RePCM: Region-Specific and Phenotype-Adaptive Bi-Ventricular Cardiac Motion Synthesis

Hao Li; Lei Li; Lingyu Chen; Xiaohan Yuan; Xuan Yang; Yanan Liu

arxiv: 2605.21237 · v1 · pith:I5GA4OM6new · submitted 2026-05-20 · 💻 cs.CV · cs.AI

RePCM: Region-Specific and Phenotype-Adaptive Bi-Ventricular Cardiac Motion Synthesis

Xuan Yang , Xiaohan Yuan , Hao Li , Lingyu Chen , Yanan Liu , Lei Li This is my paper

Pith reviewed 2026-05-21 05:34 UTC · model grok-4.3

classification 💻 cs.CV cs.AI

keywords cardiac motion synthesisbi-ventricular meshregion-specific modelingphenotype adaptationconditional VAEmotion completionfunctional partitioningdisease variability

0 comments

The pith

RePCM generates full-cycle bi-ventricular motion from one end-diastolic mesh by deriving functional regions from motion data and adapting the synthesis to disease phenotype.

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

Cardiac motion over a full cycle is needed to measure regional heart function, yet dense time sequences are rarely available in practice. The work starts from the more accessible end-diastolic frame and seeks to complete the missing motion while respecting strong regional and disease-specific differences. Global generative models tend to average these differences away. RePCM first trains a reconstruction network to obtain vertex-wise motion descriptors, clusters them into a data-driven partition of functional regions, and then injects those regions into a conditional variational autoencoder so that motion is exchanged only within matching regions. A phenotype-adaptive mixture-of-experts prior, conditioned on the input shape, further tunes the latent motion distribution to the patient's disease category. Experiments across three datasets that span different cardiovascular conditions report gains in both geometric fidelity and preservation of localized dynamics.

Core claim

RePCM shows that a data-driven functional partition obtained by clustering vertex-wise motion descriptors, when enforced through masked synchronized region exchange inside a conditional VAE and combined with a phenotype-adaptive mixture-of-experts prior conditioned on end-diastolic shape, produces motion sequences that maintain region-specific dynamics and capture inter-disease variability more faithfully than models optimized only for global patterns.

What carries the argument

The Region-Specific Injection Module that performs masked, synchronized region exchange inside a conditional VAE, together with the Phenotype-Adaptive Mixture-of-Experts prior conditioned on end-diastolic shape.

If this is right

Geometric and functional metrics improve consistently across datasets covering multiple cardiovascular diseases.
Region-specific motion patterns are retained without cross-region mixing that occurs in global models.
Inter-disease variability is captured through anatomy-guided cues supplied by the end-diastolic shape alone.
Single-frame end-diastolic meshes become sufficient input for producing temporally dense sequences usable for regional function analysis.

Where Pith is reading between the lines

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

The same region-partitioning idea could be tested on motion synthesis for other deformable organs where pathology produces localized rather than uniform changes.
The phenotype-adaptive prior opens a route to patient-specific motion templates that might support pre-procedural planning once integrated with clinical imaging pipelines.
Direct comparison of the learned clusters against standard 17-segment cardiac models would clarify whether the data-driven regions correspond to established anatomical territories.

Load-bearing premise

The data-driven clustering of vertex-wise motion descriptors produces a functional partition that meaningfully captures localized dynamics and can be enforced via masked region exchange without introducing artifacts or losing global coherence.

What would settle it

If motion sequences synthesized on a held-out dataset containing known regional wall-motion abnormalities fail to reproduce those localized defects while still matching global volume curves, the central claim would be falsified.

Figures

Figures reproduced from arXiv: 2605.21237 by Hao Li, Lei Li, Lingyu Chen, Xiaohan Yuan, Xuan Yang, Yanan Liu.

