Tuning FCMP to elicit novel time course signatures in fMRI neural activation studies

Abstract

Functional magnetic resonance imaging (fMRI) is a preferred imaging modality to infer in vivo organ function from blood flow intensities. FMRI analysis is complex due to the variety of hemodynamic response models and the presence of noise. This complexity drives the use of exploratory data analysis (EDA) to elicit intrinsic data structure. This work demonstrates the utility of a fuzzy C-means (FCM) variant that incorporates feature partitions to generalize distance metrics across spatio-temporal features. This method, FCMP, exploits this relation to generate both novel and robust data inferences. A synthetic and a hybrid fMRI dataset are examined with results compared to an industry benchmark, EvIdent®. Efficacy of FCMP is shown in terms of tunable sensitivity to novel time course signatures and adaptability with which specific signatures are integrated into the objective function. © 2007 Springer-Verlag Berlin Heidelberg.