No abstract

This work presents the application of nonlinear dynamics measures to electroencephalograms (EEG) acquired from patients with Attention Deficit/Hyperactivity Disorder (ADHD) before and after a neurofeedback therapy, with the aim to assess the effects of the neurofeedback in a quantitative way. The database contains EEG registers of seven patients acquired in eyes-closed and eyes-opened conditions, in pre-and post-treatment phases. Five measures were applied: largest Lyapunov exponent, Lempel-Ziv complexity, Hurst exponent, and multiscale entropy on two different scales. The purpose is to test whether these measures are apt to detect and quantify differences from EEG registers between pre- and post-treatment. The results indicate that these measures could have a potential utility for detection of quantitative changes in specific EEG channels. In addition, the performance of some of these measures improved when the bandwidth was reduced to 3-30 Hz.

Electroencephalogram (EEG) contains valuable information obtained noninvasively that can be used for assessment of brain's processing capacity of patients with psychiatric disorders. The purpose of the present work was to evaluate possible differences in EEG complexity between deficit (DS) and nondeficit (NDS) subtypes of schizophrenia as a reflection of the cognitive processing capacities in these groups. A particular nonlinear metric known as Lempel-Ziv complexity (LZC) was used as a computational tool in order to determine the randomness in EEG alpha band time series from 3 groups (deficit schizophrenia [n = 9], nondeficit schizophrenia [n = 10], and healthy controls [n = 10]) according to time series randomness. There was a significant difference in frontal EEG complexity between the DS and NDS subgroups ( p = .013), with DS group showing less complexity. A significant positive correlation was found between LZC values and Positive and Negative Syndrome Scale (PANSS) general psychopathology scores (ie, larger frontal EEG complexity correlated with more severe psychopathology), explained partially by the emotional component subscore of the PANSS. These findings suggest that cognitive processing occurring in the frontal networks in DS is less complex compared to NDS patients as reflected by EEG complexity measures. The data also suggest that there may be a relationship between the degree of emotionality and the complexity of the frontal EEG signal.

No abstract

We explore how the reconstruction efficiency of fast spike population codes varies with population size, population composition and code complexity. Our study is based on experiments with moving light patterns which are projected onto the isolated retina of a turtle Pseudemys scripta elegans. The stimulus features to reconstruct are sequences of velocities kept constant throughout segments of 500 ms. The reconstruction is based on the spikes of a retinal ganglion cell (RGC) population recorded extracellularly via a multielectrode array. Subsequent spike sorting yields the parallel spike trains of 107 RGCs as input to the reconstruction method, here a discriminant analysis trained and tested in jack-knife fashion. Motivated by behavioral response times, we concentrate on fast reconstruction, i.e., within 150 ms following a trigger event defined via significant changes of the population spike rate. Therefore, valid codes involve only few (≤3) spikes per cell. Using only the latency t(1) of each cell (with reference to the trigger event) corresponds to the most parsimonious population code considered. We evaluate the gain in reconstruction efficiency when supplementing t(1) by spike times t(2) and t(3). Furthermore, we investigate whether sub-populations of smaller size benefit significantly from a selection process or whether random compilations are equally efficient. As selection criteria we try different concepts (directionality, reliability, and discriminability). Finally, we discuss the implications of a selection process and its inter-relation with code complexity for optimized reconstruction.