**Figure 1.** Figure 1: Overview of the proposed RePCM framework. (a) Stage I: DataDriven Functional Partitioning. (b) Region-Specific Injection Module. (c) Stage II: Cardiac Motion Sequence Completion. mesh X0 ∈ R N×3 , our goal is to synthesize a full cardiac cycle Xˆ 0:T −1 ∈ R T ×N×3 by predicting ED-relative trajectories Xˆ T ∈ R N×(T·3): Xˆ t = X0 + Xˆ t T , t = 0, . . . , T − 1. (1) As illustrated in [PITH_FULL_IMAGE:fig… view at source ↗

**Figure 2.** Figure 2: Illustration of biventricular motion completion results. HCM: hypertrophic cardiomyopathy. DCM: dilated cardiomyopathy. NOR: normal. findings, [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗

**Figure 3.** Figure 3: Comparsion Results: (a) LV volume curves. (b) RV volume curves. Ablation Results: (c) Normalized LVV/EDV curves by disease. (d) Expertdisease usage matrix showing expert specialization [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

read the original abstract

Cardiac motion over a cardiac cycle is crucial for quantifying regional function and is strongly affected by cardiovascular diseases. Since temporally dense mesh sequences are difficult to obtain in practice, we focus on leveraging the more accessible end-diastolic frame to infer a full-cycle sequence. Due to strong regional and disease-specific differences, traditional methods often oversmooth the data by relying on generative models that are optimized for global patterns. To address this problem, we propose Region-Aware and Phenotype-Adaptive Bi-Ventricular Cardiac Motion Synthesis (RePCM) for single frame Bi-ventricular mesh motion completion. In Stage I, a reconstruction network learns vertex wise motion descriptors and clustering yields a data driven functional partition, providing an explicit motion derived region structure. In Stage II, a Region-Specific Injection Module enforces masked, synchronized region exchange within a conditional VAE, preserving localized specific dynamics and restricting cross-region mixing. A Phenotype-Adaptive Mixture-of-Experts prior conditioned on ED shape uses anatomy-guided cues to model latent motion trends and capture inter-disease variability. Experiments on three datasets covering different cardiovascular diseases show consistent gains in geometric and functional metrics and improved preservation of region specific dynamics.

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

RePCM adds data-driven region clustering plus masked injection and a phenotype MoE prior inside a cVAE to reduce oversmoothing in single-frame cardiac motion synthesis.

read the letter

This paper's core move is a two-stage setup: first learn vertex motion descriptors from limited data and cluster them into functional regions, then feed those regions into a conditional VAE that uses masked synchronized exchange to keep local dynamics separate and a mixture-of-experts prior conditioned on end-diastolic shape to handle disease variability. The abstract claims this produces better geometric and functional metrics on three cardiovascular disease datasets while preserving region-specific patterns better than global baselines. That is a targeted, practical step for applications that need regional heart function measures rather than just overall shape change. The assembly of explicit partitioning, masked injection, and anatomy-guided MoE is not a direct copy of the cited prior cardiac motion work, so the technical combination counts as new. The framing around oversmoothing in global models and the use of phenotype cues from the single available frame are reasonable responses to real constraints in cardiac mesh data. The main soft spot is the mechanical coupling across region boundaries. Cardiac tissue is continuous, so treating clusters as independent during latent sampling and decoding risks velocity or strain discontinuities at the edges even if per-region or global scores look improved. The abstract gives no boundary-specific coherence checks or strain continuity numbers, leaving that link between the architecture and the claimed regional fidelity unverified. The clustering step itself also rests on the descriptors producing partitions that actually reflect localized function, and that could vary with the input data. This is for people working on generative models for cardiac meshes or similar medical motion tasks where regional fidelity from limited frames matters. A reader who wants concrete architecture details for incorporating data-driven priors would get value from the method description. I would send it for peer review. The experiments span multiple datasets and the approach is a clear extension of standard VAE components, so referees can check the boundary behavior and the size of the reported gains.

Referee Report

2 major / 2 minor

Summary. The manuscript proposes RePCM, a two-stage framework for synthesizing full-cycle bi-ventricular cardiac motion from a single end-diastolic mesh. Stage I trains a reconstruction network to obtain vertex-wise motion descriptors and applies clustering to derive a data-driven functional partition of the mesh. Stage II employs a conditional VAE equipped with a Region-Specific Injection Module that performs masked, synchronized region exchange to enforce localized dynamics while limiting cross-region mixing, together with a Phenotype-Adaptive Mixture-of-Experts prior conditioned on ED shape to capture disease-specific motion variability. Experiments across three cardiovascular-disease datasets report gains in geometric and functional metrics together with improved region-specific fidelity relative to global generative baselines.

