Shrinkage Approach for Imaging Connectivity Analysis

• 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): Haochang Shou*+ and Ani Eloyan and Ciprian Crainiceanu
• Companies: Johns Hopkins Bloomberg School of Public Health and Johns Hopkins University and Johns Hopkins University
• Keywords: Shrinkage estimator ; measurement error ; connectivity map ; resting state fMRI

Abstract

Imaging data are observed with measurement errors that come from multiple sources. In addition to the voxel-wise white noises, there can also be systematic errors that are spatially correlated. Simply taking average over the repeated scans is not enough to reduce the effect of systematic errors. Moreover, often times there is only one scan available for a particular subject. We propose a range of estimators which shrink the subject-specific image towards population average image on the levels of individual voxels, local neighbors and the whole image. The estimators assess the signal-to-noise ratio with the information of the population image, yet preserve the subject-specific features. We have evaluated our methods on seed-based connectivity maps that are calculated using resting-state functional MRI from 21 healthy volunteers (publicly known as 'Kirby21' dataset). Our results have shown that we achieve an average of 30% improvement in mean square errors for prediction, compared to using one replicate only.