Detection of sleep and wake substates through unsupervised machine learning in rat hippocampus and primary motor cortex LFP recordings

This project investigated whether unsupervised machine learning could detect differences in global or local microstates across different rodent brain, in the effects of procedural learning, and clustering validity indices effectiveness. Previously obtained local field potential recordings of M1 and the hippocampus of freely-behaving male rats under naïve and task conditions, including transcranial direct and alternating current stimulation (tDCS; tACS), were analyzed and used to assess several methods. Two local SWS-like REM microstates were detected along with five global microstates. Learning suppressed cortical SWS-like REM microstates, but tDCS negated this effect. Calinski-Harabasz evaluated clusters had the highest sensitivity, specificity and total accuracy. Local and global brain states were effectively detected using PCA and clustering, and measures of phase-amplitude coupling were sensitive to the task conditions. These changes could underlie consolidation windows for procedural learning with potential intervention by tDCS although these results are limited to due the quality of the dataset.

Authors

• Balogun, Pauline
• University of Lethbridge. Faculty of Arts and Science

• neuroscience
• k-means
• machine learning
• hippocampus
• memory
• primary motor cortex
• clustering
• cluster validity index
• PCA
• behaviour
• tACS
• tDCS
• tPAC
• Hippocampus (Brain) -- Research
• Memory -- Research
• Motor cortex -- Research
• Sleep -- Research
• Sleep-wake cycle -- Research
• Rats as laboratory animals
• Cluster analysis -- Computer programs
• Dissertations, Academic

Identifier

hdl:10133/5685

OAI identifier oai:identifier

oai:opus.uleth.ca:10133/5685