Significance. If the central claims hold after addressing mechanical consistency, the work would offer a practical route to generating temporally dense, phenotype-aware cardiac motion sequences that respect regional functional differences, which is relevant for regional function quantification in cardiovascular imaging. The data-driven partitioning and phenotype-conditioned prior are conceptually attractive strengths; however, the significance is tempered by the absence of explicit validation that the synthesized motions remain physiologically plausible across region boundaries.

major comments (2)

[Stage II] Stage II, Region-Specific Injection Module: the masked synchronized region exchange is presented as the mechanism that preserves localized dynamics without cross-region mixing. Cardiac tissue is a mechanically continuous medium in which deformation and force propagate across any partition; the manuscript provides no boundary-specific checks (e.g., inter-region strain continuity, velocity-gradient magnitude, or edge-discontinuity metrics) to confirm that the masking operation does not introduce non-physiological jumps while aggregate geometric/functional scores remain acceptable.
[Experiments] Experiments section: the abstract states that the method yields 'consistent gains in geometric and functional metrics and improved preservation of region specific dynamics' on three datasets, yet the provided text contains no quantitative tables, ablation results isolating the contribution of the Region-Specific Injection Module versus the MoE prior, or error analysis stratified by region or disease phenotype. Without these, it is not possible to determine whether the reported improvements are load-bearing for the central claim or sensitive to post-hoc choices.

minor comments (2)

The abstract refers to 'three datasets covering different cardiovascular diseases' without naming the datasets, acquisition protocols, or disease labels; explicit identification would improve reproducibility.
[Stage I] Stage I: the description of how vertex-wise motion descriptors are clustered (algorithm, distance metric, and criterion for selecting the number of regions) is brief; a short paragraph or pseudocode would clarify the functional partition construction.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback on our manuscript. We address each major comment below, indicating the specific revisions we will make to strengthen the validation of physiological plausibility and experimental rigor.

read point-by-point responses

Referee: [Stage II] Stage II, Region-Specific Injection Module: the masked synchronized region exchange is presented as the mechanism that preserves localized dynamics without cross-region mixing. Cardiac tissue is a mechanically continuous medium in which deformation and force propagate across any partition; the manuscript provides no boundary-specific checks (e.g., inter-region strain continuity, velocity-gradient magnitude, or edge-discontinuity metrics) to confirm that the masking operation does not introduce non-physiological jumps while aggregate geometric/functional scores remain acceptable.

Authors: We agree that explicit checks for mechanical continuity across region boundaries are important for confirming physiological plausibility, given that cardiac tissue is a continuous medium. Although the masked synchronized region exchange is intended to limit cross-region mixing while preserving localized dynamics, we acknowledge that aggregate metrics alone do not fully address potential boundary artifacts. In the revised manuscript, we will incorporate boundary-specific quantitative analyses, including inter-region strain continuity, velocity-gradient magnitudes at partition edges, and edge-discontinuity metrics, computed on both synthesized and ground-truth sequences. We will also add qualitative visualizations of deformation fields across boundaries to demonstrate consistency. revision: yes
Referee: [Experiments] Experiments section: the abstract states that the method yields 'consistent gains in geometric and functional metrics and improved preservation of region specific dynamics' on three datasets, yet the provided text contains no quantitative tables, ablation results isolating the contribution of the Region-Specific Injection Module versus the MoE prior, or error analysis stratified by region or disease phenotype. Without these, it is not possible to determine whether the reported improvements are load-bearing for the central claim or sensitive to post-hoc choices.

