CardioDiT: Latent Diffusion Transformers for 4D Cardiac MRI Synthesis

• URL: http://arxiv.org/abs/2603.25194v1
• Date: Thu, 26 Mar 2026 08:59:07 GMT
• Title: CardioDiT: Latent Diffusion Transformers for 4D Cardiac MRI Synthesis
• Abstract summary: CardioDiT is a fully 4D latent diffusion framework for short-axis cine CMR synthesis based on diffusion transformers.<n>Atemporal VQ-E encodes 2D+VAt slices into compact latents, which a diffusion transformer then models jointly as complete 3D+t volumes.<n>Results show improved inter-slice consistency, temporally coherent motion, anatomical context and realistic cardiac function distributions.
• Score: 30.824485997130925
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Latent diffusion models (LDMs) have recently achieved strong performance in 3D medical image synthesis. However, modalities like cine cardiac MRI (CMR), representing a temporally synchronized 3D volume across the cardiac cycle, add an additional dimension that most generative approaches do not model directly. Instead, they factorize space and time or enforce temporal consistency through auxiliary mechanisms such as anatomical masks. Such strategies introduce structural biases that may limit global context integration and lead to subtle spatiotemporal discontinuities or physiologically inconsistent cardiac dynamics. We investigate whether a unified 4D generative model can learn continuous cardiac dynamics without architectural factorization. We propose CardioDiT, a fully 4D latent diffusion framework for short-axis cine CMR synthesis based on diffusion transformers. A spatiotemporal VQ-VAE encodes 2D+t slices into compact latents, which a diffusion transformer then models jointly as complete 3D+t volumes, coupling space and time throughout the generative process. We evaluate CardioDiT on public CMR datasets and a larger private cohort, comparing it to baselines with progressively stronger spatiotemporal coupling. Results show improved inter-slice consistency, temporally coherent motion, and realistic cardiac function distributions, suggesting that explicit 4D modeling with a diffusion transformer provides a principled foundation for spatiotemporal cardiac image synthesis. Code and models trained on public data are available at https://github.com/Cardio-AI/cardiodit.

