Abstract #3683

A Technique to Detect Outliers Automatically in Multi-Site FMRI Data

Andrew Mario Michael^1,2, Stefi A. Baum², Vince D. Calhoun^1,3

¹MIND Research Network, Albuquerque, NM, USA; ²Center for Imaging Science, Rochester Institute of Technology, Rochester, NY, USA; ³ECE, University of New Mexico, Albuquerque, NM, USA

Brain data acquisition from multiple sites is necessitated to increase the biodiversity of subjects and to improve the significance of results. However, it is nearly impossible to keep all the variables involved in data collection identical across sites and this makes outlier detection hard. In fMRI studies outliers are usually identified by the cumbersome and subjective process of visually inspecting individual subject's brain images. In a multisite setting with a large number of subjects this can be a very difficult task. We introduce a simple, easy to implement and efficient technique with minimal human intervention to more accurately detect outliers.