Alzheimer’s disease (AD), the most common cause of dementia in the elderly, is a gradually progressive degenerative neurological disorder that is characterized by increasing cognitive impairment, characteristic degenerative pathology and brain atrophy. Studies have shown that the progression of AD pathology in the brain develops in a predictable pattern and the pathological changes that take place in brain begin at the microscopic level long before the first signs of memory loss. Structural Magnetic Resonance Imaging (MRI), which has exceptional soft tissue contrast and detailed resolution, is the best way to noninvasively examine changes which occur early in the course of AD. For this dissertation, our aim is to improve the methods for measuring the atrophy of brain structures in AD, as seen on MRI, and to apply these methods to subjects with cognitive impairment. This study has established a new coordinate template to replace the widely used Montreal Neurological Institute (MNI) template for the atlas-based segmentation procedure. The new template was derived from the same structural image as the one used by the Automated Anatomical Labeling (AAL) procedure. The agreement of the newly developed coordinate template and AAL helps to estimate accurate spatial transformation parameters used in warping the AAL to individual subject images. The new template combines the spatial information of the structural image and the frequency information of MNI template. Based on the same principle, a set of customized templates has been developed. The customized template, associated atlas and customized priors match more closely the aging population than the previous template, so as to improve the atlas-based segmentation of regions of interest in AD assessment. Visual Rating System (VRS) of a single coronal slice (MB slice) in MRI has been another valuable method in the assessment of medial temporal lobe atrophy. An automated procedure has been developed in this study to measure the hippocampal area on the same coronal slice so that the labor of human experts in the VRS assessment of hippocampus will be significantly reduced. Finally the methods and materials (template and atlas) developed in this dissertation were applied to cross-sectional studies of subjects with cognitive impairment. We conducted volumetric analysis on subjects and conclude that the data from the new approaches have higher correlations with clinical data, and therefore can be reliably used as part of an AD assessment tool.