Automatic Detection of Multiple Sclerosis Lesions in MR Brain Images

Abstract This paper describes a system that locates lesions in Magnetic Resonance (MR) human brain images. The system uses new low level vision methods which successively identify the brain mass in the images, locate suspected lesions, sulci, and other normal structures. Other low level methods eliminate the majority of false positive lesions and locate certain landmarks such as the interhemispherical fissure. These methods take advantage of the special characteristics of MR images. A modeling method that employs b-spline surfaces has been developed to model the surfaces of the organs in a human brain in 3D. This method allows the model surfaces to be deformed in order to fit each individual patient's brain, and also allows the proportional deformation of the shapes of difficult-to-identify organs according to the deformation of easier-to-identify organs. The system calculates the appropriate position and orientation for the model by making use of landmarks within the patient's brain, and the moment of inertia method. The system uses both Proton Density and T2 sequences of images, in coronal and axial orientations. The various parts of the system have been tested extensively (on more than 1000 images from patients with Multiple Sclerosis lesions) with very good results. The methods developed here can also be used for other diagnostic tasks in radiology.