Time-Scale 2D Representation of EEG Long-Term Records as a Convenient Instrument for Big Data Analysis in Neurology and Psychiatry

摘要

The report is devoted to computer analysis of long-term EEG records. The visual review of such big data requires a lot of time and is not always unbiased. That is why the search for efficient instruments of long-term EEG records processing is a task of utmost importance for neurologic analysis. To solve this problem, we developed a convenient instrument that expands a one-dimensional EEG signal into a two-dimensional representation. In contrast to the well-known time-frequency representations we choose for EEG-like signals another representation domain - the time-scale one. This is done because fragments of interest in the EEG signal - seizures, coma signals, alpha-spindles and others often represent a few repeating waveforms, whose spectra are obscure. The most effective here is the correlative-type analysis, a special form of which - the multi-scale correlative analysis is put forward as the basis for a novel representation. The paper discusses in detail the multi-scale correlation algorithm of the application developed and shows how the suggested functionality of the instrument is provided by the algorithmic features. The performance of the instrument is illustrated by the results of processing rats long-term EEG, in particular, before and after traumatic brain injury.