AUTOMATIC DETECTION OF WHITE MATTER LESIONS IN MAGNETIC-RESONANCE BRAIN IMAGES

A new approach to automating radiologic diagnosis is described and tested in a system that locates multiple sclerosis lesions in magnetic resonance human brain images. This approach uses a step-by-step procedure, where the most obvious features in the images are identified first, and used to calibrate the application of the next step, until the desired features are identified. The approach stresses testing on a large number of images of the same type, and exploiting the special characteristics of this type of images. The system was designed with the variations of the various image parameters in mind, to ensure its reliability. New low-level methods of high reliability have been developed for segmenting images. A new geometric modeling method which tolerates the variability of biological structures is used to encode anatomic knowledge. The system and its components have been tested on 1132 images from 17 patients, with very good results. © 1990.