Neuronal connectivity can be reconstructed from a 3D electron microscopy (EM) image of a brain volume. Connectomics is a modern reincarnation of neuroanatomy, aiming to densely reconstruct neurons, exhaustively detect synapses, and extract a comprehensive wiring diagram from a brain volume. For the past decade, deep learning has been extensively employed to reconstruct neural circuits from EM brain images. In this seminar, I will begin by introducing our deep learning-based computational pipeline for processing petascale EM image datasets in the field of connectomics. I will then spotlight automated reconstructions from both terascale and petascale EM image datasets from the fly and mouse brains. Notably, our pipeline’s automated reconstruction has been enhanced by 30 person-years of meticulous proofreading, culminating in the very first neuronal wiring diagram of an adult brain. I will discuss how connectomes will impact the future trajectory of neuroscience. To conclude, I will share insights on the rapidly evolving landscape of connectomics, with an ambitious vision of achieving zettascale reconstruction of the human brain.
Dr. Kisuk Lee is a connectomics research scientist at Zetta AI LLC. Prior to Zetta AI, he served as a postdoctoral research associate at the Princeton Neuroscience Institute at Princeton University, working with H. Sebastian Seung. Dr. Lee received his Ph.D. in Computation from the Massachusetts Institute of Technology in 2019, and B.S. in Computer Science & Engineering and Biological Sciences from Seoul National University in 2012. His research specializes in deep learning-based image analysis for connectomics, with a specific focus on neuron reconstruction from 3D electron microscopy images.