Spatially Weighted Reduced-Rank Framework for Group Analysis of Functional Neuroimaging Data

• Activity Number: 347
• Type: Contributed
• Date/Time: Tuesday, August 5, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Imaging
• Author(s): Mihye Ahn*+ and Chao Huang and Haipeng Shen and Hongtu Zhu
• Companies: University of North Carolina at Chapel Hill and University of North Carolina at Chapel Hill and University of North Carolina at Chapel Hill and University of North Carolina at Chapel Hill
• Keywords: resting state fMRI ; functional connectivity ; lasso ; low rank representation ; singular value decomposition

Abstract

In neuroimaging studies, it is challenging to incorporate multiple subjects for group inference due to spatial-temporal functional variation. In this paper, we propose a new modelling framework for analyzing fuctional connectivity pattern across subjects by considering spatial and temporal similarity on whole brain images. To reduce noise sensitivity, we conduct the analysis in the frequency domain, and also impose sparsity on the frequency basis function for better interpretation. From the framework, we also extract the subject-specific spatial factors which enable us group comparison. We discuss optimization strategies to avoid lack of memory in practice. Numerical results show that the spatially weighted framework has lower variability regardless of poor alignment and high inter-subject variability. Finally, we apply the proposed method to the whole brain images from ADHD-200 sample.