There is an unmet medical need for identifying neuroimaging biomarkers for Alzheimer's disease (AD), the most common form of senile dementia. These biomarkers are essential for early and accurate diagnosis of AD, monitoring of AD progression, and assessment of AD-modifying therapies. In volumetric studies of the medial temporal lobe and hippocampus, magnetic resonance imaging (MRI), as a technique that can detect changes in cerebral blood flow and blood oxygenation, has shown its powerful ability to distinguish AD and mild cognitive impairment (MCI) subjects from normal controls. However, how to identify potential AD neuroimaging biomarkers from magnetic resonance (MR) images is still a very challenging task. We have thus proposed a novel method based on independent component analysis (ICA), an increasingly important biomedical signal processing technique that enables separation of blindly observed signals into original independent signals for identifying potential AD neuroimaging biomarker(s). The ICA-based method has three steps. First, all MRI scans are aligned and normalized by SPM. Then, ICA was applied to the images for extracting a potential neuroimaging biomarker. Finally, the separated independent component coefficients were fed into a classifying machine that is able to discrinate AD and MCI from control subjects. The experimental results on the MRI data from the Open Access Series of Imaging Studies showed that that our ICA-based method can discern AD and MCI cases from agematched controls.