Deformable registration and generative modelling of aortic anatomies by auto-decoders and neural ODEs

2025-06-01

AI Summary

This work addresses deformable vascular registration and synthetic generation, proposing an implicit deformation modeling framework integrating neural ordinary differential equations (ODEs) and a self-decoder. The aorta is represented as a weighted point cloud; continuous spatial deformations are modeled via neural ODEs, while the self-decoder jointly optimizes low-dimensional latent codes and network parameters. Our approach innovatively combines implicit point-cloud representation, Chamfer-distance-driven forward/backward ODE integration, and disentangled latent-space optimization. It enables cross-subject generalization, parameter sharing, and lightweight inference via latent-code fine-tuning only; novel anatomically plausible aortic structures can be sampled directly from the latent space. Evaluated on healthy aortic data, the method achieves sub-millimeter registration accuracy, produces photorealistic synthetic geometries, and significantly outperforms conventional approaches in computational efficiency—effectively balancing high precision with real-time applicability.

Technology Category

• Machine Learning: Deep Generative Models & Autoencoders
• Computer Vision: Diffusion Models for Vision
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

This work introduces AD-SVFD, a deep learning model for the deformable registration of vascular shapes to a pre-defined reference and for the generation of synthetic anatomies. AD-SVFD operates by representing each geometry as a weighted point cloud and models ambient space deformations as solutions at unit time of ODEs, whose time-independent right-hand sides are expressed through artificial neural networks. The model parameters are optimized by minimizing the Chamfer Distance between the deformed and reference point clouds, while backward integration of the ODE defines the inverse transformation. A distinctive feature of AD-SVFD is its auto-decoder structure, that enables generalization across shape cohorts and favors efficient weight sharing. In particular, each anatomy is associated with a low-dimensional code that acts as a self-conditioning field and that is jointly optimized with the network parameters during training. At inference, only the latent codes are fine-tuned, substantially reducing computational overheads. Furthermore, the use of implicit shape representations enables generative applications: new anatomies can be synthesized by suitably sampling from the latent space and applying the corresponding inverse transformations to the reference geometry. Numerical experiments, conducted on healthy aortic anatomies, showcase the high-quality results of AD-SVFD, which yields extremely accurate approximations at competitive computational costs.

Problem

• Deformable registration of vascular shapes to a reference
• Generation of synthetic anatomies using neural ODEs
• Efficient shape representation with low-dimensional latent codes

Innovation

• Auto-decoder structure for efficient weight sharing
• Neural ODEs for deformable registration of shapes
• Implicit shape representations enable generative applications