Related papers

• Learning to Beat: Phenotype-Guided Latent Flow with Regional Motion Priors for Biventricular Motion Synthesis [29.123055769120246]
  Full-cycle biventricular motion synthesis from a single end-diastolic (ED) mesh.<n>A functional partition learned from reconstructed motion organizes the ventricular surface into regions with coherent dynamics.<n>Experiments on ACDC, M&Ms, and M&Ms-2 demonstrate consistent improvements in geometric accuracy and functional fidelity.
  arXiv  Detail & Related papers  (2026-08-20T07:37:44Z)
• Personalized 4D Whole-Heart Mesh Reconstruction from Cine MRI via Multi-Scale Temporal Modeling and Differentiable Contour Rendering [7.111748961723236]
  We propose a novel end-to-end framework for reconstructing temporally resolved whole-heart meshes from 2D cine MRI sequences.<n>The framework incorporates a differentiable contour inspired by the Beer-Lambert attenuation principle, enabling anatomy-aware supervision of 3D+t mesh deformation.<n>The proposed method achieved a whole-heart mean absolute error of 1.68 $pm$ 0.31 mm and a motion jitter of 0.77 $pm$ 0.17 $mathrmmm/mathrmframe3$, outperforming existing methods with lower reconstruction error and substantially improved motion smoothness.
  arXiv  Detail & Related papers  (2026-07-02T09:42:46Z)
• Anatomy-Guided Residual Motion Diffusion for Controllable 4D Cardiac MRI Synthesis [15.392128655814451]
  We propose a 4D controllable generative framework for anatomically consistent data augmentation.<n>A semi-supervised variational autoencoder learns a compact latent representation of anatomical volumes.<n>Anatomical structure is disentangled from temporal dynamics through a cascaded latent diffusion model.
  arXiv  Detail & Related papers  (2026-06-25T08:48:45Z)
• Temporally Consistent and Controllable Video Generation of 2D Cine CMR via Latent Space Motion Modeling [7.215198952705435]
  We propose a generative method for synthesizing temporally coherent and anatomically consistent cardiac sequences.<n>Our model generates anatomically and pathologically diverse sequences with high temporal coherence and strong fidelity to input prompts.<n>These experimental results highlight its potential to produce high-fidelity, on-demand medical data, offering a scalable solution to data scarcity.
  arXiv  Detail & Related papers  (2026-06-08T11:05:20Z)
• Conditional Latent Diffusion Model with Fourier-based Motion Modelling for Virtual Population Synthesis [41.938132992227196]
  In cardiovascular applications, virtual anatomy is typically represented as a 3D+t mesh sampled from a generative model.<n>We propose 4D F-MeshLDM, a conditional generative framework comprising a convolutional mesh VAE to encode meshes.<n>We demonstrate that 4D F-MeshLDM outperforms state-of-the-art baselines in anatomical fidelity and achieves near-zero cycle closure error.
  arXiv  Detail & Related papers  (2026-06-02T16:10:00Z)
• Sculpt4D: Generating 4D Shapes via Sparse-Attention Diffusion Transformers [70.55535980817577]
  Sculpt4D is a native 4D generative framework that seamlessly integrates efficient temporal modeling into a pretrained 3D Transformer.<n>We show that Sculpt4D establishes a new state-of-the-art in temporally coherent 4D synthesis and charts a path toward efficient and scalable 4D generation.
  arXiv  Detail & Related papers  (2026-04-23T12:18:55Z)
• Orthogonal Spatial-temporal Distributional Transfer for 4D Generation [58.30004328671699]
  We propose a framework that transfers rich spatial priors from existing 3D diffusion models and temporal priors from video diffusion models to enhance 4D synthesis.<n>We develop a spatial-temporal-disentangled 4D (STD-4D) Diffusion model, which synthesizes 4D-aware videos through disentangled spatial and temporal latents.<n> Experiments demonstrate that our method significantly outperforms existing approaches, achieving superior spatial-temporal consistency and higher-quality 4D synthesis.
  arXiv  Detail & Related papers  (2026-03-05T11:52:21Z)
• Chain of Flow: A Foundational Generative Framework for ECG-to-4D Cardiac Digital Twins [4.615980870195537]
  Chain of Flow (COF) is an ECG-driven generative framework that reconstructs full 4D cardiac structure and motion from a single cardiac cycle.<n>By enabling full 4D organ reconstruction directly from ECG, COF transforms cardiac digital twins into fully generative, patient-specific virtual hearts.
  arXiv  Detail & Related papers  (2026-02-26T12:06:43Z)
• 3D Cardiac Anatomy Generation Using Mesh Latent Diffusion Models [9.749160335898255]
  We investigate the application of Latent Diffusion Models for generating 3D meshes of human cardiac anatomies.<n>We apply the proposed model on a dataset of 3D meshes of left ventricular cardiac anatomies from patients with acute myocardial infarction.<n>The MeshLDM successfully captures characteristics of the cardiac shapes at end-diastolic (relaxation) and end-systolic (contraction) cardiac phases, generating meshes with a 2.4% difference in population mean compared to the gold standard.
  arXiv  Detail & Related papers  (2025-08-18T16:53:20Z)
• Articulated Kinematics Distillation from Video Diffusion Models [64.18346374534352]
  Articulated Kinematics Distillation (AKD) is a framework for generating high-fidelity character animations.<n>AKD uses a skeleton-based representation for rigged 3D assets, drastically reducing the Degrees of Freedom (DoFs)<n>AKD distills complex, articulated motions while maintaining structural integrity.
  arXiv  Detail & Related papers  (2025-04-01T21:37:57Z)
• X$^{2}$-Gaussian: 4D Radiative Gaussian Splatting for Continuous-time Tomographic Reconstruction [64.2059940799033]
  Current methods discretize temporal resolution into fixed phases with respiratory gating devices.<n>X$2$-Gaussian, a novel framework, enables continuous-time 4DCT reconstruction by integrating dynamic radiative splatting with self-supervised respiratory motion learning.
  arXiv  Detail & Related papers  (2025-03-27T17:59:57Z)
• Towards Unified and Lossless Latent Space for 3D Molecular Latent Diffusion Modeling [90.23688195918432]
  3D molecule generation is crucial for drug discovery and material science.<n>Existing approaches typically maintain separate latent spaces for invariant and equivariant modalities.<n>We propose textbfUAE-3D, a multi-modal VAE that compresses 3D molecules into latent sequences from a unified latent space.
  arXiv  Detail & Related papers  (2025-03-19T08:56:13Z)
• MedTet: An Online Motion Model for 4D Heart Reconstruction [59.74234226055964]
  We present a novel approach to reconstruction of 3D cardiac motion from sparse intraoperative data.<n>Existing methods can accurately reconstruct 3D organ geometries from full 3D volumetric imaging.<n>We propose a versatile framework for reconstructing 3D motion from such partial data.
  arXiv  Detail & Related papers  (2024-12-03T17:18:33Z)
• A personalized time-resolved 3D mesh generative model for unveiling normal heart dynamics [6.6350578770951385]
  We develop a conditional generative model, MeshHeart, to learn the distribution of shape and motion patterns for the left ventricles of the heart.<n>Based on MeshHeart, we investigate the latent space of 3D+t cardiac mesh sequences and propose a distance metric, latent delta, which quantifies the deviation of a real heart from its personalised normative pattern.
  arXiv  Detail & Related papers  (2024-09-20T18:08:37Z)
• Epicardium Prompt-guided Real-time Cardiac Ultrasound Frame-to-volume Registration [50.602074919305636]
  This paper introduces a lightweight end-to-end Cardiac Ultrasound frame-to-volume Registration network, termed CU-Reg.<n>We use epicardium prompt-guided anatomical clues to reinforce the interaction of 2D sparse and 3D dense features, followed by a voxel-wise local-global aggregation of enhanced features.
  arXiv  Detail & Related papers  (2024-06-20T17:47:30Z)
• DMCVR: Morphology-Guided Diffusion Model for 3D Cardiac Volume Reconstruction [33.59945107137013]
  Current cardiac MRI-based reconstruction technology is 2D with limited through-plane resolution. We propose a morphology-guided diffusion model for 3D cardiac volume reconstruction, DMCVR, that synthesizes high-resolution 2D images and corresponding 3D reconstructed volumes.
  arXiv  Detail & Related papers  (2023-08-18T00:48:30Z)
• DeepRecon: Joint 2D Cardiac Segmentation and 3D Volume Reconstruction via A Structure-Specific Generative Method [12.26150675728958]
  We propose an end-to-end latent-space-based framework, DeepRecon, that generates multiple clinically essential outcomes. Our method identifies the optimal latent representation of the cine image that contains accurate semantic information for cardiac structures. In particular, our model jointly generates synthetic images with accurate semantic information and segmentation of the cardiac structures.
  arXiv  Detail & Related papers  (2022-06-14T20:46:11Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.