AutomaticSleepScoringTool

May 12, 2022

The identification of behavioral states based on physiological measures such as EEG, EMG, and LFP is a non-trivial task and much needed for advancing research on topics such as sleep. Jacob Ellen and Michael Dash from Middlebury College created an automated neural network pipeline for classifying behavioral states in rats. Their AutomaticSleepScoringTool takes inputs from electrophysiological readings (EEG/EMG/LFP) and classifies behavioral states automatically. With minimal manual classification and low computational complexity, this artificial neural network can be used for accurate and efficient behavioral state classifications. A small initial training set (around 40 minutes) is recommended to achieve high classification performance. Additionally, algorithmic errors of behavioral states can be manually re-scored to improve models fit. This algorithm uses R with additional packages (Keras and TensorFLow) and is entirely open-source.

Thanks to Caleb Darden, a rising senior majoring in Neuroscience at American University, for writing up this post.

Read the Paper

Learn more about using AutomaticSleepScoringTool

AutomaticSleepScoringTool on GitHub

Get access to the software for AutomaticSleepScoringTool from GitHub!

Have questions? Send us an email!