Evaluating Longitudinal Reliability and Cross-Subject Sensitivity of Structural Connectivity Networks Computed Using Probabilistic Fiber Tracking

Alex Smith¹, Madhura Ingalhalikar¹, Ragini Verma¹

This study assesses the reliability of graph metrics extracted from structural networks constructed via probabilistic tractography, and the sensitivity of such networks to individual subjects. Both weighted and binarized networks were evaluated. It was found that subjects could be differentiated based on their weighted structural connectivity patterns, suggesting that such networks can be used in population studies. The intraclass correlation coefficients (ICCs) of the assessed metrics tended to be moderately high across network types, indicating longitudinal robustness across scan sessions. The ICCs tended to be lower in the weighted networks, possibly due to their decreased uniformity versus the binarized versions.