The purpose of Functional Magnetic Resonance Imaging (fMRI) is to map areas of increased neuronal activity of the human brain. fMRI has been applied to investigate a variety of neuronal processes from activities in the primary sensory and motor cortices to cognitive functions such as perception or learning. Robust anisotropic diffusion of statistical parametric maps (RADSPM) is a new technique to improve functional Magnetic Resonance Imaging. RADSPM attempts to improve voxel classification based on robust anisotropic diffusion (RAD) to include the spatial relationship between active voxels. This paper compares two fMRI postprocessing techniques used to identify areas of increased neuronal activity, a widely used method, correlation analysis, and RADSPM. In recent years, the use of ROC analysis has been extended from its original use in communication systems to machine learning, pattern classification and fMRI. We proposed to use ROC curves and the area under the curve (AUC) not only as a final performance evaluation and visualizing technique but as a gauging parameter procedure in RADSPM. We give a brief review of the main methods and conclude presenting experimental results and suggesting further research alternatives.
Workshop de Computación Gráfica, Imágenes y Visualización (WCGIV)
Red de Universidades con Carreras en Informática (RedUNCI)