A MR brain image study consists of several image series obtained at different contrast and resolution. The study can however be summarized by a few relevant slices that document the patient condition effectively. Summarization is equivalent to a clinical compression and can be used in teaching files, for efficient integration into an electronic medical record, and for referring physicians. Automated image summarization is based on a novel combination of text and image processing techniques: (i) structured text related to patient presenting symptom and/or reason for scan is processed using expert rules to infer the differential diagnosis, relevant structure (defined as the location of the abnormality), optimum orientation/contrast of the images, and (ii) automated image registration of the patient study to a labeled customizable atlas to locate structures within the patient images. The MR atlas is customizable for contrast and intensity to match the current patient data to improve the accuracy of registration. For a given brain MR image study and patient presentation, the module will automatically identify the relevant slices showing the abnormality.