Abstract:Data-driven cell tracking and segmentation methods in biomedical imaging require diverse and information-rich training data. In cases where the number of training samples is limited, synthetic computer-generated data sets can be used to improve these methods. This requires the synthesis of cell shapes as well as corresponding microscopy images using generative models. To synthesize realistic living cell shapes, the shape representation used by the generative model should be able to accurately represent fine details and changes in topology, which are common in cells. These requirements are not met by 3D voxel masks, which are restricted in resolution, and polygon meshes, which do not easily model processes like cell growth and mitosis. In this work, we propose to represent living cell shapes as level sets of signed distance functions (SDFs) which are estimated by neural networks. We optimize a fully-connected neural network to provide an implicit representation of the SDF value at any point in a 3D+time domain, conditioned on a learned latent code that is disentangled from the rotation of the cell shape. We demonstrate the effectiveness of this approach on cells that exhibit rapid deformations (Platynereis dumerilii), cells that grow and divide (C. elegans), and cells that have growing and branching filopodial protrusions (A549 human lung carcinoma cells). A quantitative evaluation using shape features and Dice similarity coefficients of real and synthetic cell shapes shows that our model can generate topologically plausible complex cell shapes in 3D+time with high similarity to real living cell shapes. Finally, we show how microscopy images of living cells that correspond to our generated cell shapes can be synthesized using an image-to-image model.

Genetic cellular neural networks for generating three-dimensional geometry

SDF-based 3D Generative Adversarial Networks for Image and Geometry Generation

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

Implicit Neural Representations for Generative Modeling of Living Cell Shapes

Growing 3D Artefacts and Functional Machines with Neural Cellular Automata

MagicClay: Sculpting Meshes With Generative Neural Fields

A Generative Model for Creating Path Delineated Helical Proteins

Automatic design of gene regulatory mechanisms for spatial pattern formation

Cellular Genetic Algorithm for Supporting Creative Design of 3D Animation Modelings

MeshXL: Neural Coordinate Field for Generative 3D Foundation Models

Synthetic Generation of Three-Dimensional Cancer Cell Models from Histopathological Images

MeshingNet: A New Mesh Generation Method based on Deep Learning

Neural Progressive Meshes

Epigenetic Tracking, a Method to Generate Arbitrary Shapes By Using Evolutionary-Developmental Techniques

Multi-chart Generative Surface Modeling

Neural population geometry: An approach for understanding biological and artificial neural networks

Evo-Devo Algorithms: Gene-Regulation for Digital Architecture

Neural Volumetric Mesh Generator

Evolutionary Design Based on an Improved Genetic Algorithm and Generative Three-dimensional Shape

NeuSDFusion: A Spatial-Aware Generative Model for 3D Shape Completion, Reconstruction, and Generation

StructureNet: Hierarchical Graph Networks for 3D Shape Generation