work page arXiv 2019
[15]

Walker, Steven Hawken, and Adrian D

Youssef Megahed, Robin Ducharme, Aylin Erman, Mark C. Walker, Steven Hawken, and Adrian D. C. Chan. USF - MAE : Ultrasound Self - Supervised Foundation Model with Masked Autoencoding , January 2026. URL https://papers.ssrn.com/abstract=5900025

2026
[16]

Vogt, and Sven Wellmann

Holger Michel, Ece Ozkan, Kieran Chin-Cheong, Anna Badura, Verena Lehnerer, Stephan Gerling, Julia E. Vogt, and Sven Wellmann. Automated detection of neonatal pulmonary hypertension in echocardiograms with a deep learning model. Pediatric Research, pp.\ 1--8, September 2025. ISSN 1530-0447. doi:10.1038/s41390-025-04404-3. URL https://www.nature.com/articl...

work page doi:10.1038/s41390-025-04404-3 2025
[17]

Masoud Mokhtari, Neda Ahmadi, Teresa S. M. Tsang, Purang Abolmaesumi, and Renjie Liao. GEMTrans : A General , Echocardiography -based, Multi - Level Transformer Framework for Cardiovascular Diagnosis , August 2023. URL http://arxiv.org/abs/2308.13217. arXiv:2308.13217 [cs]

work page arXiv 2023
[18]

Badano, James N

Victor Mor-Avi, Alexandra Blitz, Marcus Schreckenberg, Karima Addetia, Kalie Kebed, Gregory Scalia, Luigi P. Badano, James N. Kirkpatrick, Pedro Gutierrez-Fajardo, Ana Clara Tude Rodrigues, Anita Sadeghpour, Edwin S. Tucay, Aldo D. Prado, Wendy Tsang, Kofo O. Ogunyankin, Alexander Rossmanith, Georg Schummers, Dorottya Laczik, Federico M. Asch, and Roberto...

work page doi:10.1007/s10554-023-02960-5 2023
[19]

Enhancing cardiac function assessment: Developing and validating a domain adaptive framework for automating the segmentation of echocardiogram videos

Mojdeh Nazari, Hassan Emami, Reza Rabiei, Hamid Reza Rabiee, Arsalan Salari, and Hossein Sadr. Enhancing cardiac function assessment: Developing and validating a domain adaptive framework for automating the segmentation of echocardiogram videos. Computerized Medical Imaging and Graphics, 124: 0 102627, September 2025. ISSN 0895-6111. doi:10.1016/j.compmed...

work page doi:10.1016/j.compmedimag.2025.102627 2025
[20]

Langlotz, Paul A

David Ouyang, Bryan He, Amirata Ghorbani, Neal Yuan, Joseph Ebinger, Curtis P. Langlotz, Paul A. Heidenreich, Robert A. Harrington, David H. Liang, Euan A. Ashley, and James Y. Zou. Video-based AI for beat-to-beat assessment of cardiac function. Nature, 580 0 (7802): 0 252--256, April 2020. ISSN 0028-0836, 1476-4687. doi:10.1038/s41586-020-2145-8. URL htt...

work page doi:10.1038/s41586-020-2145-8 2020
[21]

Sutter, Yurong Hu, Sebastian Balzer, and Julia E

Ece Ozkan, Thomas M. Sutter, Yurong Hu, Sebastian Balzer, and Julia E. Vogt. M(otion)- Mode Based Prediction of Ejection Fraction Using Echocardiograms . In Ullrich Köthe and Carsten Rother (eds.), Pattern Recognition , volume 14264, pp.\ 307--320. Springer Nature Switzerland, Cham, 2024. ISBN 978-3-031-54604-4 978-3-031-54605-1. doi:10.1007/978-3-031-546...

work page doi:10.1007/978-3-031-54605-1_20 2024
[22]

Reddy, Leo Lopez, David Ouyang, James Y

Charitha D. Reddy, Leo Lopez, David Ouyang, James Y. Zou, and Bryan He. Video- Based Deep Learning for Automated Assessment of Left Ventricular Ejection Fraction in Pediatric Patients . Journal of the American Society of Echocardiography, 36 0 (5): 0 482--489, May 2023. ISSN 08947317. doi:10.1016/j.echo.2023.01.015. URL https://linkinghub.elsevier.com/ret...

