Petascale Connectomics and Beyond - IPAM at UCLA

Explore a comprehensive lecture on petascale connectomics and its future developments. Delve into the world of brain connectivity mapping as Sebastian Seung from Princeton University presents groundbreaking research at IPAM's Diffractive Imaging with Phase Retrieval Workshop. Learn about the evolution of connectome reconstruction techniques, from manual labor to automated systems using convolutional neural networks. Discover the progress made in mapping the fruit fly Drosophila's connectome and ongoing efforts in cerebral cortex reconstruction. Gain insights into how connectomic information reveals innate nervous system structures and its potential impact on brain learning theories. Examine the challenges and possibilities of exascale connectomics, including plans for mapping an entire mouse brain and incorporating electron tomography for improved axial resolution. Understand the technological revolution driven by the fusion of advanced imaging techniques and computational power in neuroscience research.

Intro
Serial section electron microscopy (EM)
EM image dataset from fruit fly brain
Software stack for petascale connectomics
Completion of Drosophila connectome expected 2023
What can we do with connectomes?
What can a computer scientist do with connectomes?
Graph reduction: 50,000 fly neurons→ 28 clusters
Example cluster: ellipsoid body (important in navigation)
Artificial neural networks
Brains
NIH BRAIN CONNECTS (beginning 2023)
Competing proposals for exascale connectomics
Serial section tilt series tomography
Why is this challenging?
Tomography with convolutional nets
Tomography with 5 dual axis simulated tilts
Image reconstruction from 5 dual axis simulated tilts
How to generate ground truth from real data?
New implementation of iterative reconstruction
Application to mouse retina
Plan for handling real images
Technological revolution driven by imaging + computing