Cluster-Span Threshold: An unbiased threshold for binarising weighted complete networks in functional connectivity analysis

TL;DR

The paper introduces the Cluster-Span Threshold (CST), an unbiased method for binarizing weighted networks in functional connectivity analysis, which balances clustering and spanning properties to improve network topology assessment.

Contribution

It proposes a novel thresholding technique based on clustering coefficient balance, moving away from arbitrary fixed-density thresholds in network analysis.

Findings

CST effectively distinguishes task-related differences in EEG data.

CST provides a sensitive and objective thresholding method.

Compared to other thresholds, CST enhances functional connectivity analysis.

Abstract

We propose a new unbiased threshold for network analysis named the Cluster-Span Threshold (CST). This is based on the clustering coefficient, C, following logic that a balance of `clustering' to `spanning' triples results in a useful topology for network analysis and that the product of complementing properties has a unique value only when perfectly balanced. We threshold networks by fixing C at this balanced value, rather than fixing connection density at an arbitrary value, as has been the trend. We compare results from an electroencephalogram data set of volunteers performing visual short term memory tasks of the CST alongside other thresholds, including maximum spanning trees. We find that the CST holds as a sensitive threshold for distinguishing differences in the functional connectivity between tasks. This provides a sensitive and objective method for setting a threshold on weighted complete networks which may prove influential on the future of functional connectivity research.