work page doi:10.1016/j.echo.2023.01.015 2023
[23]

Sutter, Ece Ozkan, and Julia E

Yves Stebler, Thomas M. Sutter, Ece Ozkan, and Julia E. Vogt. Temporal Representation Learning for Real - Time Ultrasound Analysis , September 2025. URL http://arxiv.org/abs/2509.01433. arXiv:2509.01433 [eess]

work page arXiv 2025
[24]

Multi- View Echocardiographic Embedding for Accessible AI Development

Takeshi Tohyama, Ahram Han, Dukyong Yoon, Kenneth Paik, Brian Gow, Nura Izath, Jacques Kpodonu, and Leo Anthony Celi. Multi- View Echocardiographic Embedding for Accessible AI Development . medRxiv, pp.\ 2025.08.15.25333725, October 2025. doi:10.1101/2025.08.15.25333725. URL https://pmc.ncbi.nlm.nih.gov/articles/PMC12393585/

work page doi:10.1101/2025.08.15.25333725 2025
[25]

Wang, Q., Zhang, Z., Xie, B., Jin, X., Wang, Y ., Wang, S., Zheng, L., Yang, X., and Zeng, W

Zhan Tong, Yibing Song, Jue Wang, and Limin Wang. VideoMAE : Masked Autoencoders are Data - Efficient Learners for Self - Supervised Video Pre - Training , October 2022. URL http://arxiv.org/abs/2203.12602. arXiv:2203.12602 [cs]

work page arXiv 2022
[26]

simple-icd-10: A simple python library for ICD -10 codes, 2025

Stefano Travasci. simple-icd-10: A simple python library for ICD -10 codes, 2025. URL https://simpleicd10.stefanotravasci.it/

2025
[27]

Comprehensive echocardiogram evaluation with view primed vision language AI

Milos Vukadinovic, I.-Min Chiu, Xiu Tang, Neal Yuan, Tien-Yu Chen, Paul Cheng, Debiao Li, Susan Cheng, Bryan He, and David Ouyang. Comprehensive echocardiogram evaluation with view primed vision language AI . Nature, pp.\ 1--8, November 2025. ISSN 1476-4687. doi:10.1038/s41586-025-09850-x. URL https://www.nature.com/articles/s41586-025-09850-x

work page doi:10.1038/s41586-025-09850-x 2025
[28]

Assran, Q

Limin Wang, Bingkun Huang, Zhiyu Zhao, Zhan Tong, Yinan He, Yi Wang, Yali Wang, and Yu Qiao. VideoMAE V2 : Scaling Video Masked Autoencoders with Dual Masking . In 2023 IEEE / CVF Conference on Computer Vision and Pattern Recognition ( CVPR ) , pp.\ 14549--14560, Vancouver, BC, Canada, June 2023. IEEE. ISBN 979-8-3503-0129-8. doi:10.1109/CVPR52729.2023.01...

work page doi:10.1109/cvpr52729.2023.01398 2023
[29]

EchoCardMAE : Video Masked Auto - Encoders Customized for Echocardiography

Xuan Yang, Rui Xu, Xinchen Ye, Zhihui Wang, Miao Zhang, Yi Wang, Xin Fan, Hongkai Wang, Qingxiong Yue, Xiangjian He, and Yen-Wei Chen. EchoCardMAE : Video Masked Auto - Encoders Customized for Echocardiography . In James C. Gee, Daniel C. Alexander, Jaesung Hong, Juan Eugenio Iglesias, Carole H. Sudre, Archana Venkataraman, Polina Golland, Jong Hyo Kim, a...

work page doi:10.1007/978-3-032-05169-1_17 2025
[30]

Echo- Vision - FM : A Pre -training and Fine -tuning Framework for Echocardiogram Videos Vision Foundation Model , October 2024

Ziyang Zhang, Qinxin Wu, Sirui Ding, Xiaolong Wang, and Jiancheng Ye. Echo- Vision - FM : A Pre -training and Fine -tuning Framework for Echocardiogram Videos Vision Foundation Model , October 2024. URL https://www.medrxiv.org/content/10.1101/2024.10.09.24315195v2. Pages: 2024.10.09.24315195

work page doi:10.1101/2024.10.09.24315195v2 2024