Optimization of Recurrence Quantification Analysis for Detecting the Presence of Multiple Sclerosis

Abstract

Purpose

The visual patterns in recurrence plots of time-series data can be quantified using recurrence quantification analysis (RQA), a phase-space-based method. The ability to quantitate recurrence plots affords the possibility of using them to solve central biomedical problems, for example detecting the presence of neurological diseases. Our goal was to assess this application by statistically comparing the values of plot-based quantifiers of electroencephalograms (EEGs) from patients having multiple sclerosis (MS) with values from the EEGs of control subjects.

Methods

First, employing a model system consisting of the addition of known deterministic signals to the EEG of normal subjects, we empirically identified the embedding conditions that facilitated detection of the effect of the addition of the signals. Second, we used the conditions thus identified to compare EEGs from 10 patients with MS and 10 age- and gender-matched control subjects, using seven standard recurrence-plot quantifiers.

Results

We identify embedding dimension of 5 points and time delay of 5 points as conditions that maximize the ability of RQA to detect the presence of deterministic activity in EEGs time series sampled at 500 Hz. The values of the RQA quantifiers computed from the EEGs of the MS patients were significantly greater than the corresponding values from the controls, indicating that the presence of the disease was associated with detectable changes in the EEG (family-wise error < 0.05%).

Conclusions

Recurrence plots detected the occurrence of alterations in EEGs associated with the presence of MS, indicating a decreased complexity (increased order) of brain electrical activity associated with brain disease.

http://bit.ly/2Tsvek5