Title: Novel tools for automated anatomical mapping and behavioral tracking

When: Thursday, November 19th, 3pm

Where:

https://columbiauniversity.zoom.us/j/93796332541?pwd=TVRHc1RwOVUyR2RlN3M5d0trWHk1Zz09  

Meeting ID: 937 9633 2541 

Passcode: 033904

Speakers: 

Luke Hammond (Peterka Lab/Cellular Imaging):

Title: Automating imaging and analysis of whole brains and spinal cords
Abstract: Reconstructing serial tissue sections into whole brains and analyzing them within a common coordinate framework is an essential capability for the discovery of novel circuits and studies in neuroscience. Many existing solutions require coding experience, commercial software, or labor-intensive interaction and annotation for researchers. With an emphasis on accessibility and automation, BrainJ enables high-throughput analysis of serial tissue sections imaged using confocal or widefield imaging techniques. Developed in Fiji, this approach leverages freely available tools, including machine learning pixel classification for cell detection and mesoscale mapping of axons and dendrites. By creating a 3D spinal cord atlas, we have extended this approach to perform comparative analyses of neurons and their projections in the whole mouse spinal cord.

Anqi Wu  (Paniski Lab):  
Title: Exploiting unlabeled frames to build better models for behavioral video analysis  
Abstract: Video cameras are finding increasing use in the study of animal behavior over short ranges and rendering fruitful descriptions for neuro-behavioral analysis in neuroscience. However, analyzing the high dimensional videos with raw pixels is a challenging statistical problem. To address this, we aim at extracting meaningful information from videos by finding suitable low-dimensional representations using pose tracking, segmentation, and video compression. We propose a novel tracking algorithm that leverages spatial statistics imposed by skeletal constraints and temporal continuity, and exploits both labeled and unlabeled frames to achieve significantly more accurate and robust tracking over different species of animals in various tasks. In turn, these tracking improvements enhance performance on downstream applications, including robust unsupervised segmentation of behavioral syllables, and estimation of interpretable disentangled low-dimensional representations of the full behavioral video.

Motor Club Slack Workspace:

motorclubworkspace.slack.com