ATLAS TO PATIENT REGISTRATION WITH BRAIN TUMOR BASED ON A NEW MESH-FREE METHOD

Abstract

An atlas is an anatomical representation containing all brain structures well identified in a stereotaxic space from a single subject or a population. Atlases provide information about the organization and localization of the different brain tissues. One may take advantage of this well-organized information for the analysis and processing of a patient’s image by warping the atlas to the patient’s image, and establishing a one-to-one correspondence between the two images. This process is also known as atlas to patient’s image registration and is quite useful for brain tissues segmentation and registering different images to a common reference space. Also, a registered atlas is a model of the patient that can be used for simulation of medical procedures, such as recession and needle insertions. In the presence of brain tumors, the task of atlas to a patient’s image registration become more challenging because tumors deform the brain structures and cause intensity variations in the affected areas, augmenting the dissimilarity between the atlas and patient images. As a consequence, most of the deformable registration based on intensity or shape similarities between images fail in this cases. In order to overcome this issue, some methods involve the use of bio-mechanical models to simulate the tumor’s mass-effect and, in this way, can simulate realistic deformation of the brain structure in the atlas according to the patient’s image. However, these approaches have some weaknesses, mainly related to the assumption of a spherical tumor growth model, the use of Finite Element Method to simulate large deformations, and the computational time that they require. We propose a new approach for atlas to patient’s image registration with tumor based on bio-mechanical deformation of the brain. Contrary to other approaches, our method simulates tumor growth with irregular shape by segmenting from the multi-modal magnetic resonance images of a specific patient. We have developed a totally new mesh free method for the bio-mechanical deformation avoiding the limitations of traditional finite element methods. Experimental results look structurally very similar to the patient’s image and outperform two of the top ranking algorithms.