Background: Normal brain function depends on intact interactions between multiple neuronal ensembles. Interactions within and between local networks comprising multiple neuronal types may occur on a range of time scales thus affecting the estimation of interaction strength. A common technique to investigate functional interactions within neuronal ensembles is pairwise cross-correlation analysis. However, conventional cross-correlation methods address the question of whether an observed peak in the cross-correlation is statistically significant relative to the null hypothesis which assumes a lack of correlation. Ultimately, these methods were not designed to evaluate the strength of the observed interactions. New method: We devised four complementary measures - Triplets, Bin crossing, Bin height and Entropy - for assessing the strength of neuronal interactions; each is sensitive to different features of the cross-correlogram peak such as height, width and smoothness. Results: First, a comparison of five prevalent methods for evaluating whether an observed peak in neuronal cross-correlogram is significant allowed their ranking from the most conservative to the more sensitive for purposes of selecting the appropriate method based on the data structure and preferred strategy. Second, the performance of the four measures we derived improved with interaction strength and the number of spikes in the cross-correlogram. The four measures also enabled the reconstruction of interaction parameters of simulated networks including the detection of time-dependent alterations. Conclusions: We suggest that the combination of several measures of peak characteristics helps rectify the individual shortcomings of specific measures and can yield a broad coverage of interaction strengths and widths.
Bibliographical noteFunding Information:
The study was financed in part by a REALNET ( FP7-ICT270434 ) grant from the European Commission .
- Confidence limits
- Normal distribution
- Poisson distribution