Path Analysis of the Visual Attention Network for fMRI Data

• Activity Number: 187
• Type: Contributed
• Date/Time: Tuesday, August 10, 2004 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistical Computing
• Author(s): Jieun Kim*+ and Wei Zhu and Thomas Ernst
• Companies: SUNY, Stony Brook and SUNY, Stony Brook and Brookhaven National Laboratory
• Keywords: Dynamic Causal Modeling (DCM) ; fMRI ; structural equation modeling (SEM) ; Principal Components Analysis (PCA) ; AR(p) model

Abstract

The ultimate goal for brain connectivity study is to propose, test, modify, and compare certain directional brain pathways. Path analysis or structural equation modeling is ideal for such studies. We report the path analysis of the visual attention network using fMRI data of 28 subjects (14 male,14 female). The path analysis was done in three approaches: (1)The PCA was performed to reduce the multisubject multivariate time series data to the group-level eigen-image, and path analysis was performed on this assuming independent observations. (2) Average the multivariate time series data across subjects and perform path analysis in two stages-- a suitable AR(p) model is determined for each relevant brain region, and the model goodness of fit from path analysis is examined. This approach tend to produce a similar pathway as approach (1). However, the model goodness of fit would increase because the AR(p) procedure reflects the time dependent nature of the data. (3) Finally,the original multisubject, multivariate time series data was used to fit the visual attention pathway. We used Dynamic Causal Modeling (DCM) analysis for this approach, and the comparison with SEM analysis was performed.