Extraction of Discriminative Functional MRI Activation Patterns and an Application to Alzheimer's Disease

We propose a novel Dynamic Recursive Partitioning approach for discovering discriminative patterns of functional MRI activation. The goal is to efficiently identify spatial regions that are associated with non-spatial variables through adaptive recursive partitioning of the 3D space into a number of hyperrectangles utilizing statistical tests. As a case study, we analyze fMRI datasets obtained from a study that explores neuroanatomical correlates of semantic processing in Alzheimer’s disease. We seek to discover brain activation areas that discriminate controls from patients. We evaluate the results by presenting classification experiments that utilize information extracted from these regions. The discovered areas elucidated large hemispheric and lobar differences being consistent with prior findings. The overall classification accuracy based on activation patterns in these areas exceeded 90%. The proposed approach being general enough has great potential for elucidating structure-function relationships and can be valuable to human brain mapping.