JGrid/FCC: A Tool for Efficient Knowledge Extraction from MRI Scans

MRI (Magnetic Resonance Imaging) allows to display different kinds of soft tissue with highest resolution and has attracted increased interest in the analysis of anatomical differences between normal and pathologic populations in the field of neuroscience. However, to get a deeper insight into complex neurological abnormalities like dementia or somatoform disorder large-scale analysis is indispensable but mostly difficult and time consuming. In this paper, we propose JGrid/FCC, a highly efficient software solution for the analysis of MR images using a combination of several data mining techniques including feature selection, clustering and classification (FCC), in order to provide a concise classification of pathogenous areas in the brain. Using the newly developed distributed computing software JGrid, the time usage of the FCC software was reduced by 80% on a cluster of 10 workstations.