Abstract
Neuroscientists use genetically encoded indicators and advanced microscopy to optically record the activities of numerous neurons in animal brains at high speed. The extraction of neural activity from individual neurons in imaging movies requires a multistep video processing pipeline, including correcting motion artifacts, segmenting spatial footprints of neurons, extracting temporal traces, and inferring activity spikes. Many neuron segmentation and spike inference algorithms have been developed using various machine learning techniques, including unsupervised learning, supervised learning, and a combination of both. The neuron segmentation algorithms work for different input data types, including 2D summary images, 3D videos as a whole or in blocks, or 3D videos frame-by-frame. These methods can replace tedious human labeling and automate the analysis of neural activity. Fast online processing methods potentially enable real-time closed-loop neuroscience experiments.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have