Authors: We appreciate this observation and agree that detailed quantitative support is essential to substantiate the central claims. The current version summarizes results across three datasets but lacks the requested breakdowns. In the revised manuscript, we will include full quantitative tables with geometric and functional metrics, ablation studies that isolate the individual contributions of the Region-Specific Injection Module and the Phenotype-Adaptive Mixture-of-Experts prior, and error analyses stratified by cardiac region and disease phenotype. These additions will clarify the load-bearing nature of the improvements and their robustness. revision: yes

Circularity Check

0 steps flagged

No significant circularity; pipeline uses standard components without self-referential reductions

full rationale

The described method consists of a Stage I reconstruction network that learns vertex-wise motion descriptors followed by clustering to obtain a data-driven functional partition, then a Stage II conditional VAE incorporating a Region-Specific Injection Module for masked region exchange and a Phenotype-Adaptive Mixture-of-Experts prior. No equations, derivations, or self-citations are presented that define outputs in terms of the same fitted quantities or reduce claimed improvements to inputs by construction. Experimental gains on three external datasets are reported as evidence rather than tautological results, and the approach relies on conventional VAE and clustering elements without load-bearing self-referential steps.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review yields no explicit free parameters, axioms, or invented entities; the method implicitly assumes standard VAE training and clustering produce usable region structures.

pith-pipeline@v0.9.0 · 5744 in / 1230 out tokens · 26843 ms · 2026-05-21T05:34:55.615830+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

21 extracted references · 21 canonical work pages · 1 internal anchor

[1]

Medical image analysis 26(1), 133–145 (2015)

Bai,W.,Shi,W.,deMarvao,A.,Dawes,T.J.,O’Regan,D.P.,Cook,S.A.,Rueckert, D.: A bi-ventricular cardiac atlas built from 1000+ high resolution mr images of healthy subjects and an analysis of shape and motion. Medical image analysis 26(1), 133–145 (2015)

work page 2015
[2]

Bernard, O., Lalande, A., Zotti, C., Cervenansky, F., Yang, X., Heng, P.A., Cetin, I., Lekadir, K., Camara, O., Ballester, M.A.G., et al.: Deep learning techniques for automatic mri cardiac multi-structures segmentation and diagnosis: is the problem solved? IEEE transactions on medical imaging37(11), 2514–2525 (2018)

work page 2018
[3]

IEEE Transactions on Medical Imaging40(12), 3543–3554 (2021)

Campello, V.M., Gkontra, P., Izquierdo, C., Martin-Isla, C., Sojoudi, A., Full, P.M., Maier-Hein, K., Zhang, Y., He, Z., Ma, J., et al.: Multi-centre, multi-vendor and multi-disease cardiac segmentation: the m&ms challenge. IEEE Transactions on Medical Imaging40(12), 3543–3554 (2021)

work page 2021
[4]

Progress in biophysics and molecular biology122(3), 215–226 (2016)

Carruth, E.D., McCulloch, A.D., Omens, J.H.: Transmural gradients of myocar- dial structure and mechanics: implications for fiber stress and strain in pressure overload. Progress in biophysics and molecular biology122(3), 215–226 (2016)

work page 2016
[5]

Korean journal of radiology20(1), 114–125 (2019)

Cui, C., Yin, G., Lu, M., Chen, X., Cheng, S., Li, L., Yan, W., Song, Y., Prasad, S., Zhang, Y., et al.: Retrospective electrocardiography-gated real-time cardiac cine mri at 3t: comparison with conventional segmented cine mri. Korean journal of radiology20(1), 114–125 (2019)

work page 2019
[6]

Medical image analysis75, 102278 (2022)

Di Folco, M., Moceri, P., Clarysse, P., Duchateau, N.: Characterizing interactions between cardiac shape and deformation by non-linear manifold learning. Medical image analysis75, 102278 (2022)

work page 2022
[7]

In: International Conference on Medical Image Computing and Computer-Assisted Intervention

Dou, H., Huang, J., Zakeri, A., Zhou, Z., Mu, T., Duan, J., Frangi, A.F.: 4d car- diosynth: Synthesising dynamic virtual heart populations through spatiotemporal disentanglement. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. pp. 3–12. Springer (2025)

work page 2025
[8]

In: Proceedings of the 28th ACM international conference on multimedia

Guo, C., Zuo, X., Wang, S., Zou, S., Sun, Q., Deng, A., Gong, M., Cheng, L.: Action2motion: Conditioned generation of 3d human motions. In: Proceedings of the 28th ACM international conference on multimedia. pp. 2021–2029 (2020) 10 X. Yang et al

work page 2021
[9]

