Our goal is to leverage Google expertise and resources to advance understanding of the structure and function of the brain.
About our team
A major hypothesis in modern neuroscience is that neuron-to-neuron connectivity structure in the brain can be linked to function -- how the brain encodes memories, extracts features from perceptual stimuli, and makes decisions. However, the structure of these brain networks has remained largely unknown, due to technical difficulties involved in imaging and reconstructing the brain in 3D.
New microscopy techniques have begun to address the challenge of imaging the brain in 3D at nanometer resolution, and 4D at cellular resolution, but this has led to a huge bottleneck in the subsequent step of data analysis. Our goal is to help solve some of these data analysis problems and thus enable a high-throughput approach to studying the network architecture of the brain.
Team focus summaries
Automated 3d Brain Reconstruction
We develop algorithms and software for automating the process of aligning, segmenting, and annotating petabyte-scale 3d images of brain tissue.
In collaboration with colleagues at the Max Planck Institute of Neurobiology, we published "High-Precision Automated Reconstruction of Neurons with Flood-Filling Networks" in Nature Methods, which shows how a new type of recurrent neural network can improve the accuracy of automated interpretation of connectomics data by an order of magnitude over previous deep learning techniques.
Neuroglancer is a WebGL-based viewer for petascale multidimensional data. It is capable of displaying arbitrary (non axis-aligned) cross-sectional views of volumetric data, as well as 3-D meshes and line-segment based models (skeletons).
Cloud-first library for reading and writing large multi-dimensional arrays, with advanced concurrency and indexing capabilities.