Abstract:We propose a deep learning method to model and generate synthetic aortic shapes based on representing shapes as the zero-level set of a neural signed distance field, conditioned by a family of trainable embedding vectors with encode the geometric features of each shape. The network is trained on a dataset of aortic root meshes reconstructed from CT images by making the neural field vanish on sampled surface points and enforcing its spatial gradient to have unit norm. Empirical results show that our model can represent aortic shapes with high fidelity. Moreover, by sampling from the learned embedding vectors, we can generate novel shapes that resemble real patient anatomies, which can be used for in-silico trials.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to generate realistic 3D aortic shapes in in - silico trials in the cardiovascular field. Specifically, the researchers proposed a deep - learning - based method to model and generate synthetic aortic shapes. These shapes can be used to simulate medical procedures such as Transcatheter Aortic Valve Implantation (TAVI), thereby accelerating the testing process of medical devices and being able to take into account a wider range of population anatomical differences. ### Main contributions of the paper include: 1. **Proposing a new deep - learning method**: By representing shapes as the zero - level set of the Neural Signed Distance Field (NSDF), this method can learn the geometric features of different aortic shapes. 2. **Improving the shape reconstruction accuracy**: Experimental results show that this model can reconstruct aortic shapes with high fidelity. 3. **Generating novel and anatomically reasonable shapes**: By sampling from the learned embedding vectors, new shapes can be generated. These shapes are similar to the anatomical structures of real patients and are suitable for in - silico trials. 4. **Exploring the diversity of generated shapes**: The study found that by adjusting the number of embedding vectors used in the generation process, a balance can be found between the diversity and consistency of shapes. ### Method overview: - **Model architecture**: An 8 - layer fully - connected autoencoder network with skip - connections was used to improve the stability of the network and the effective use of high - dimensional embedding vectors. - **Loss function**: The loss function consists of three parts: - The first part ensures that the output of the network on surface points is close to zero and maintains the consistency of normal vectors. - The second part is implicit geometric regularization, which encourages the network to generate a signed distance field. - The third part is the \(L_2\) regularization of the embedding vectors. - **Dataset**: The training dataset includes 97 normalized aortic root (non - closed) mesh models, with 500,000 surface points and their unit normal vectors sampled for each mesh model. ### Experimental results: - **Shape reconstruction**: The similarity between the reconstructed shapes and the original shapes was evaluated by the Chamfer distance, and the results showed that most shapes could be accurately reconstructed. - **Shape generation**: New shapes were generated by interpolating multiple embedding vectors. Experiments showed that increasing the number of interpolation vectors can reduce the variance of the generated shapes, but at the same time it will also reduce the diversity. ### Conclusions and future work: - **Conclusions**: This method can effectively generate synthetic aortic shapes for cardiovascular in - silico trials, but it has certain limitations in terms of diversity and variance. - **Future work**: Exploring more advanced generation techniques (such as generative adversarial networks or diffusion models), introducing attention mechanisms, and expanding the functions of the model, such as generating features such as calcification, will further improve the accuracy and effectiveness of the model.

Generative 3D Cardiac Shape Modelling for In-Silico Trials

Shape of my heart: Cardiac models through learned signed distance functions

SDF4CHD: Generative Modeling of Cardiac Anatomies with Congenital Heart Defects

Interpretable cardiac anatomy modeling using variational mesh autoencoders

Learning Interpretable Anatomical Features Through Deep Generative Models: Application to Cardiac Remodeling

Patient-specific Heart Geometry Modeling for Solid Biomechanics using Deep Learning

A Generative Shape Compositional Framework to Synthesise Populations of Virtual Chimaeras

A Generative Shape Compositional Framework to Synthesize Populations of Virtual Chimeras

Generation of patient specific cardiac chamber models using generative neural networks under a Bayesian framework for electroanatomical mapping

Synthetic Data Generation for 3D Myocardium Deformation Analysis

A Deep-Learning Approach For Direct Whole-Heart Mesh Reconstruction

Spatio-temporal neural distance fields for conditional generative modeling of the heart

3D Cardiac Shape Prediction with Deep Neural Networks: Simultaneous Use of Images and Patient Metadata

A Personalised 3D+t Mesh Generative Model for Unveiling Normal Heart Dynamics

Generative modeling of living cells with SO(3)-equivariant implicit neural representations

Guiding 3D U-nets with signed distance fields for creating 3D models from images

A Deep Learning-Based Method for Rapid 3D Whole-Heart Modeling in Congenital Heart Disease

Learning Whole Heart Mesh Generation From Patient Images For Computational Simulations

Blood Vessel Geometry Synthesis using Generative Adversarial Networks

Neural Deformable Models for 3D Bi-Ventricular Heart Shape Reconstruction and Modeling from 2D Sparse Cardiac Magnetic Resonance Imaging

Discriminative and Generative Models for Anatomical Shape Analysison Point Clouds with Deep Neural Networks