Auto-Encoding Variational Bayes

Kingma, D.P., Welling, M.: Auto-encoding variational bayes. arXiv preprint arXiv:1312.6114 (2013)

work page internal anchor Pith review Pith/arXiv arXiv 2013
[10]

Medical image analysis97, 103293 (2024)

Kong, F., Stocker, S., Choi, P.S., Ma, M., Ennis, D.B., Marsden, A.L.: Sdf4chd: Generative modeling of cardiac anatomies with congenital heart defects. Medical image analysis97, 103293 (2024)

work page 2024
[11]

IEEE Journal of Biomedical and Health Informatics27(7), 3302–3313 (2023)

Martín-Isla, C., Campello, V.M., Izquierdo, C., Kushibar, K., Sendra-Balcells, C., Gkontra, P., Sojoudi, A., Fulton, M.J., Arega, T.W., Punithakumar, K., et al.: Deep learning segmentation of the right ventricle in cardiac mri: the m&ms chal- lenge. IEEE Journal of Biomedical and Health Informatics27(7), 3302–3313 (2023)

work page 2023
[12]

Circulation: Cardiovascular Imaging9(2), e003866 (2016)

Muraru, D., Onciul, S., Peluso, D., Soriani, N., Cucchini, U., Aruta, P., Romeo, G., Cavalli, G., Iliceto, S., Badano, L.P.: Sex-and method-specific reference values for right ventricular strain by 2-dimensional speckle-tracking echocardiography. Circulation: Cardiovascular Imaging9(2), e003866 (2016)

work page 2016
[13]

In: Medical Imaging 2007: Physiology, Function, and Structure from Medical Images

Ordas, S., Oubel, E., Leta, R., Carreras, F., Frangi, A.F.: A statistical shape model of the heart and its application to model-based segmentation. In: Medical Imaging 2007: Physiology, Function, and Structure from Medical Images. vol. 6511, pp. 490–500. SPIE (2007)

work page 2007
[14]

In: Proceedings of the AAAI conference on artificial intelligence

Perez, E., Strub, F., De Vries, H., Dumoulin, V., Courville, A.: Film: Visual rea- soning with a general conditioning layer. In: Proceedings of the AAAI conference on artificial intelligence. vol. 32 (2018)

work page 2018
[15]

In: Proceedings of the IEEE/CVF international con- ference on computer vision

Petrovich, M., Black, M.J., Varol, G.: Action-conditioned 3d human motion syn- thesis with transformer vae. In: Proceedings of the IEEE/CVF international con- ference on computer vision. pp. 10985–10995 (2021)

work page 2021
[16]

Nature Machine Intelligence7(5), 800–811 (2025)

Qiao, M., McGurk, K.A., Wang, S., Matthews, P.M., O’Regan, D.P., Bai, W.: A personalized time-resolved 3d mesh generative model for unveiling normal heart dynamics. Nature Machine Intelligence7(5), 800–811 (2025)

work page 2025
[17]

IEEE transactions on medical imaging43(3), 1259–1269 (2023)

Qiao, M., Wang, S., Qiu, H., De Marvao, A., O’Regan, D.P., Rueckert, D., Bai, W.: Cheart: A conditional spatio-temporal generative model for cardiac anatomy. IEEE transactions on medical imaging43(3), 1259–1269 (2023)

work page 2023
[18]

jacc cardiovasc imaging (2024)

Smiseth, O., Rider, O., Cvijic, M., Valkovic, L., Remme, E., Voigt, J.: Myocardial strain imaging: theory, current practice, and the future. jacc cardiovasc imaging (2024)

work page 2024
[19]

JACC: Car- diovascular Imaging18(3), 340–381 (2025)

Smiseth, O.A., Rider, O., Cvijic, M., Valkovič, L., Remme, E.W., Voigt, J.U.: Myocardial strain imaging: theory, current practice, and the future. JACC: Car- diovascular Imaging18(3), 340–381 (2025)

work page 2025
[20]

In: International Conference on Medical Image Computing and Computer-Assisted Intervention

