Segmentação e quantificação de tecidos das regiões hipocampais em imagens de ressonância magnética aplicadas ao auxílio ao diagnóstico do Alzheimer

Abstract

Alzheimer’s disease (AD), is a neurodegenerative disease that destroys brain cells, causing losses in cognitive functions and the ability to perform simple tasks daily. AD has a progressive course and its prevalence increases exponentially with age, being 5% in the population with over 60 years and 20% in those over 80 years of age. In AD, generally, the hippocampi, which are brain structures linked to the formation of memories, are the first structures to be damaged. Posteriorly, other parts of the brain are also affected and begin to present atrophy as more neurons die. Magnetic resonance imaging (MRI) has been used clinically with great success in the diagnosis and monitoring of AD as it allows the evaluation of structural changes in regions of the brain related to cognitive functions. The atrophy of the hippocampi, and the consequent reduction in their volumes, have been recognized as an early characteristic of the degenerative process and in DA and it appears to correspond with the onset of memory decline. The main objective of this work was the development of a system that allows the segmentation and quantification of white matter, gray matter, and cerebrospinal fluid present in the hippocampal regions. It is believed that from the quantification of tissues it is possible to evaluate the stage of degeneration of the hippocampal regions and, with that, aid in the diagnosis and follow-up of AD. This work had the objective of developing and providing access through a web platform to a computational technique based on the Gaussian Mixture Model (GMM) and the expectation-maximization algorithm (EM) that performs automatic segmentation tissue in MR images and to evaluate its results quantitatively by comparing the results obtained by the technique developed with those obtained by the state-of-the-art software FMRIB’s Automated Segmentation Tool (FAST), by using metrics like Jaccard, Dice, False Positives and False Negatives.