The eigenimage filter was used to evaluate the results of a MRI study of cerebral ischemia in a rat model. This linear filter segments a desired feature in an image sequence from other features which may interfere with its observation. The animals were imaged temporally, after occlusion of the middle cerebral artery, to investigate the evolution of the ischemic process. The temporal evolution of ischemia was evaluated by analysis of the "eigenimages," calculated T2 and T1 map images, and images for the angles between signature vectors defined in the eigenimage technique. The eigenimages and angle map images demonstrated an improved visibility of the lesion at all time points, as compared to the original images and T2 and T1 map images. The eigenimages also demonstrated signal intensity changes within the area of ischemia. These changes are speculated to be related to variations in local cerebral blood flow resulting in varying degrees of tissue damage. The eigenimage intensities and the angles between signature vectors demonstrated time-related changes similar to the T2 and T1 values. Since the eigenimage filter and angle calculations are not dependent upon physical models (like T2 and T1), and the errors associated with these models, they may be preferable as methods for tissue characterization.