Sørensen, K., Diez, P., Margeta, J., El Youssef, Y., Pham, M., Pedersen, J.J., Kühl, T., De Backer, O., Kofoed, K., Camara, O., et al.: Spatio-temporal neural distance fields for conditional generative modeling of the heart. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. pp. 422–432. Springer (2024)

work page 2024
[21]

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

Wu, Z., Yu, C., Wang, F., Bai, X.: Animateanymesh: A feed-forward 4d foundation model for text-driven universal mesh animation. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 13557–13568 (2025)

work page 2025

[1] [1]

Medical image analysis 26(1), 133–145 (2015)

Bai,W.,Shi,W.,deMarvao,A.,Dawes,T.J.,O’Regan,D.P.,Cook,S.A.,Rueckert, D.: A bi-ventricular cardiac atlas built from 1000+ high resolution mr images of healthy subjects and an analysis of shape and motion. Medical image analysis 26(1), 133–145 (2015)

work page 2015

[2] [2]

Bernard, O., Lalande, A., Zotti, C., Cervenansky, F., Yang, X., Heng, P.A., Cetin, I., Lekadir, K., Camara, O., Ballester, M.A.G., et al.: Deep learning techniques for automatic mri cardiac multi-structures segmentation and diagnosis: is the problem solved? IEEE transactions on medical imaging37(11), 2514–2525 (2018)

work page 2018

[3] [3]

IEEE Transactions on Medical Imaging40(12), 3543–3554 (2021)

Campello, V.M., Gkontra, P., Izquierdo, C., Martin-Isla, C., Sojoudi, A., Full, P.M., Maier-Hein, K., Zhang, Y., He, Z., Ma, J., et al.: Multi-centre, multi-vendor and multi-disease cardiac segmentation: the m&ms challenge. IEEE Transactions on Medical Imaging40(12), 3543–3554 (2021)

work page 2021

[4] [4]

Progress in biophysics and molecular biology122(3), 215–226 (2016)

Carruth, E.D., McCulloch, A.D., Omens, J.H.: Transmural gradients of myocar- dial structure and mechanics: implications for fiber stress and strain in pressure overload. Progress in biophysics and molecular biology122(3), 215–226 (2016)

work page 2016

[5] [5]

Korean journal of radiology20(1), 114–125 (2019)

Cui, C., Yin, G., Lu, M., Chen, X., Cheng, S., Li, L., Yan, W., Song, Y., Prasad, S., Zhang, Y., et al.: Retrospective electrocardiography-gated real-time cardiac cine mri at 3t: comparison with conventional segmented cine mri. Korean journal of radiology20(1), 114–125 (2019)

work page 2019

[6] [6]

Medical image analysis75, 102278 (2022)

Di Folco, M., Moceri, P., Clarysse, P., Duchateau, N.: Characterizing interactions between cardiac shape and deformation by non-linear manifold learning. Medical image analysis75, 102278 (2022)

work page 2022

[7] [7]

In: International Conference on Medical Image Computing and Computer-Assisted Intervention

Dou, H., Huang, J., Zakeri, A., Zhou, Z., Mu, T., Duan, J., Frangi, A.F.: 4d car- diosynth: Synthesising dynamic virtual heart populations through spatiotemporal disentanglement. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. pp. 3–12. Springer (2025)

work page 2025

[8] [8]

In: Proceedings of the 28th ACM international conference on multimedia

Guo, C., Zuo, X., Wang, S., Zou, S., Sun, Q., Deng, A., Gong, M., Cheng, L.: Action2motion: Conditioned generation of 3d human motions. In: Proceedings of the 28th ACM international conference on multimedia. pp. 2021–2029 (2020) 10 X. Yang et al

work page 2021

[9] [9]

Auto-Encoding Variational Bayes

Kingma, D.P., Welling, M.: Auto-encoding variational bayes. arXiv preprint arXiv:1312.6114 (2013)

work page internal anchor Pith review Pith/arXiv arXiv 2013

[10] [10]

Medical image analysis97, 103293 (2024)

Kong, F., Stocker, S., Choi, P.S., Ma, M., Ennis, D.B., Marsden, A.L.: Sdf4chd: Generative modeling of cardiac anatomies with congenital heart defects. Medical image analysis97, 103293 (2024)

work page 2024

[11] [11]

