A signal analysis approach to rat sleep scoring instrumentation

Abstract

Automated rat sleep analysis focuses on the statistically regular waveforms of the EEG, such as theta and delta rhythms. Such stochastic processes can be quantified in several manners. Time domain statistics such as auto- and cross-correlations produce outputs that are difficult to use and are best performed in software. Frequency domain statistics like spectral density accurately quantify the sleep state by power-frequency distributions but also require sophisticated computer processing. Continuous frequency analysis, using pass-band filtering, accurately measures signal power in an on-line fashion and employs relatively inexpensive hardware to estimate power by integrating the square of the signal. This method differs substantively from other previously reported systems which rely on signal amplitude analysis. Comparison of this system with a human scorer indicates high degrees of validity and reproducibility.