547 – Statistical Challenges and Solutions for Neuroimaging Studies
Image Registration Techniques Alter Image Properties in fMRI
Kevin K. Liu
Marquette University
Daniel B. Rowe
Marquette University
Functional magnetic resonance imaging (fMRI) can find active parts of the brain by observing changes in blood oxygenation. After collecting complex valued data, reconstruction is necessary to obtain an image. Often times it becomes preferable to perform image processing in order to either enhance the overall quality or a particular aspect of an image. Registration is an image processing operation that aligns two or more images through geometric transformations. Important registration operations commonly used in fMRI include motion correction and spatial normalization. This is a broad topic consisting of many approaches that vary in accuracy, computational efficiency, and complexity. It has been shown in recent studies that some image processing operations can induce artificial correlation of non-biological origins between voxels. Quantifying any potentially induced correlation from image registration can allow for future development of statistical models that greatly improve the accuracy and reliability of fMRI studies.