Seung Lab

Computational Neuroscience @ Princeton

The Seung Lab uses techniques from machine learning and social computing to extract brain structure from light and electron microscopic images.  EyeWire showcases our approach by mobilizing gamers from around the world to create 3D reconstructions of neurons by interacting with a deep convolutional network.

The Seung Lab also develops computational methods for relating brain structure to function. To establish this relationship, we often reconstruct the connectivity of the same neurons after observation of their activity via two-photon imaging. We also classify neurons into cell types that have characteristic structural and functional properties. The latter approach was used to create a new model for how direction selectivity emerges in the mammalian retina, based on the 3D reconstructions from EyeWire. For further insights into function, we also reconstruct neural circuits after manipulation of neurons via viral or genetic techniques.

We are best known for our work on the reconstruction of neural circuits using serial electron microscopy. At the present time, we are only able to reconstruct a tiny fraction of neural circuitry in the mammalian brain. A long-term goal is to accelerate image analysis until it becomes practical to reconstruct complete connectomes of mammalian brains. This technological goal complements exciting progress in the imaging of neural activity. Knowing the connectivity and activity of every cell in a neural circuit will radically accelerate the progress of neuroscience.

We also study brain structure using light microscopy, which has inferior spatial resolution to electron microscopy but provides a larger field of view and molecular information from fluorescent probes. We use light microscopy for the structural classification of cell types and cross-validation with molecularly defined types. We also develop computational tools for handling whole brain images from light microscopy.