Classificação de estágios de Alzheimer utilizando aprendizado profundo em imagens de ressonância magnética

Abstract

As life expectancy increases around the world, there is a growth in the prevalence of age-related diseases, such as dementias like Alzheimer’s Disease (AD). Estimates indicate that by 2050, there will be approximately 131.5 million cases of AD, highlighting the need for effective methods of early diagnosis. Traditionally, Alzheimer’s diagnosis involves physical examinations and the evaluation of imaging exams, such as magnetic resonance imaging (MRI), but evaluating these images can be a difficult task. In view of this, recent work demonstrates that deep learning algorithms can assist in classifying the stages of the disease. Thus, this work proposed the implementation and comparison of models based on convolutional neural networks (CNN), applied to magnetic resonance images. In addition, the approach included the use of image preprocessing techniques and comparative analysis of the models to identify the best classification architecture, aiming to contribute to a more accurate and early diagnosis of dementia. At the end of the experiments, the results demonstrated that the selected architecture achieved 96.17% validation accuracy.