Comparison of structural connectivity metrics for multimodal brain image analysis

Multimodal brain image analysis is gaining popularity as exemplified in the recent surge of techniques dedicated to fusing structural and functional connectivity information. The performance of such endeavors relies on the metrics used to quantify connection strength. Compared to functional connectivity (FC) metrics, structural connectivity (SC) metrics received less scrutiny by the neuroimaging community despite being widely utilized in the literature. In this paper, we analyze the performance of commonly used SC metrics. Specifically, we analyze the relationship between SC and FC during resting-state and different tasks with the assumption of an inherent dependence between brain structure and function. Among the tested metrics, we show that parcel volume-normalized fiber count correlates best with FC, and that total fiber length has the least bias in favor of shorter distances between brain regions. We also show relatively consistent SC-FC correlation across tasks, supporting the notion that SC constitutes the backbone of brain connectivity facilitating a diverse repertoire of functional connectivity patterns.