The diagnostic tool to detect early stages of Alzheimer’s Disease, a progressive neurodegenerative disease, is lacking until today. FDG-PET (Fluorodeoxyglucose-Positron Emission Tomography) shows hypometabolic areas in the brains of pre-demented, i.e. patients suffering from mild cognitive impairment (MCI). The reduced activity may be attributed to disrupted connectivity of the resting-, or default-mode network of the brain [1]. In this contribution, we study the detection of such a network using the framework of blind signal processing, a technique to identify hidden sources within a multivariate mixture using source characteristics such as statistical independence or sparseness. The results are compared to FDG-PET data.