Study of 3D MR image registration techniques and probabilistic anatomical atlases for the application in the detection of multiple sclerosis plaques

Abstract

The use of advanced non-invasive medical imaging techniques, both for the detection and treatment of brain diseases, has become more common in the medical field. One example of such a technique is Magnetic Resonance Imaging (MRI), which presents high resolution and good differentiation between tissues. Thanks to this technique, Magnetic Resonance Imaging (MRI) has been successfully used in the diagnosis of brain diseases such as Multiple Sclerosis (MS), Alzheimer and gliomas. Because of the multi-spectral characteristic of the MRI, the combined use of different contrast images (T1, T2-weighted, PD, etc) and anatomic and functional images generates a great number of images that need to be carefully analyzed by radiologists and radiotherapists. Since MR image analysis tends to be very long and error-prone - such as visual fatigue and opinion divergence -, in the last years, a lot of effort has been put into the development of automatic computational systems, in order to help doctors with the task of analyzing such images. A very important step in this kind of system consists of the registration, or alignment, of different images. Such images may come from different subjects or from the same subject, but are obtained with different parameters or in different instances. The alignment of images of a certain patient with the anatomic atlases has been recently used with success in the development of automatic methods for 3D MRI image segmentation. Automatic computational techniques of image registration can, therefore, give a great contribution to the whole process of diagnosing and keeping track of the evolution of diseases. Henceforth, the purpose of this project is to study the most common image registration techniques used in 3D MRI images mainly applied in the detection of MS.(AU)