IEEE Journal of Biomedical and Health Informatics27(7), 3302–3313 (2023)

Martín-Isla, C., Campello, V.M., Izquierdo, C., Kushibar, K., Sendra-Balcells, C., Gkontra, P., Sojoudi, A., Fulton, M.J., Arega, T.W., Punithakumar, K., et al.: Deep learning segmentation of the right ventricle in cardiac mri: the m&ms chal- lenge. IEEE Journal of Biomedical and Health Informatics27(7), 3302–3313 (2023)

work page 2023

[12] [12]

Circulation: Cardiovascular Imaging9(2), e003866 (2016)

Muraru, D., Onciul, S., Peluso, D., Soriani, N., Cucchini, U., Aruta, P., Romeo, G., Cavalli, G., Iliceto, S., Badano, L.P.: Sex-and method-specific reference values for right ventricular strain by 2-dimensional speckle-tracking echocardiography. Circulation: Cardiovascular Imaging9(2), e003866 (2016)

work page 2016

[13] [13]

In: Medical Imaging 2007: Physiology, Function, and Structure from Medical Images

Ordas, S., Oubel, E., Leta, R., Carreras, F., Frangi, A.F.: A statistical shape model of the heart and its application to model-based segmentation. In: Medical Imaging 2007: Physiology, Function, and Structure from Medical Images. vol. 6511, pp. 490–500. SPIE (2007)

work page 2007

[14] [14]

In: Proceedings of the AAAI conference on artificial intelligence

Perez, E., Strub, F., De Vries, H., Dumoulin, V., Courville, A.: Film: Visual rea- soning with a general conditioning layer. In: Proceedings of the AAAI conference on artificial intelligence. vol. 32 (2018)

work page 2018

[15] [15]

In: Proceedings of the IEEE/CVF international con- ference on computer vision

Petrovich, M., Black, M.J., Varol, G.: Action-conditioned 3d human motion syn- thesis with transformer vae. In: Proceedings of the IEEE/CVF international con- ference on computer vision. pp. 10985–10995 (2021)

work page 2021

[16] [16]

Nature Machine Intelligence7(5), 800–811 (2025)

Qiao, M., McGurk, K.A., Wang, S., Matthews, P.M., O’Regan, D.P., Bai, W.: A personalized time-resolved 3d mesh generative model for unveiling normal heart dynamics. Nature Machine Intelligence7(5), 800–811 (2025)

work page 2025

[17] [17]

IEEE transactions on medical imaging43(3), 1259–1269 (2023)

Qiao, M., Wang, S., Qiu, H., De Marvao, A., O’Regan, D.P., Rueckert, D., Bai, W.: Cheart: A conditional spatio-temporal generative model for cardiac anatomy. IEEE transactions on medical imaging43(3), 1259–1269 (2023)

work page 2023

[18] [18]

jacc cardiovasc imaging (2024)

Smiseth, O., Rider, O., Cvijic, M., Valkovic, L., Remme, E., Voigt, J.: Myocardial strain imaging: theory, current practice, and the future. jacc cardiovasc imaging (2024)

work page 2024

[19] [19]

JACC: Car- diovascular Imaging18(3), 340–381 (2025)

Smiseth, O.A., Rider, O., Cvijic, M., Valkovič, L., Remme, E.W., Voigt, J.U.: Myocardial strain imaging: theory, current practice, and the future. JACC: Car- diovascular Imaging18(3), 340–381 (2025)

work page 2025

[20] [20]

In: International Conference on Medical Image Computing and Computer-Assisted Intervention

Sørensen, K., Diez, P., Margeta, J., El Youssef, Y., Pham, M., Pedersen, J.J., Kühl, T., De Backer, O., Kofoed, K., Camara, O., et al.: Spatio-temporal neural distance fields for conditional generative modeling of the heart. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. pp. 422–432. Springer (2024)

work page 2024

[21] [21]

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

Wu, Z., Yu, C., Wang, F., Bai, X.: Animateanymesh: A feed-forward 4d foundation model for text-driven universal mesh animation. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 13557–13568 (2025)

work page 2025