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

David Wiesner, Julian Suk, Sven Dummer, Tereza Necasová, Vladimír Ulman, David Svoboda, Jelmer M. Wolterink

Abstract: Highlights•Living cell shapes of diverse cell types are generated in 3D+time.•A neural network is optimized to implicitly represent time-evolving cell shapes.•Shape synthesis is possible in virtually unlimited spatial and temporal resolution.•The latent codes and rotations of shapes are disentangled.•Synthesized cell shapes can be used to generate corresponding microscopy images.