Flow-based conditional cardiac anatomy generation for virtual cohorts

2026-08-10

AI Summary

Constructing clinically representative synthetic cohorts of medical images is highly challenging due to limitations in cohort size, subgroup sparsity, and data-sharing constraints. To address this, this work proposes the CAN-FLOW framework, which first learns a geometry-only latent representation of cardiac shape deformations via diffeomorphic shape momentum and then models the joint distribution of this latent space with demographic metadata—specifically sex, age, and BMI—using conditional normalizing flows. This approach enables disentangled representation learning and metadata-conditioned generation. Notably, it represents the first application of conditional normalizing flows to cardiac anatomical synthesis. Evaluated on 2,208 healthy participants from the UK Biobank, the method generates biventricular structures that more accurately reproduce clinical phenotype distributions, metadata dependencies, subgroup variability, and high-dimensional shape diversity.

Abstract

Cardiac digital twin research is moving from subject-specific anatomical replicas toward virtual cohorts that represent clinically relevant population subgroups. Yet access to representative imaging-derived anatomy datasets remains limited by cohort size, subgroup sparsity, and data-sharing constraints. Conditional generative models could help address this gap, but virtual cohorts are useful only if they preserve realistic, metadata-dependent anatomical variability. Existing cardiac anatomy generators largely rely on conditional variational autoencoders (cVAEs), which couple representation learning and metadata conditioning through a shared regularized latent prior. We introduce CAN-FLOW, a two-step Conditional ANatomy generation framework based on normalizing FLOWs that first learns geometry-only latent representations of diffeomorphic cardiac shape momenta and then models their sex-, age-, and body-mass-index-dependent distribution with a conditional normalizing flow. We trained CAN-FLOW on 2,208 healthy UK Biobank subjects and compared it with cVAEs across regularization strengths. CAN-FLOW generated plausible stochastic biventricular anatomies that better reproduced clinical phenotype distributions, metadata-dependent trends, subgroup variability, point-cloud coverage, and high-dimensional shape variability. Together, these results establish CAN-FLOW as a shareable framework for generating realistic, stochastically varying, metadata-conditioned biventricular anatomies for virtual cohort construction and in silico clinical trial workflows.

Problem

• virtual cohorts
• cardiac anatomy generation
• metadata-dependent variability
• data scarcity
• anatomical realism

Innovation

• conditional normalizing flows
• cardiac anatomy generation
• virtual cohorts
• diffeomorphic shape momenta
• metadata-conditioned generation