|
Abstract:
|
Functional connectivity (FC) plays a central role in understanding the basis of neurological and psychiatric disorders. FC also holds promise for predicting clinical outcomes, which stands to have strong translational impact. These objectives rely heavily on proper characterization of FC. The issue of proper characterization involves at least two key components. First, it is critical to consider the most effective portrayal of FC to answer the research question. Examples include decisions to estimate static or time-varying FC (correlation matrix) and to determine the appropriate spatial scale. Secondly, it is essential to quantify accurate estimates of FC. There is little guidance suggesting the best approach for a given analysis. In this talk, we present results from an analysis conducted to determine FC features that represent predictors of Parkinson's disease (PD). We evaluate various approaches to characterizing FC, compare the associated performances for prediction, and discuss the relative merits of each approach in context of our PD analysis. Overall, we achieve high prediction accuracy to separate PD patients from healthy controls with a parsimonious set of connect
|
