Abstract
Alzheimer's disease is a neurogenerative disease that alters memories, cognitive functions leading to death. Early diagnosis of the disease, by detection of the preliminary stage, called Mild Cognitive Impairment (MCI), remains a challenging issue. In this respect, we introduce, in this paper, a powerful classification architecture that implements the pre-trained network AlexNet to automatically extract the most prominent features from Magnetic Resonance Imaging (MRI) images in order to detect the Alzheimer's disease at the MCI stage. The proposed method is evaluated using a big database from OASIS Database Brain. Various sections of the brain: frontal, sagittal and axial were used. The proposed method achieved 96.83% accuracy by using 420 subjects: 210 Normal and 210 MRI.