We develop algorithms, systems and software architectures for automating reconstruction of accurate representations of neural tissue structures, such as nanometer-scale neurons' morphology and synaptic connections in the mammalian cortex.

The field of connectomics faces unprecedented "big data" challenges. To reconstruct neuronal connectivity, automated pixel-level segmentation is required for petabytes of streaming electron microscopy data. Existing algorithms provide relatively good accuracy but are unacceptably slow, and would require years to extract connectivity graphs from even a single cubic millimeter of neural tissue. We offer few viable real-time solutions optimized for shared-memory multicore systems, capable of processing data at near the terabyte-per-hour pace of multi-beam electron microscopes. As an example, our deep learning architecture and algorithms were used to reconstruct, from image stack to skeletons, a somatosensory cortex dataset (Kasthuri et al.; 463 GB capturing 120,000 cubic microns) in a matter of hours on a single multicore machine rather than the weeks it has taken in the past on much larger distributed systems, attaining the accuracy of state of the art slower reconstruction systems.