Detecção e análise de assimetrias estruturais hipocampais em imagens de ressonância magnética aplicadas ao auxílio ao diagnóstico do Alzheimer

Abstract

Alzheimer’s disease (AD) is the main cause of dementia and affects more than 30% of population older than 85 years old. As the elderly population increases, AD has become a relevant public health problem. However, the majority of AD cases are diagnosed at a late stage, even when the patient has already shown severe symptoms of the disease. In this scenario, the importance of the use and development of automated computational tools that aid in the diagnosis of AD is essential. Thus, an automatic hippocampal structural asymmetry detection and analysis technique is proposed in this work. Such technique extracts statistical measurements from hippocampal regions in magnetic resonance (MR) images filtered using a bank of tridimensional (3D) log-Gabor filters, followed by image classification as cognitively normal (CN) and AD through a Support Vector Machine. The bests results achieved represent the average values of AUC, accuracy and F1 being 0.88, 80.43% e 0.75, respectively. Aiming to make the technique available for usage, a microservices-based web application was also proposed, allowing users to interact with the system and request for a MR image to be processed.