Conditional Latent Diffusion Model with Fourier-based Motion Modelling for Virtual Population Synthesis

2026-06-02

AI Summary

Existing methods struggle to generate 4D cardiac dynamic meshes with explicit periodicity, limiting the reliability of in silico clinical trials. This work proposes the 4D F-MeshLDM framework, which introduces truncated Fourier series coefficients as an explicit motion representation within a diffusion model for the first time. By integrating a convolutional mesh VAE, a conditional latent diffusion prior, and an affine modulation mechanism, the framework enables the generation of anatomically faithful and temporally coherent cardiac mesh sequences governed by clinical covariates. Evaluated on 5,000 UK Biobank cases, the method substantially outperforms current approaches, achieving near-zero periodic closure error, superior anatomical fidelity, and accurate preservation of key clinical functional metrics.

Abstract

In-silico trials of medical devices require the generation of virtual populations of anatomies. In cardiovascular applications, virtual anatomy is typically represented as a 3D+t mesh sampled from a generative model. However, most existing mesh generators focus on static anatomy, while sequence models often lack explicit periodicity. To this end, we propose 4D F-MeshLDM, a conditional generative framework comprising a convolutional mesh VAE to encode meshes, a structural latent space that parameterises motion using a truncated Fourier series, and a diffusion prior that learns the latent distribution over Fourier coefficient tokens. By conditioning the diffusion process on clinical covariates via affine modulation, we enable controllable synthesis. Sampling tokens and performing inverse Fourier synthesis yield cycle-consistent latent trajectories, which can be decoded into 3D+t cardiac mesh sequences. Experiments on 5,000 UK Biobank subjects demonstrate that 4D F-MeshLDM outperforms state-of-the-art baselines in anatomical fidelity and achieves near-zero cycle closure error. Furthermore, the generated cohorts accurately preserve clinical functional indices, highlighting the potential of our framework for reliable in-silico cardiac trials.

Problem

• virtual population synthesis
• 4D cardiac mesh
• periodic motion modelling
• in-silico trials
• anatomical fidelity

Innovation

• Latent Diffusion Model
• Fourier-based Motion Modelling
• Virtual Population Synthesis
• Cycle-consistent 4D Mesh Generation
• Conditional Generative Modeling