Learning Biological Priors for Large-scale Connectomics Reconstruction

Par

Donglai Wei

Boston College

Résumé:

The field of connectomics aims to reconstruct the brain's wiring diagram from nanometer-resolution 3D microscopy image volumes to enable new insights into the workings of brains. These new insights could inspire novel artificial intelligence algorithms and benefit the treatment development for neurodegenerative diseases. However, the current reconstruction process is time-consuming due to the costly annotation and complex neuron morphology. In this talk, I will present two deep learning methods to leverage biological priors to speed up the reconstruction of the nodes (neurons) and edges (synapses) of the wiring diagram. First, for 3D synapse detection, we exploited the unsupervised appearance prior to improve deep active learning (ECCV 2020). Next, for neuron segmentation, we built and optimized a biologically constrained graph over initial 3D segments for error detection and correction (CVPR 2019). With these methods, we have densely annotated connectomics datasets that are at least 100x larger than the prior art (MICCAI 2021).

An interesting read on Connectomics (Yann LeCun is a co-author): The mind of a mouse. https://pubmed.ncbi.nlm.nih.gov/32946777/

Biographie :

Donglai Wei is an assistant professor in the Computer Science Department at Boston College. His research focuses on developing novel registration and reconstruction algorithms for large-scale (currently petabyte-scale) connectomics datasets to empower neuroscience discoveries. During his Ph.D. at MIT under Prof. William Freeman, he worked on video understanding problems, including arrow of time and Vimeo-90K benchmark. Since his postdoc at Harvard University, he has embarked on the quest to reconstruct the brain's wiring diagram in collaboration with Prof. Hanspeter Pfister, Prof. Jeff Lichtman, and Prof. Ed Boyden.